Unmanned Aerial Vehicles (UAV)

and Drones
Unmanned Aerial Vehicles
(UAV) and Drones

Edited by:
Zoran Gacovski

ARCLER
P r e s s

www.arclerpress.com
Unmanned Aerial Vehicles (UAV) and Drones
Zoran Gacovski

Arcler Press
224 Shoreacres Road
Burlington, ON L7L 2H2
Canada
www.arclerpress.com
Email: [email protected]

e-book Edition 2021

ISBN: 978-1-77407-976-8 (e-book)

This book contains information obtained from highly regarded resources. Reprinted
material sources are indicated. Copyright for individual articles remains with the au-
thors as indicated and published under Creative Commons License. A Wide variety of
references are listed. Reasonable efforts have been made to publish reliable data and
views articulated in the chapters are those of the individual contributors, and not neces-
sarily those of the editors or publishers. Editors or publishers are not responsible for
the accuracy of the information in the published chapters or consequences of their use.
The publisher assumes no responsibility for any damage or grievance to the persons or
property arising out of the use of any materials, instructions, methods or thoughts in the
book. The editors and the publisher have attempted to trace the copyright holders of all
material reproduced in this publication and apologize to copyright holders if permission
has not been obtained. If any copyright holder has not been acknowledged, please write
to us so we may rectify.

Notice: Registered trademark of products or corporate names are used only for explana-
tion and identification without intent of infringement.

© 2021 Arcler Press

ISBN: 978-1-77407-773-3 (Hardcover)

Arcler Press publishes wide variety of books and eBooks. For more information about
Arcler Press and its products, visit our website at www.arclerpress.com
DECLARATION

Some content or chapters in this book are open access copyright free
published research work, which is published under Creative Commons
License and are indicated with the citation. We are thankful to the
publishers and authors of the content and chapters as without them this
book wouldn’t have been possible.
 ABOUT THE EDITOR

Dr. Zoran Gacovski has earned his PhD degree at Faculty of Electrical
engineering, Skopje. His research interests include Intelligent systems and
Software engineering, fuzzy systems, graphical models (Petri, Neural and
Bayesian networks), and IT security. He has published over 50 journal and
conference papers, and he has been reviewer of renowned Journals. Currently,
he is a professor in Computer Engineering at European University, Skopje,
Macedonia.
 TABLE OF CONTENTS

List of Contributors........................................................................................xv
List of Abbreviations................................................................................... xxiii
Preface.................................................................................................... ....xxv

Section 1 Design and Development of UAV-s

Chapter 1 Design and Development of a Fly-by-Wireless UAV Platform.................... 3

Introduction................................................................................................ 3
Wireless Network Technologies.................................................................. 5
Global Electronics Architecture................................................................... 7
Onboard Wireless System........................................................................... 8
Experimental Results................................................................................. 13
Conclusion............................................................................................... 15
References................................................................................................ 17

Chapter 2 A Review of Control Algorithms for Autonomous Quadrotors................. 19

Abstract ................................................................................................... 19
Introduction ............................................................................................. 20
Mathematical Model ................................................................................ 21
Survey of Control Algorithms ................................................................... 22
Comparison of Control Algorithms ........................................................... 30
Discussion and Conclusion ...................................................................... 30
References ............................................................................................... 32

Chapter 3 Central Command Architecture for High-Order Autonomous

Unmanned Aerial Systems........................................................................ 37
Abstract.................................................................................................... 37
Introduction.............................................................................................. 38
Method..................................................................................................... 40
 Results...................................................................................................... 46
Summary And Conclusions....................................................................... 51
Appendix: The Shortest Lines Between Two Point Sets Do Not Cross......... 52
References................................................................................................ 56

Chapter 4 Quadcopter Design for Payload Delivery................................................. 59

Abstract ................................................................................................... 59
Introduction ............................................................................................. 60
Methodology ........................................................................................... 61
Simulation................................................................................................ 67
Simulation Results And Discussion........................................................... 70
Conclusion And Future Work.................................................................... 72
Conflicts Of Interest.................................................................................. 74
References................................................................................................ 75

Section 2 Trajectory Control Techniques

Chapter 5 Advanced UAV Trajectory Generation: Planning and Guidance.............. 79

Introduction ............................................................................................. 79
Motion-Planning Methodologies .............................................................. 81
Trajectory Generation And Guidance Module - Tg2m ............................... 86
Tg2m Simulation .................................................................................... 105
Final Observations ................................................................................. 109
References ............................................................................................. 110

Chapter 6 Modelling and Control Prototyping of Unmanned Helicopters.............. 113

Introduction ........................................................................................... 113
Helicopters Flight Principles .................................................................. 115
Model Description ................................................................................. 117
Identification Results .............................................................................. 127
Control Prototyping................................................................................ 130
Control Results ...................................................................................... 136
Final Observations And Future Work ..................................................... 138
References ............................................................................................. 140

x
Chapter 7 Contour Based Path Planning with B-Spline Trajectory Generation for
Unmanned Aerial Vehicles (UAVs) over Hostile Terrain........................ 141
Abstract ................................................................................................. 141
Introduction ........................................................................................... 142
Modelling............................................................................................... 144
Problem Formulation ............................................................................. 149
Generation Of Nodes ............................................................................ 151
Implementation ..................................................................................... 151
Conclusions And Future Works .............................................................. 158
Acknowledgements ............................................................................... 158
References.............................................................................................. 159

Chapter 8 Trajectory Tracking of Quadrotor Aerial Robot Using Improved

Dynamic Inversion Method.................................................................... 163
Abstract.................................................................................................. 163
Introduction............................................................................................ 164
Quadrotor Mechanism............................................................................ 165
Dynamic Model of Quadrotor................................................................ 166
Improved Dynamic Inverse Controller.................................................... 169
Simulation Results.................................................................................. 172
Conclusion............................................................................................. 173
References.............................................................................................. 175

Section 3 Quadrotors and Small-size Aerial Vehicles

Chapter 9 Neural Network Control and Wireless Sensor Network-based

Localization of Quadrotor UAV Formations.......................................... 179
Introduction ........................................................................................... 179
Background ........................................................................................... 181
Leader-Follower Formation Tracking Control .......................................... 184
Optimized Energy-Delay Sub-Network Routing (OEDSR)
Protocol For Uav Localization And Discovery .............................. 200
Simulation Results ................................................................................. 205
Conclusions ........................................................................................... 207
References ............................................................................................. 208

xi
Chapter 10 Dubins Waypoint, Navigation of
Small-Class Unmanned Aerial Vehicles.................................................. 209
Abstract.................................................................................................. 209
Introduction............................................................................................ 210
Method................................................................................................... 212
Results.................................................................................................... 220
Conclusion............................................................................................. 223
Acknowledgements................................................................................ 224
Nomenclature......................................................................................... 224
Conflicts Of Interest................................................................................ 225
References.............................................................................................. 226

Chapter 11 Modelling Oil-Spill Detection with Swarm Drones................................ 229

Abstract.................................................................................................. 229
Introduction............................................................................................ 230
Pollutant Dispersion Model.................................................................... 231
Swarm Design........................................................................................ 234
Experimentation...................................................................................... 237
Macroscopic Model................................................................................ 244
Discussion.............................................................................................. 250
Conflict Of Interests................................................................................ 252
Acknowledgment.................................................................................... 252
References.............................................................................................. 253

Chapter 12 A Survey of Modelling and Identification of Quadrotor Robot.............. 255

Abstract.................................................................................................. 255
Introduction............................................................................................ 256
Characteristics Of Quadrotor.................................................................. 258
6-Dof Airframe Dynamics....................................................................... 260
Basic Dynamic Model Of A Quadrotor................................................... 263
Aerodynamic Effects............................................................................... 270
Identification Of A Quadrotor Model...................................................... 275
Conclusion And Perspective.................................................................... 284
Conflict Of Interests................................................................................ 285
Acknowledgments.................................................................................. 285
References.............................................................................................. 286

xii
Chapter 13 Visual Flight Control of a Quadrotor
Using Bioinspired Motion Detector........................................................ 295
Abstract.................................................................................................. 296
Introduction............................................................................................ 296
Bioinspired Image Processing................................................................. 298
Multicamera 3D Pose Estimation............................................................ 300
Controller............................................................................................... 302
Experiments And Results......................................................................... 304
Conclusions And Future Works............................................................... 310
Acknowledgments.................................................................................. 310
References.............................................................................................. 311

Section 4 Applications of UAV-s and Drones

Chapter 14 Development of Rescue Material

Transport UAV (Unmanned Aerial Vehicle)........................................... 315
Abstract.................................................................................................. 315
Introduction............................................................................................ 316
Development Of Software And Hardware............................................... 317
Research And Development Results........................................................ 321
Conclusion............................................................................................. 324
Acknowledgements................................................................................ 324
Conflicts Of Interest................................................................................ 324
References.............................................................................................. 325

Chapter 15 Railway Transport Infrastructure Monitoring by UAVs and Satellites.... 327

Abstract.................................................................................................. 327
Introduction............................................................................................ 328
Advantages And Disadvantages Of Remotely Piloted Aircrafts................. 329
Analysis Of Satellite Resources For Obtaining Information...................... 331
The Databases Of Satellite Information................................................... 332
Comparison Of Information Received From Different Sources................ 333
Reconstruction Of 3D Models Of The Railway Infrastructure
Objects From A Set Of 2D Images................................................. 336
Conclusion............................................................................................. 339
Acknowledgements................................................................................ 340
Conflicts Of Interest................................................................................ 340

xiii
 References.............................................................................................. 341

Chapter 16 Assessment of UAV Based Vegetation Indices for Nitrogen

Concentration Estimation in Spring Wheat............................................ 343
Abstract.................................................................................................. 344
Introduction............................................................................................ 345
Materials And Methods........................................................................... 348
Results And Discussions......................................................................... 353
Conclusions............................................................................................ 362
Acknowledgements................................................................................ 362
Conflicts Of Interest................................................................................ 362
References.............................................................................................. 363

Chapter 17 Research and Teaching Applications of Remote Sensing

Integrated with GIS: Examples from the Field........................................ 369
Abstract.................................................................................................. 369
Introduction............................................................................................ 370
Wildlife Monitoring On Catalina Island, Pilot 2018................................ 372
Autonomous Terrain Modeling................................................................ 377
Integrating Remote Sensing And Uas Into Curriculum............................. 379
Discussion.............................................................................................. 382
Conclusion............................................................................................. 383
Acknowledgements................................................................................ 383
Conflicts Of Interest................................................................................ 383
References ............................................................................................. 384

Chapter 18 Development of Drone Cargo Bay with Real-Time

Temperature Control.............................................................................. 385
Abstract.................................................................................................. 385
Introduction............................................................................................ 386
Related Works........................................................................................ 387
Configuration Of Temperature Control System........................................ 387
Hardware Design.................................................................................... 388
Software Design...................................................................................... 390
Prototype Design.................................................................................... 392
Conclusion............................................................................................. 394

xiv
Acknowledgements................................................................................ 395
Conflicts of Interest................................................................................. 395
References.............................................................................................. 396

Index...................................................................................................... 399

xv
 LIST OF CONTRIBUTORS

Paulo Carvalhal
University of Minho Portugal

Cristina Santos
University of Minho Portugal

Manuel Ferreira
University of Minho Portugal

Luís Silva
University of Minho Portugal

José Afonso
University of Minho Portugal

Andrew Zulu
Department of Mechanical and Marine Engineering, Polytechnic of Namibia, Windhoek,
Namibia

Samuel John
Department of Mechanical and Marine Engineering, Polytechnic of Namibia, Windhoek,
Namibia

Larry M. Silverberg
Mechanical and Aerospace Engineering, North Carolina State University, Raleigh, USA

Chad Bieber
Mechanical and Aerospace Engineering, North Carolina State University, Raleigh, USA

Gordon Ononiwu
Department of Electrical/Electronic Engineering, Federal University of Technology,
Owerri, Nigeria

Ondoma Onojo
Department of Electrical/Electronic Engineering, Federal University of Technology,
Owerri, Nigeria

xvii
Oliver Ozioko
Department of Electrical/Electronic Engineering, Federal University of Technology,
Owerri, Nigeria

Onyebuchi Nosiri
Department of Electrical/Electronic Engineering, Federal University of Technology,
Owerri, Nigeria

Antonio Barrientos
Universidad Politécnica de Madrid – (Robotics and Cybernetics Group), Spain

Pedro Gutiérrez
Universidad Politécnica de Madrid – (Robotics and Cybernetics Group), Spain

Julián Colorado
Universidad Politécnica de Madrid – (Robotics and Cybernetics Group), Spain

Jaime del-Cerro
Universidad Politécnica de Madrid – Robotics and Cybernetics Group, Spain

Alexander Martínez
Universidad Politécnica de Madrid – Robotics and Cybernetics Group, Spain

Ee-May Kan
Nanyang Technological University, Singapore City, Singapore

Meng-Hiot Lim
Nanyang Technological University, Singapore City, Singapore

Swee-Ping Yeo
National University of Singapore, Singapore City, Singapore

Jiun-Sien Ho
Temasek Polytechnic, Singapore City, Singapore

Zhenhai Shao
University of Electronic Science Technology of China, Chengdu, China.

Lifeng Wang
Field Bus Technology & Automation Lab, North China University of Technology,
Beijing, China

xviii
Yichong He
Field Bus Technology & Automation Lab, North China University of Technology,
Beijing, China

Zhixiang Zhang
Field Bus Technology & Automation Lab, North China University of Technology,
Beijing, China

Congkui He
Field Bus Technology & Automation Lab, North China University of Technology,
Beijing, China

Travis Dierks
Missouri University of Science and Technology, United States of America

S. Jagannathan
Missouri University of Science and Technology, United States of America

Dahan Xu
Mechanical and Aerospace Engineering, North Carolina State University, Raleigh, USA

F. Aznar
Department of Computer Science and Artificial Intelligence, University of Alicante,
03090 Alicante, Spain

M. Sempere
Department of Computer Science and Artificial Intelligence, University of Alicante,
03090 Alicante, Spain

M. Pujol
Department of Computer Science and Artificial Intelligence, University of Alicante,
03090 Alicante, Spain

R. Rizo
Department of Computer Science and Artificial Intelligence, University of Alicante,
03090 Alicante, Spain

M. J. Pujol
Department of Applied Mathematics, University of Alicante, 03090 Alicante, Spain

Xiaodong Zhang
School of Automation, Shenyang Aerospace University, Shenyang 110136, China

xix
Xiaoli Li
School of Automation and Electrical Engineering, University of Science and Technology
Beijing, Beijing 100083, China

Kang Wang
School of Automation and Electrical Engineering, University of Science and Technology
Beijing, Beijing 100083, China

Yanjun Lu
School of Automation, Shenyang Aerospace University, Shenyang 110136, China

Lei Zhang
Institute of Automatic Control Engineering (LSR), Technische Universität München,
80290 München, Germany

Tianguang Zhang
Institute of Automatic Control Engineering (LSR), Technische Universität München,
80290 München, Germany

Haiyan Wu
Institute of Automatic Control Engineering (LSR), Technische Universität München,
80290 München, Germany

Alexander Borst
Department of Systems and Computational Neurobiology, Max Planck Institute of
Neurobiology, Am Klopferspitz 18, D-82152 Martinsried, Germany

Kolja Kühnlenz
Institute of Automatic Control Engineering (LSR), Technische Universität München,
80290 München, Germany
Institute for Advanced Study (IAS), Technische Universität München, 80290 München,
Germany

Daeil Jo
Department of Industrial Engineering, College of Engineering, Ajou University, Suwon,
South Korea

Sergey I. Ivashov
Remote Sensing Laboratory, Bauman Moscow State Technical University, Moscow,
Russia

Alexander B. Tataraidze
Remote Sensing Laboratory, Bauman Moscow State Technical University, Moscow,
Russia

xx
Vladimir V. Razevig
Remote Sensing Laboratory, Bauman Moscow State Technical University, Moscow,
Russia

Eugenia S. Smirnova
Remote Sensing Laboratory, Bauman Moscow State Technical University, Moscow,
Russia

Olga S. Walsh
Department of Plant Sciences, Southwest Research and Extension Center, University of
Idaho, Moscow, ID, USA.

Sanaz Shafian
Department of Plant Sciences, Southwest Research and Extension Center, University of
Idaho, Moscow, ID, USA.

Juliet M. Marshall
Department of Entomology, Plant Pathology, and Nematology, Idaho Falls Research
and Extension Center, University of Idaho, Idaho Falls, ID, USA.

Chad Jackson
Department of Entomology, Plant Pathology, and Nematology, Idaho Falls Research
and Extension Center, University of Idaho, Idaho Falls, ID, USA.

Jordan R. McClintick-Chess
Department of Plant Sciences, Southwest Research and Extension Center, University of
Idaho, Moscow, ID, USA.

Steven M. Blanscet
BASF-Chemical Co., Caldwell, ID, USA.

Kristin Swoboda
Take Flight UAS, LLC, Boise, ID, USA.

Craig Thompson
Take Flight UAS, LLC, Boise, ID, USA.

Kelli M. Belmont
Seminis, Payette, ID, USA.

Willow L. Walsh
Vallivue High School, Caldwell, ID, USA.

xxi
Laura C. Loyola
Spatial Sciences Institute, University of Southern California, Los Angeles, CA, USA.

Jason T. Knowles
Spatial Sciences Institute, University of Southern California, Los Angeles, CA, USA.
GeoAcuity, Los Angeles, CA, USA.

Andrew J. Marx
Spatial Sciences Institute, University of Southern California, Los Angeles, CA, USA.

Ryan McAlinden
Spatial Sciences Institute, University of Southern California, Los Angeles, CA, USA.
Institute for Creative Technologies, University of Southern California, Los Angeles,
CA, USA.

Steven D. Fleming
Spatial Sciences Institute, University of Southern California, Los Angeles, CA, USA.

Sedam Lee
Department of Industrial Engineering, College of Engineering, Ajou University, Suwon,
South Korea

Yongjin Kwon
Department of Industrial Engineering, College of Engineering, Ajou University, Suwon,
South Korea

xxii
 LIST OF ABBREVIATIONS

AP Access Point
AVCL Aerial Vehicle Control Language
AOA Angle of Attack
AED Automatic Cardioverter Defibrillator
AUAS Autonomous Unmanned Aerial Systems
BET Blade Element Theory
CRP Calibrated Reflectance Panel
CIC Catalina Island Conservancy
COA Civil Certificate of Waiver Authorization
CW Clockwise
CHs Cluster Heads
COTS Commercial off the Shelf
CASI Compact Airborne Spectrographic Imager
DLS Downwelling Light Sensor
DWN Dubins Waypoint Navigation
ERS Earth Remote Sensing
ESC Electronic Speed Controllers
EMD Elementary Motion Detector
ERR Experimental Railway Ring
FAA Federal Aviation Administration
FSM Finite State Machine
FL Formation Leader
GIST Geographic Information Sciences & Technology
GIS Geographic Information System
GPS Global Positioning System receivers
GUI Graphical User Interface
GNDVI Green Normalized Difference Vegetation Index
GTS Guaranteed Time Slot
IMU Inertial Measurement Units
ICT Institute for Creative Technologies
ICE Internet Communications Engine
IFT Iterative Feedback Tuning
ILC Iterative Learning Control
LQR Linear Quadratic Regulator
MM Master Node
MAVs Microaerial Vehicles
MP Mission Planner
MFAC Model Free Adaptive Control
MT Momentum Theory
NN Neural Networks
NDVI Normalized Difference Vegetation Index
OFN Obstacle Field Navigation
OWT One World Terrain
PLR Packet Loss Ratio
PLS Partial Least Squares
PCA Principal Component Analysis
PID Proportional Integral Derivative
PWM Pulse Width Modulation
RCS Radar Cross Section
RN Relay Nodes
RPL Remote License Pilots
RMSE Root Mean Squared Error
SMC Sliding Mode Control
SSI Spatial Sciences Institute
TIFF Tag Image File Format
UAS Unmanned Aerial Systems
VIs Vegetation Indices
VTOL Vertical Take-Off And Landing
VM Virtual Machine
VR/AR Virtual Reality/Augmented Reality
VRFT Virtual Reference Feedback Tuning
WN Waypoint Navigation
 PREFACE

Drones are defined as unmanned aerial vehicles (UAVs) that can fly over
space from a few tens of meters - up to thousands of kilometers, or small
(micro and nano) quadrotors that can move indoors. Today – there is a
rising need of these types of flying vehicles - for civil and for military
applications. There is also significant interest for development of new
types of drones that can fly autonomously – around different locations
and environments, and will accomplish different technical missions
(agriculture, public events – sports/ concerts, space research etc.).
Within the past decade – wide spectra of applications have emerged that
resulted in many variations in the size and the design of the UAV-s (drones).
The year of 2015 was declared as a year that exponentially boosted the
drone applications in many human-oriented fields and environments,
especially in agriculture and forestry (75% of all applications).
The drones categories differ among themselves by their configuration,
platform or application. There are 3 major classes defined in the literature,
and the class I has 4 subclasses (a, b, c, d):
Class I-a) – Nano drones, weight < 200 g.
Class I–b) – Micro drones, weight between 200g – 2 kg.
Class I–c) – Mini drones, weight between 2 kg – 20 kg.
Class I–d) – Small drones, weight between 20 kg – 150 kg.
Class II – Tactical drones, weight between 150-600 kg.
Class III – Strike/ Hale drones, weight over 600 kg.
This edition covers different topics from UAV-s and drones – such as
design and development of UAV-s, trajectory control techniques, small
drones and quadrotors, and application scenarios of UAV-s and drones.
Section 1 focuses on design and development of UAV-s, describing design
and development of a fly-by-wireless UAV platform, review of control
algorithms for autonomous quadrotors, central command architecture for
high-order autonomous unmanned aerial systems, and quadcopter design
for payload delivery.
Section 2 focuses on trajectory control techniques, describing
development of rescue material transport UAV, railway transport
infrastructure monitoring by UAVs and satellites, assessment of UAV
based vegetation indices for nitrogen concentration estimation in spring
wheat, research and teaching applications of remote sensing integrated
with GIS: examples from the field, and development of drone cargo bay
with real-time temperature control.
Section 3 focuses on small drones and quadrotors, describing neural
network control and wireless sensor network-based localization of
quadrotor UAV formations, Dubin’s waypoint, navigation of small-class
unmanned aerial vehicles, modelling oil-spill detection with swarm
drones, a survey of modelling and identification of quadrotor robot, and
visual flight control of a quadrotor using bioinspired motion detector.
Section 4 focuses on application scenarios of UAV-s and drones,
describing .

xxvi
 SECTION 1
DESIGN AND
DEVELOPMENT OF UAV-S
 CHAPTER

1
Design and Development of a
Fly-by-Wireless UAV Platform

Paulo Carvalhal, Cristina Santos, Manuel Ferreira, Luís Silva and

José Afonso
University of Minho Portugal

INTRODUCTION
The development of unmanned aerial vehicles (UAVs) has become an active
area of research in recent years, and very interesting devices have been
developed. UAVs are important instruments for numerous applications, such
as forest surveillance and fire detection, coastal and economic exclusive
zone surveillance, detection of watershed pollution and military missions.

Citation: Paulo Carvalhal, Cristina Santos, Manuel Ferreira, Luis Silva and Jose Afon-
so (January 1st 2009). “Design and Development of a Fly-by-Wireless UAV Platform”,
Aerial Vehicles, Thanh Mung Lam, IntechOpen, DOI: 10.5772/6464.
Copyright: © 2009 The Author(s). Licensee IntechOpen. This chapter is distributed
under the terms of the Creative Commons Attribution-NonCommercial-ShareAlike-3.0
License
4 Unmanned Aerial Vehicles (UAV) and Drones

The work described in this chapter is part of a larger project, named AIVA,
which involves the design and development of an aerial platform, as well
as the instrumentation, communications, flight control and artificial vision
systems, in order to provide autonomous takeoff, flight mission and landing
maneuvers. The focus of the chapter is on one of the main innovative
aspects of the project: the onboard wireless distributed data acquisition and
control system. Traditionally, UAVs present an architecture consisting of
one centralized and complex unit, with one or more CPUs, to which the
instrumentation devices are connected by wires. At the same time, they
have bulky mechanical connections. In the approach presented here, dubbed
“fly-by-wireless”, the traditional monolithic processing unit is replaced by
several less complex units (wireless nodes), spread out over the aircraft.
In that way, the nodes are placed near the sensors and controlled surfaces,
creating a network of nodes with the capacity of data acquisition, processing
and actuation.
This proposed fly-by-wireless platform provides several advantages
over conventional systems, such as higher flexibility and modularity, as well
as easier installation procedures, due to the elimination of connecting cables.
However, it also introduces several challenges. The wireless network that
supports the onboard distributed data acquisition and control system needs
to satisfy demanding requirements in terms of quality of service (QoS), such
as sustainable throughput, bounded delay and reliable packet delivery. At
the same time, it is necessary to guarantee that the power consumption of the
battery powered wireless nodes is small, in order to increase the autonomy of
the system. Currently there are many different wireless network technologies
available in the market. Section 2 presents an overview of the most relevant
technologies and discusses their suitability to meet the above requirements.
Based on this analysis, we chose the Bluetooth wireless network technology
as the basis for the design and development of a prototype of the fly-by-
wireless system. The system was implemented using commercial off-the-
shelf components, in order to provide a good trade-off between development
costs, reliability and performance. Some other objectives were also pursued
in the development of the system, namely the design of a framework
where communication between nodes is effective and independent of the
technology adopted, the development of a design approach to model the
embedded system and the development of an application oriented operating
system with a modular structure.
The following sections are organized as follows: section 2 presents an
overview of available wireless network technologies, taking into account
 Design and Development of a Fly-by-Wireless UAV Platform 5

the requirements of the application; section 3 presents the global electronics

architecture of the UAV platform, while section 4 describes the developed
onboard wireless system; section 5 presents experimental performance
results obtained with this system, and section 6 presents the conclusions and
addresses the future work.

WIRELESS NETWORK TECHNOLOGIES

Most wireless networks technologies available nowadays can be subdivided
in a few categories: satellite networks, mobile cellular networks, broadband
wireless access, wireless local networks (WLAN) and wireless personal
networks (WPAN). The former three differ substantially from the latter two.
One difference is that the network infrastructure does not belong to the user,
but to the network operator, which charges the user for the services provided.
Other difference is that they provide coverage over a large area. On the
other hand, WLAN and WPAN are short range technologies in which all the
communications equipment usually belongs to the user. These characteristics
are more adequate for the intended application, so the remainder of this
section will focus on wireless network technologies belonging to these two
categories.
The most widespread type of WLAN nowadays is the IEEE 802.11
(IEEE, 2007), also known as WiFi. These networks are available on multiple
physical options and operating frequency bands. However, all these versions
use the same MAC (Medium Access Control) protocol; a contention based
CSMA/CA (Carrier Sense Multiple Access/Collision Avoidance) mechanism
known as DCF (Distributed Control Function). Given its statistical nature,
this protocol is not adequate to provide the QoS guarantees required by the
onboard wireless data acquisition and control system due to the probability
of collisions.
In order to support real-time traffic, the 802.11 standard defines an
alternative MAC protocol know as PCF (Point Coordination Function),
based on a polling scheme, which is capable of providing QoS guarantees.
However, unlike the DCF protocol, the implementation of PCF is not
mandatory, and the availability of products that support it is scarce. More
recently, a newer standard, the IEEE 802.11e (IEEE, 2007), designed to
improve the efficiency and QoS support of 802.11 networks was released,
but its availability on the market is also low.
Concurrently to the development of the 802.11, the European
Telecommunications Standards Institute (ETSI) has developed another
6 Unmanned Aerial Vehicles (UAV) and Drones

WLAN standard: HIPERLAN/2 (ETSI, 2002). HIPERLAN/2 networks are

designed to operate at the 5 GHz band using OFDM (Orthogonal Frequency
Division Modulation). Its physical layer is similar to the one used by the
IEEE 802.11a due to agreements made by the two standard bodies. On
the other hand, the MAC protocols used by these networks are radically
different. HIPERLAN/2 uses a demand based dynamic TDMA (Time
Division Multiple Access) protocol, which is able to provide extensive
support of QoS to multiple types of traffic, including those generated by
data acquisition and control systems (Afonso & Neves, 2005). However, the
802.11 standard won the battle for the wireless LAN market and as such no
available HIPERLAN/2 products are known at the moment.
Due to its design characteristics, 802.11 and HIPERLAN/2 modules
present relatively high power consumption. Although these networks can be
suitable to interconnect devices like computers, there is an enormous potential
market to provide wireless communication capabilities to smaller and cheaper
devices running on batteries without the need of frequent recharging. Such
devices include computer peripherals, biomedical monitoring appliances,
surveillance units and many other sensing and actuation devices. To provide
communication capabilities to such devices, various low cost short range
networks, known collectively by the term wireless personal area network
(WPAN), are being developed.
At the IEEE, the task of standardization of WPAN networks is under the
scope of the IEEE 802.15 group. One of these standards, the IEEE 802.15.4
(IEEE, 2006) defines the physical and MAC layer of ZigBee (ZigBee, 2006),
which aims to provide low power and low bit rate WPANs with the main
purpose of enabling wireless sensor network applications. At the physical
layer, the IEEE 802.15.4 relies on direct sequence spread spectrum (DSSS)
to enhance the robustness against interference, and provides gross data rates
of 20/40 kbps, at the 868/915 band, and 250 kbps, at the 2.4 GHz band.
As in 802.11 networks, the basic ZigBee MAC protocol is a contention
based CSMA/CA mechanism. A complementary mechanism defined in the
802.15.4 standard, the guaranteed time slot (GTS), enables the provision of
some QoS guarantees to real-time traffic.
Bluetooth (Bluetooth, 2003) is another WPAN technology. It operates
in the 2.4 GHz band using frequency hopping spread spectrum (FSSS)
and provides a gross data rate of 1 Mbps. Bluetooth operates using a star
topology, called piconet, formed by one master and up to seven active slaves.
Transmissions can achieve a range of 10 or 100 m, depending of the class
 Design and Development of a Fly-by-Wireless UAV Platform 7

of the device. At the MAC layer, the Bluetooth devices uses a polling based
protocol that provides support for both real-time and asynchronous traffic.
Bluetooth provides better overall characteristics than the other networks
discussed here for the desired application. It drains much less power than
802.11 and HIPERLAN/2, uses a MAC protocol that provides support
for real-time traffic, and provides a higher gross data rate than ZigBee.
Bluetooth spread spectrum covers a bandwidth of 79 MHz while ZigBee
operates in a band of less than 5 MHz, what makes the former more robust
against interference. Moreover, Bluetooth provides an adaptive frequency
hopping mechanism that avoids frequency bands affected by interference.
Given these characteristics and the availability of the technology at the time
of development, Bluetooth was chosen as the supporting wireless network
technology for the development of the prototype of the system described in
the following sections.

GLOBAL ELECTRONICS ARCHITECTURE

The global view of architectural model of the onboard computing and
communication system of the AIVA fly-by-wireless UAV platform is
presented in Figure 1. It is a multitasking/multiprocessor based system
connected by an asynchronous local bus that allows for speed adaptation of
different tasks/processors. The system architecture supports one processing
unit for a Bluetooth piconet master node, one flight controller unit, one data
logger and earth link, and one embedded vision system (EVS). In each of
these nodes many critical processes are permanently running.

Figure 1. Global electronics architecture of the AIVA UAV platform This ar-
chitecture allows an easy way to introduce or remove processing units from the
platform. For instance new sensors or new vision units can be included. In the
first case a new module must be connected to the Bluetooth piconet, and in the
second case the new module is connected to the local multi-access bus.
8 Unmanned Aerial Vehicles (UAV) and Drones

ONBOARD WIRELESS SYSTEM

The AIVA UAV platform implements an onboard wireless distributed data
acquisition and control system based on Bluetooth (BT) wireless network
technology, represented by the Bluetooth piconet of Figure 1. The general
architecture of a wireless node is presented on Figure 2. Each node is
composed by a commercial off-the-shelf Bluetooth module that contains
the radio electronics, a microcontroller that runs the code that controls the
behavior of the node, and a local bus that provides interfacing between the
node components, as well as specific sensors and/or actuators according to
the purpose of the node.

Figure 2. Architecture of a Bluetooth wireless node

Physical Architecture
The physical part of the platform is built around a low power Texas
Instruments MSP430 microcontroller, a Von-Neumann 16 bit RISC
architecture with mixed program, data and
I/O in a 64Kbytes address space. Besides its low power profile, which
uses about 280 µA when operating at 1 MHz @ 2.2 Vdc, MSP430 offers some
interesting features, like single cycle register operations, direct memory-
to-memory transfers and a CPU independent hardware multiplication
unit. From the flexibility perspective, a flexible I/O structure capable of
independently dealing with different I/O bits, in terms of data direction,
interrupt programming, and edge triggering selection; two USARTs
supporting SPI or UART protocols; an onboard 12 bit SAR ADC with 200
kHz rate; and PWM capable timers, are all relevant features.
 Design and Development of a Fly-by-Wireless UAV Platform 9

The Bluetooth modules chosen for the implementation of the wireless nodes
are OEM serial adapter devices manufactured by connectBlue. The master
node uses an OEMSPA33i module and the slave nodes use OEMSPA13i
modules (connectBlue, 2003). These modules include integrated antennas;
nevertheless, we plan to replace them with modules with external antennas
in future versions of the platform, to be able to shield the modules in order
to increase the reliability of the system against electromagnetic interference.
While the module used on the master (OEMSPA33i) allows up to seven
simultaneous connections, the module used on the slaves (OEMSPA13i) has
a limitation of only three simultaneous connections. However, this limitation
does not represent a constraint to the system because the slaves only need to
establish one connection (to the master).
The connectBlue modules implement a virtual machine (VM) that enables
the provision of a serial interface abstraction to the microcontroller, so Bluetooth
stack details can be ignored and focus can be directed to the application. The
manufacturer’s virtual machine implements a wireless multidrop access scheme
where the master receives all frames sent by the slaves and all slaves can listen to
the frames sent by the master, in a point-to-multipoint topology.The AIVA onboard
wireless system is composed by one Bluetooth piconet containing seven nodes:
one master (MM - Master Module) and six slaves (SAM - Sensing & Actuation
Modules). The nodes are spread over the aircraft structure, as shown in Figure 3.
The master node (MM) is placed at the fuselage body, and acts as the network
and flight controller, onboard data logger, and communications controller for the
link with the ground station. On each wing, there is a SAM node for an electrical
propulsion motor and for control surfaces (ailerons and flaps). These wing nodes
are responsible for motor speed control and operating temperature monitoring, as
well as control surfaces actuation and position feedback.
In the fuselage body, there are other two SAM nodes, one for a GPS module
and other for an inertial measurement unit (IMU), which provide information
assessment for navigational purposes. At the tail, there is another SAM node
for elevator and rudder control, and position feedback. Finally, at the nose
there is a SAM node connected to a nose probe consisting of a proprietary
design based on six independent pressure sensors that give valuable digital
information for flight control. This node also contains an ultrasonic probe
that provides information for support of the automatic take-off and landing
system. Figure 4 displays the physical layout of the nose node. The Bluetooth
module is in the lower corner, the microcontroller is on the left hand side and
the sensor hardware on the right hand side of the board.
10 Unmanned Aerial Vehicles (UAV) and Drones

Figure 3. Node distribution on the aircraft structure

Figure 4. Physical layout of the nose node

Logical Architecture
The logical architecture of the developed system is a two layered state
machine implementation, composed by a transport layer and an application
layer. The transport layer provides a packet delivery service, under control
of master node, capable of transparent delivery of data packets across the
network.
The transport layer is application independent, and interfaces with the
top level application layer by means of a data space for buffering and a set
of signaling control bits that allow total synchronization between the two
layers. The hierarchy and signaling between the two layers is represented
in Figure 5.
 Design and Development of a Fly-by-Wireless UAV Platform 11

Figure 5. Hierarchy and signaling at the logical level of the platform.

The asynchronous reception process delivers characters to upper
processes. Analyzing the hierarchy from the lower level to the upper level,
CharReady condition goes TRUE every time a new character arrives to
the interface. The next process in the hierarchy is PacketAssembler, a state
machine that performs packet re-assembly, reconstructing the original
packet from a group of segments, and delivers packets for the next process
in the hierarchy. When MsgRcvd (message received) goes TRUE, a new
message is ready for processing. Thus, for incoming data, the model at
layer 1 receives characters and delivers ready-to-process messages to the
application layer. When the application layer understands that the message is
ready to process, a command processor for incoming messages is activated
in order to decode the embedded command and semantics contained in the
message, to eventually execute some action, and to pass relevant information
for the final application.
For outgoing data, the resident application eventually makes
available some data to transmit to the master, signaling this event with a
DataReady signal. This causes the output command processor to execute
its cycle, preparing one message to be sent. When the message is ready,
OutCommandReady goes TRUE, signaling to the lower layer that there is
a message to send to the network. At this phase, frame segmentation starts
(if needed) by means of a state machine for packet disassembly. This state
machine breaks the original message in smaller segments prepared to be
serialized. Each time a segment is ready to be sent, Msg2Send goes TRUE
and serialization is triggered. So, for outgoing data, the transport layer
12 Unmanned Aerial Vehicles (UAV) and Drones

receives messages from application layer, and sends segments to the radio
module in order to be sent over the wireless medium.
Layer 2 is application dependent, and has no knowledge of the lower layer
internal mechanisms, just the services and interfaces available from it. That
means that its logical architecture can be used in other applications. For the fly-
by-wireless application, its main goal is to replicate a system table among all
network nodes, at the maximum possible frequency. This system table maintains
all critical system values, describing the several sensors and actuators, status
parameters, and loop feedback information. Each network node is mapped to a
table’s section, where all related variables from sensing, actuators and metering
are located. This layer is responsible for cyclic refreshing the respective table
contents, based on local status, and also for cyclic actuation according to data
sent from the master node (flight controller orders). This way, the whole system
is viewed as a resident two-dimensional array located at master, with different
partial copies distributed and synchronized among the slave nodes.

Other Design Issues

All the Bluetooth modules in the developed platform are configured in non
discoverable mode, which contributes to the security of the system. The node
discovery process of Bluetooth is a slow process, in the order of seconds,
however it is not a problem since the master stores the addresses of all slave
nodes that should participate in the piconet, so this process is avoided. The
piconet formation is performed on the ground before the takeoff procedure,
so the associated delay does not constitute a problem as well.
The use of Bluetooth technology limits the piconet operation to a
maximum of seven active slaves; however, this limitation is not of major
concern on the developed system, since only six slaves are used, and could
only impose some restrains if node number should be raised. The number
of slaves in the network could be increased by interconnecting a number of
piconets to form a scatternet. That way, a device participating in two piconets
could relay traffic between both piconets. However, this architecture would
probably have a negative impact in the performance of the network, making
it more difficult to provide QoS guarantees to the application. Moreover,
currently there are very few actual implementations of scatternets available.
Given that free space propagation loss is proportional to the square
of the distance, it is not expected that the onboard wireless network will
either suffer or induce interference on other networks operating in the same
frequency band, such as the widely deployed WiFi networks, since the
 Design and Development of a Fly-by-Wireless UAV Platform 13

former operates in the sky most of time, while the later are normally based
on the ground.

EXPERIMENTAL RESULTS
The performance of the developed wireless system was evaluated in
laboratory. The experimental setup used to achieve the results presented in
this section is composed by 6 slaves sending data periodically to the master
(uplink direction) at the same predefined sampling rate. Each sampling
packet has a length of 15 octets, which is the maximum packet length
supported by the transport layer due to a limitation imposed by the virtual
machine used by the Bluetooth module.
Figure 6 presents the aggregated uplink throughput that reaches the
master node as a function of the sampling rate used by the 6 slaves. Since
Bluetooth uses a contention-free MAC protocol, the uplink transmissions are
not affected by collisions, so the network throughput increases linearly with
the offered load until the point it reaches saturation, which in this scenario
corresponds to the situation where the slaves transmit data at sampling rates
higher than 200 Hz. As this figure shows, the maximum throughput available
to the application is about 160 kbps, which is significantly lower than the gross
data rate provided by Bluetooth (1 Mbps). This difference can be explained
by the overhead introduced by the Bluetooth protocol and the virtual machine,
including the gap between the packets, the FEC (Forward Error Correction)
and ARQ (Automatic Repeat reQuest) mechanisms, the packet headers, as
well as the overhead introduced by control packets such as the POLL packet,
that is sent by the master to grant permission to slaves to transmit.

Figure 6. Uplink throughput as a function of the sampling rate

14 Unmanned Aerial Vehicles (UAV) and Drones

Figure 7 presents the packet loss ratio (PLR) averaged over the 6 slaves as a
function of the sampling rate. As the figure shows, the PLR is limited to less
than 0.5 % in the region where the network is not congested, but increases
rapidly after the saturation point. The flight control application should be
able to tolerate such small losses; otherwise a change in the supporting
wireless technology should be made in the attempt to obtain higher link
reliability.

Figure 7. Packet loss ratio as a function of the sampling rate

Figure 8. Complementary cumulative distribution of the delay

 Design and Development of a Fly-by-Wireless UAV Platform 15

Concerning to the delay experienced by the packets as they travel from the
slaves to the master, results showed that the delay is not adversely affected
by the rise in the offered load, as long as the network operates below the
saturation point. For sampling rates up to 200 Hz, the registered average
delay was 27 ms and the standard deviation was 16 ms.
Figure 8 presents the complementary cumulative distribution (P{X>x})
for the random variable X, which represents the delay for all the samples
collected using sampling rates in the range from 0 to 200 Hz. With this
chart, it is possible to see the probability that the delay exceeds a given delay
bound, which is an important metric for real-time applications such as the
one considered in this chapter. The chart shows, for instance, that less than
1 % of the sample packets suffer a delay higher than 90 ms, while less than
0.1 % of the packets suffer a delay higher than 120 ms.
Experimental tests were also made with a varying number of slaves in the
piconet (from 1 to 6), both in the uplink and downlink direction. The average
delay measured in the downlink direction (from the master to the slaves) was
slightly higher than the one registered in the uplink direction, but below 40 ms,
for the measurements made with up to 4 slaves. However, the average master-
to-slave delay with 5 slaves in the network ascended to 600 ms, while with 6
slaves the performance was even worse, with the average delay reaching 1000
ms.

CONCLUSION
This chapter presented the design and development of a fly-by-wireless
UAV platform built on top of Bluetooth wireless technology. The developed
onboard wireless system is composed by one master node, connected to the
flight controller and six slave nodes spread along the aircraft structure and
connected to several sensors and actuators.
In order to assess the suitability of the developed system, several
performance evaluation tests were carried out. The experimental results
showed that, for the slave-to-master direction, the system prototype is
able to support a sampling rate of up to 200 Hz for each of the 6 slaves
simultaneously without significant performance degradation in terms of
throughput, loss or delay. On the other hand, although the master-to-slave
delay with 1 to 4 slaves in the network is low, its value increases significantly
with 5 and 6 slaves, which is unacceptable given the real-time requirements
of the desired application.
16 Unmanned Aerial Vehicles (UAV) and Drones

This problem is caused by implementation issues related to the proprietary

embedded virtual machine provided by the manufacturer of the Bluetooth
module that is used in the master node of the prototype.
The approach of relying on the virtual machine provided by the
manufacturer, which hides the Bluetooth protocol stack functionality,
allowed the development focus to be directed to the application, reducing
the development costs. The disadvantage, however, is the lack of control of
the behavior of the system at the Bluetooth stack level, which impedes the
optimization of the performance of the system at this level and the correction
of problems such as the verified with the master-to-slave delay. The solution
to the detected problem can pass either by the replacement of the Bluetooth
module by a newer version (already available) from the same manufacturer
or by the direct interaction with the Bluetooth stack, with the bypass of the
virtual machine.
Despite the limitations of the current prototype, the overall results
provided by the experimental tests are satisfactory. Nevertheless, further
tests are needed in order to evaluate the behavior of the system under more
harsh interference conditions, as well as in a real scenario onboard the
aircraft.
 Design and Development of a Fly-by-Wireless UAV Platform 17

REFERENCES
1. Afonso, J. A. & Neves, J. E. (2005), Fast Retransmission of Real-Time
Traffic in HIPERLAN/2 Systems, Proceedings of Advanced Industrial
Conference on Telecommunications (AICT2005), pp. 34-38, ISBN
0-7695-2388-9, Lisbon, Portugal, July 2005, IEEE Computer Society.
2. Bluetooth SIG (2003), Specification of the Bluetooth system. Available
at http://www.bluetooth.org/.
3. connectBlue (2003), Serial Port Adapter – 2nd Generation, User
Manual. Available at http://www.connectblue.com/.
4. ETSI TR 101 683 V1.1.1 (2000), Broadband Radio Access Networks
(BRAN)—HIPERLAN Type 2—Data Link Control (DLC) Layer—
Part 1: Basic Data Transport Functions.
5. IEEE Std 802.11 (2007), IEEE Standard for Information Technology—
Telecommunications and information exchange between systems—
Local and metropolitan area networks—Specific requirements—
Part11: Wireless LAN Medium Access Control (MAC) and Physical
Layer (PHY) Specifications.
6. IEEE Std 802.15.4 (2006), IEEE Standard for Information Technology—
Telecommunications and information exchange between systems—
Local and metropolitan area networks—Specific requirements—Part
15.4: Wireless Medium Access Control (MAC) and Physical Layer
(PHY) Specifications for Low-Rate Wireless Personal Area Networks
(WPANs).
7. ZigBee Standards Organization (2006), ZigBee Specification.
Available at http://www.zigbee.org/.
 CHAPTER

2
A Review of Control Algorithms for
Autonomous Quadrotors

Andrew Zulu, Samuel John

Department of Mechanical and Marine Engineering, Polytechnic of Namibia,
Windhoek, Namibia

ABSTRACT
The quadrotor unmanned aerial vehicle is a great platform for control
systems research as its nonlinear nature and under-actuated configuration
make it ideal to synthesize and analyze control algorithms. After a brief
explanation of the system, several algorithms have been analyzed including
their advantages and disadvantages: PID, Linear Quadratic Regulator

Citation: Zulu, A. and John, S. (2014), “A Review of Control Algorithms for Au-
tonomous Quadrotors”. Open Journal of Applied Sciences, 4, 547-556. doi: 10.4236/
ojapps.2014.414053.
Copyright: © 2014 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
20 Unmanned Aerial Vehicles (UAV) and Drones

(LQR), Sliding mode, Backstepping, Feedback linearization, Adaptive,

Robust, Optimal, L1, H∞, Fuzzy logic and Artificial neutral networks. The
conclusion of this work is a proposal of hybrid systems to be considered as
they combine advantages from more than one control philosophy.
Keywords: Control, Algorithms, Quadrotors, Intelligent Control,
Optimal Control, Robust Control, Adaptive Control, Linear Control,
Nonlinear Control

INTRODUCTION
The quadrotor unmanned aerial vehicle (UAV) are helicopters with four
rotors typically designed in a cross configuration with two pairs of opposite
rotors rotating clockwise and the other rotor pair rotating counter-clockwise
to balance the torque. The roll, pitch, yaw and up-thrust actions are controlled
by changing the thrusts of the rotors using pulse width modulation (PWM)
to give the desired output.
Civilian use of unmanned/micro aerial vehicles (UAVs/MAVs) include
aerial photography for mapping and news coverage, inspection of power
lines, atmospheric analysis for weather forecasts, traffic monitoring in urban
areas, crop monitoring and spraying, border patrol, surveillance for illegal
imports and exports, fire detection and control, search and rescue operations
for missing persons and natural disasters.
In this work, the focus is to realize the best configuration and control
scheme of the quadrotor MAV for application in game counting in the
protected game reserves in Africa. The current situation in most African
countries is that researchers and game rangers either count from road
vehicles by driving by the road or use conventional helicopters. The former
is ineffective as the count is limited to only animals coming close to the road
whereas the later leads to inaccurate counts as the animals run for safety
when a helicopter hovers above them. Several advantages accrue to the
quadrotor which has led to much attention from researchers. This is mainly
due to its simple design, high agility and maneuverability, relatively better
payload, vertical take-off and landing (VTOL) ability.
The quadrotor does not have complex mechanical control linkages due
to the fact that it relies on fixed pitch rotors and uses the variation in motor
speed for vehicle control [1]. However, these advantages come at a price
as controlling a quadrotor is not easy because of the coupled dynamics
and its commonly under-actuated design configuration [2]. In addition, the
 A Review of Control Algorithms for Autonomous Quadrotors 21

dynamics of the quadrotor are highly non-linear and several uncertainties

are encountered during its missions [3], thereby making its flight control a
challenging venture. This has led to several control algorithms proposed in
the literature. In this work, a review of the prominent controllers applied to
the quadrotor is reviewed.
The rest of this paper is organized as follows: Section 2 provides a brief
description of the mathematical model of the quadrotor; Section 3 gives a
review of the popular controllers proposed for the quadrotor and a discussion
of hybrid control systems; in Section 4 a tabular comparison is presented
and the conclusions are given in Section 5.

MATHEMATICAL MODEL
In this section a brief description of the main mathematical equation for the
quadrotor is explained.
In flight mode, the quadrotor is able to generate its lift forces by controlling
the fixed pitch angle rotor speeds. Figure 1 presents the schematic diagram.
The center of mass is the origin O of the coordinate system (see Figure 1)
and the forward direction is arbitrarily fixed in the x-axis. To lift off from
ground, all four rotors are rotated at the same speed in the sense shown. A
total thrust equal to the weight of the system stabilizes the system in hover.
To roll about the x-axis, differential thrust of rotors 2 and 4 is applied. To
pitch about the y-axis, differential thrust of rotors 1 and 3 is applied. The
yaw motion is achieved by the differential thrusts of the opposite rotors (1/3
or 2/4) while also adjusting the constant thrusts of the remaining opposite
pairs to maintain altitude.
Using the Newton-Euler formalism, the equations of motion of the
quadrotor are given in [4]:

Figure 1. Schematic of quadrotor showing rotor configurations.

22 Unmanned Aerial Vehicles (UAV) and Drones

where Φ , θ and ψ are the roll, pitch and yaw angles respectively; are
the mass moments of inertia in the x , y , and z axes respectively; Jr is the rotor inertia;
Ω is the angular velocity of the rotor; l is the length of the rotor arm from the origin of
the coordinate system; and the inputs to the four rotors are given by the expressions:

where b and d are thrust and drag coefficients respectively. For detailed
derivation of the equations, consult [4].

SURVEY OF CONTROL ALGORITHMS

Due to the nature of the dynamics of the quadrotor, several control algorithms
have been applied to it. As to be expected, each control scheme has its
advantages and disadvantages. The control schemes used could be broadly
categorized as linear and non-linear control schemes. In this review a broad
range of controllers within these categories are discussed.

Proportional Integral Derivative (PID)

The PID controller has been applied to a broad range of controller
applications. It is indeed the most applied controller in industry [5]. The
classical PID linear controller has the advantage that parameter gains are
easy to adjust, is simple to design and has good robustness. However some of
 A Review of Control Algorithms for Autonomous Quadrotors 23

the major challenges with the quadrotor include the non-linearity associated
with the mathematical model and the imprecise nature of the model due to
unmodeled or inaccurate mathematical modeling of some of the dynamics.
Therefore applying PID controller to the quadrotor limits its performance.
A PID controller was used for the attitude control of a quadrotor in [6],
while a dynamic surface control (DSC) was used for the altitude control.
Applying Lyapunov stability criteria, Lee et al. were able to prove that all
signals of the quadrotor were uniformly ultimately bounded. This meant that
the quadrotor was stable for hovering conditions. From the simulation and
experimental plots however, it reveals the PID controller to have performed
better in the pitch angle tracking, whereas large steady state errors could be
observed in the roll angle tracking.
In another work by Li and Li [7], a PID controller was applied to regulate
both position and orientation of a quadrotor. The PID parameter gains were
chosen intuitively. The performance of the PID controller indicated relatively
good attitude stabilization. The response time was good, with almost zero
steady state error and with a slight overshoot.
It is generally established in the literature that the PID controller has
been successfully applied to the quadrotor though with some limitations.
The tuning of the PID controller could pose some challenges as this must
be conducted around the equilibrium point, which is the hover point, to give
good performance.
Figure 2 shows the general block diagram of a PID controller for the
quadrotor.

Linear Quadratic Regulator/Gaussian-LQR/G

The LQR optimal control algorithm operates a dynamic system by
minimizing a suitable cost function. Boubdallar and co-researchers applied
the LQR algorithm to a quadrotor and compared its performance to that
of the PID controller in [8]. However, the PID is applied on the quadrotor
simplified dynamics and the LQR on the complete model. Both approaches
provided average results but it implicitly was clear that the LQR approach
had better performance considering the fact that it was applied to a more
complete dynamic model.
A simple path-following LQR controller was applied by Cowling et al. in
[9] on a full dynamic model of the quadrotor. It was shown that accurate path
following was achieved in simulation using optimal real-time trajectories
despite the presence of wind and other disturbances. The controller seemed
24 Unmanned Aerial Vehicles (UAV) and Drones

to lose tracking after avoiding an obstacle. Its performance in the presence

of many obstacles still needed to be analyzed.
In combination with a Linear Quadratic Estimator (LQE) and Kalman
Filter, the LQR algorithm transforms into the Linear Quadratic Gaussian
(LQG). This algorithm is for systems with Gaussian noise and incomplete
state information. The LQG with integral action was applied in [10] for
stabilization of attitude of a quadrotor with good results in hover mode. The
advantage of this LQG controller is the fact that you do not need to have
complete state information to implement it.
Figure 3 shows the general block diagram of an LQG controller for the
quadrotor.

Sliding Mode Control (SMC)

Sliding mode control is a nonlinear control algorithm that works by applying
a discontinuous control signal to the system to command it to slide along a
prescribed path. Its main advantage is that it does not simplify the dynamics
through linearization and has good tracking. The sliding mode controller was
applied to stabilize cascaded under-actuated systems in [11]. The quadrotor
system was subdivided into the full-actuated and underactuated systems. The
under-actuated system, to which SMC was applied, was further subdivided
into underactuated subsystems. Results showed good stability and robustness
of the system. Chattering effect of SMC was observed but minimized with a
continuous approximation of a pre-determined “sign” function.

Figure 2. Block diagram of PID controller applied to the quadrotor.

Figure 3. Block diagram of LQG controller applied to the quadrotor.

 A Review of Control Algorithms for Autonomous Quadrotors 25

A sliding mode controller based on Lyapunov stability theory is articulated

by Runcharoon and Srichatrapimuk in [12]. The SMC controller was able
to stably drive the quadrotor to a desired position and yaw. Tracking was
equally good with injected noise, which showed good robustness for the
SMC controller.
Figure 4 shows the general block diagram of an SMC controller for the
quadrotor.

(Integrator) Backstepping Control

Backstepping control is a recursive algorithm that breaks down the controller
into steps and progressively stabilizes each subsystem. Its advantage is that
the algorithm converges fast leading to less computational resources and it
can handle disturbances well. The main limitation with the algorithm is its
robustness is not good. Madani and co-researchers [13] applied backstepping
control to stabilize a quadrotor system consisting of an under-actuated, fully
actuated and propeller subsystems. Good tracking was achieved for position
and yaw angle. Using Lyapunov stability theory, roll and pitch angles were
stabilized.
The backstepping approach was also applied in [1] for attitude
stabilization of a quadrotor. Using Lyapunov stability analysis, the closed-
loop attitude system was found to be asymptotically stable with all states
uniformly ultimately bounded in the presence of external disturbance. It was
also implicit that the quaternion formulation also helped in the computational
side for stabilization in addition to avoiding singularity.
To increase robustness (to external disturbances) of the general
backstepping algorithm, an integrator is added and the algorithm becomes
Integrator backstepping control as articulated by Fang and Gao in [14].
The integral approach was shown to eliminate the steady-state errors of
the system, reduce response time and restrain overshoot of the control
parameters.

Adaptive Control Algorithms

Adaptive control algorithms are aimed at adapting to parameter changes
in the system. The parameters are either uncertain or varying with time. A
continuous time-varying adaptive controller, which shows good performance,
was implemented by Diao et al. in [15] with known uncertainties in mass,
moments of inertia and aerodynamic damping coefficients.
26 Unmanned Aerial Vehicles (UAV) and Drones

In other work by Palunko et al., an adaptive control scheme using feedback

linearization (FBL) is implemented [16] for quadrotors with dynamic
changes in the center of gravity. It was observed that when the center of
gravity changes, PD and regular feedback linearization techniques were not
able to stabilize the system but the adaptive controller was able to stabilize
it. The paper touches on a very important futuristic quadrotor that will self-
configure in real time when the center of gravity changes.
An adaptive control technique based on rectilinear distance (L1 ) norm
was used in [17] with a tradeoff between control performance and robustness.
The modified (linearized) model was able to compensate for constant and
moderate wind gusts.
Figure 5 shows the general block diagram of an adaptive controller
for the quadrotor clearing showing the parameter estimator and quadrotor
model.

Figure 4. Block diagram of an SMC controller applied to the quadrotor.

Figure 5. Block diagram of an adaptive controller applied to the quadrotor.

Robust Control Algorithms

Robust control algorithms are designed to deal with uncertainty in the system
parameters or disturbances. This guarantees controller performance within
acceptable disturbance ranges or un-modeled system parameters. A major
limitation normally observed with robust controllers is poor tracking ability.
A robust controller was implemented by Bai et al. in [18] for an attitude
 A Review of Control Algorithms for Autonomous Quadrotors 27

controller subsystem based on linear control and robust compensation.

Experimental results validated the effectiveness of the controller for
controlling attitude.
In another work by Tony and Mackunisy a robust tracking algorithm is
presented [19] and achieves asymptotic stability in the presence of parametric
uncertainties and unknown nonlinear disturbances. This was achieved with
inexpensive control law, no observers, functional approximators or online
adaptive updating laws.

Optimal Control Algorithms

Optimization algorithms are designed to minimize a variable and get the best
cost function from a set of alternatives. A specialized case of optimization
is known as convex optimization. This is a mathematical optimization
technique that deals with minimizing a convex variable in a convex set. The
common optimal algorithms include LQR, L1 , H∞ and Kalman filter. A major
limitation of optimization algorithms is generally their poor robustness. A
controller that is both relatively robust and L1 -optimal was applied in [20]
for tracking of both attitude and heading. The performance of the controller
was experimentally validated and was efficient in error minimization and
rejection of persistent disturbances.
An optimization algorithm based on H∞ looping was applied by
Falkenberg et al. in [21] to a quadrotor with simplified dynamics for iterative
parameter identification and attitude control. The algorithm had superior
performance with respect to disturbance rejection even under disturbance
saturation conditions. However, the ant-windup compensator based on
Riccati equations did not seem to show computational efficiency.
An integral optimal predictive H∞ controller was implemented by Raffo
et al. in [22] for stabilization of rotational movement of a quadrotor and for
path following (using model predictive control). The controller achieved
sustained disturbance rejection and good robustness. Integral action was key
to achieving good tracking in this algorithm.

Feedback Linearization
Feedback linearization control algorithms transform a nonlinear system
model into an equivalent linear system through a change of variables.
Some limitation of feedback linearization is the loss of precision due to
linearization and requiring an exact model for implementation. Output
feedback linearization was implemented as an adaptive control strategy in
28 Unmanned Aerial Vehicles (UAV) and Drones

[16] for stabilization and trajectory tracking on a quadrotor that had dynamic
changes of its center of gravity. The controller was able to stabilize the
quadrotor and reconfigure it in real time when the center of gravity changed.
Feedback linearization and input dynamic inversion were implemented
by Roza and Maggiore in [23] to design a path-following controller which
allowed the designer to specify the speed profile and yaw angle as a function
of displacement along the path. Two simulation cases with the quadrotor
travelling at different speeds along the path were considered. Both cases
showed convergence of velocity and yaw angle.
Feedback linearization was compared with adaptive sliding mode control
in [24]. The feedback controller, with simplified dynamics, was found to be
very sensitive to sensor noise and not robust. The sliding mode controller
performed well under noisy conditions and adaptation was able to estimate
uncertainties such as ground effect.
Hence, feedback linearization nonlinear control shows good tracking
but poor disturbance rejection. However, feed-back linearization applied
with another algorithm with less sensitivity to noise give good performance.

Intelligent Control (Fuzzy Logic and Artificial Neural Net-

works)
Intelligent control algorithms apply several artificial intelligence approaches,
some biologically-inspired, to control a system. Examples include fuzzy
logic, neural networks, machine learning, and genetic algorithm. They
typically involve considerable uncertainty and mathematical complexity.
This complexity and abundant computational resources required are
limitations to the use of intelligent systems.
Intelligent control is not limited to fuzzy logic and neural networks but
the two are the most widely used. Many other algorithms exist and continue
to be formulated. Fuzzy logic algorithms deal with many-valued logic, not
discrete levels of truth. An intelligent fuzzy controller was applied by Santos
et al. in [25] to control the position and orientation of a quadrotor with good
response in simulation. However, a major limitation of this work was the
trial and error approach for tuning of input variables.
Artificial neural networks are biologically inspired by the central
nervous system and brain. A robust neural networks algorithm was applied
to a quadrotor by Nicol et al. in [26] to stabilize against modeling error
and considerable wind disturbance. The method showed improvements
 A Review of Control Algorithms for Autonomous Quadrotors 29

with respect to achieving desired attitude and reducing weight drift. Output
feedback control was implemented on a quadrotor using neural networks [27]
for leader-follower quadrotor formation to learn the complete dynamics of
the UAV including un-modeled dynamics. A virtual neural network control
was implemented to control all six degrees of freedom from four control
inputs. Using Lyapunov stability theory, it was shown that the position,
orientation, velocity tracking errors, observer estimation errors and virtual
control were semi-globally uniformly ultimately bounded.
An adaptive neural network scheme was applied in [28] for quadrotor
stabilization in the presence of a sinusoidal disturbance. The proposed
solution of two parallel single hidden layers proved fruitful as reduced
tracking error and no weight drift were achieved.
Figure 6 shows the general block diagram of a fuzzy logic controller
implemented on the quadrotor.

Hybrid Control Algorithms

It is evident that even the best linear or nonlinear algorithms had limitations
and no single controller had it all. Researchers have tackled this by combining
the philosophies of one or more algorithms. Here are few examples, which
are not in any way exhaustive of what is in literature.
A hybrid fuzzy controller with backstepping and sliding mode control
was implemented in [29] and successfully eliminated chattering effect of
the sliding mode control algorithm. A feedback linearization controller was
applied parallel with a high-order sliding mode controller to a quadrotor in
[30]. The sliding mode controller was used as an observer and estimator of
external disturbances. The system showed good disturbance rejection and
robustness.

Figure 6. Block diagram an FLC controller applied to the quadrotor.

Adaptive control via backstepping combined with neural networks was
presented by Madani and co-researchers in [31]. The backstepping was used
to achieve good tracking of desired translational positions and yaw angle
30 Unmanned Aerial Vehicles (UAV) and Drones

whilst maintaining stability of roll and pitch angles. Neural networks were
used to compensate for unmodeled dynamics. The major contribution of the
paper was the fact that the controller did not require the dynamic model and
other parameters. This meant greater versatility and robustness.

COMPARISON OF CONTROL ALGORITHMS

Table 1 summarizes the comparison of the various algorithms as applied
to quadrotors with all things being equal. The performance of a particular
algorithm depends on many factors that may not even be modeled. Hence,
this table serves as “fuzzy” guide in accordance with what is presented in
this paper and common knowledge.

DISCUSSION AND CONCLUSION

This survey acts as a stepping stone into further work that will be conducted
by the authors on control of quadrotors with tilting propellers. The tilting
propellers further increase the number of actuated inputs to give six control
inputs for six control outputs, which makes the quadrotor fully actuated.
This research direction is relatively new with few examples in literature but
with very promising performance achievement (see [32]-[34]). This work
will focus on optimal control algorithms to achieve asymptotic stability of
the quadrotor under any conditions.

Table 1. Comparison of quadrotor control algorithms.

This paper has reviewed several common control algorithms that have
 A Review of Control Algorithms for Autonomous Quadrotors 31

been used on quadrotors in literature. As evident from the review, no single

algorithm presents the best of the required features. It also been discussed
that getting the best performance usually requires hybrid control schemes that
have the best combination of robustness, adaptability, optimality, simplicity,
tracking ability, fast response and disturbance rejection among other factors.
However, such hybrid systems do not guarantee good performance; hence
a compromise needs to be found for any control application on which
of the factors would be most appropriate. The jury is still out as to what
mathematical model would give the best overall performance.
The designed quadrotor will be used for game counting at the National
Parks; hence it should have the following important characteristics: high
endurance, low noise, high agility, high cruising and VTOL ability. The
quadrotor will be a great aid for nature conservationists that are looking
for innovative ways to combat poaching and get accurate animal statistics.
Future possibilities with this type of quadrotor would include the possibility
of equipping it with a manipulator that can do other things such as spraying
of medicines or tagging.
32 Unmanned Aerial Vehicles (UAV) and Drones

REFERENCES
1. Huo, X., Huo, M. and Karimi, H.R. (2014) Attitude Stabilization
Control of a Quadrotor UAV by Using Backstepping Approach.
Mathematical Problems in Engineering, 2014, 1-9.
2. Mahony, R., Kumar, V. and Corke, P. (2012) Multirotor Aerial Vehicles:
Modeling, Estimation, and Control of Quadrotor. Robotics Automation
Magazine, 19, 20-32. http://dx.doi.org/10.1109/MRA.2012.2206474
3. Lee, B.-Y., Lee, H.-I. and Tahk, M.-J. (2013) Analysis of Adaptive
Control Using On-Line Neural Networks for a Quadrotor UAV.
13th International Conference on Control, Automation and Systems
(ICCAS), 20-23 October 2013, 1840- 1844.
4. Bouabdallah, S. (2007) Design and Control of Quadrotors with
Application to Autonomous Flying. Ph.D. Dissertation, Lausanne
Polytechnic University, Lausanne.
5. John, S. (2013) Artificial Intelligent-Based Feedforward Optimized
PID Wheel Slip Controller. AFRICON, 12 September 2013, Pointe-
Aux-Piments, 1-6.
6. Lee, K.U., Kim, H.S., Park, J.-B. and Choi, Y.-H. (2012) Hovering
Control of a Quadroto. 12th International Conference on Control,
Automation and Systems (ICCAS), 17-21 October 2012, 162-167.
7. Li, J. and Li, Y. (2011) Dynamic Analysis and PID Control for a
Quadrotor. International Conference on Mechatronics and Automation
(ICMA), 7-10 August 2011, 573-578.
8. Bouabdallah, S., Noth, A. and Siegwart, R. (2004) PID vs LQ Control
Techniques Applied to an Indoor Micro Quadrotor. Proceedings of
IEEE/RSJ International Conference on Intelligent Robots and Systems
(IROS 2004), Vol. 3, 28 September-2 October 2004, 2451-2456.
9. Cowling, I.D., Yakimenko, O.A., Whidborne, J.F. and Cooke, A.K.
(2007) A Prototype of an Autonomous Controller for a Quadrotor UAV.
European Control Conference, July 2007, Kos, 1-8.
10. Minh, L.D. and Ha, C. (2010) Modeling and Control of Quadrotor
MAV Using Vision-Based Measurement. International Forum on
Strategic Technology (IFOST), 13-15 October 2010, 70-75.
11. Xu, R. and Ozguner, U. (2006) Sliding Mode Control of a Quadrotor
Helicopter. Proceedings of the 45th IEEE Conference on Decision and
Control, San Diego, 13-15 December 2006, 4957-4962.
12. Runcharoon, K. and Srichatrapimuk, V. (2013) Sliding Mode Control
 A Review of Control Algorithms for Autonomous Quadrotors 33

of Quadrotor. Proceedings of the 2013 International Conference on

Technological Advances in Electrical, Electronics and Computer
Engineering (TAEECE), Konya, 9-11 May 2013, 552-557.
13. Madani, T. and Benallegue, A. (2006) Backstepping Control for a
Quadrotor Helicopter. Proceedings of the 2006 IEEE/RSJ International
Conference on Intelligent Robots and Systems, Beijing, 9-15 October
2006, 3255-3260.
14. Fang, Z. and Gao, W. (2011) Adaptive Integral Backstepping Control
of a Micro-Quadrotor. Proceedings of the 2nd International Conference
on Intelligent Control and Information Processing (ICICIP), Harbin,
25-28 July 2011, 910- 915.
15. Diao, C., Xian, B., Yin, Q., Zeng, W., Li, H. and Yang, Y. (2011)
A Nonlinear Adaptive Control Approach for Quadrotor UAVs.
Proceedings of the 8th Asian Control Conference (ASCC), Kaohsiung,
15-18 May 2011, 223-228.
16. Palunko, I. and Fierro, R. (2011) Adaptive Control of a Quadrotor with
Dynamic Changes in the Center of Gravity. Proceedings of the 18th
IFAC World Congress, Milan, 28 August-2 September 2011, 2626-
2631.
17. De Monte, P. and Lohmann, B. (2013) Position Trajectory Tracking of
a Quadrotor Helicopter Based on L1 Adaptive Control. Proceedings
of the 2013 European Control Conference (ECC), Zurich, 17-19 July
2013, 3346-3353.
18. Bai, Y., Liu, H., Shi, Z. and Zhong, Y. (2012) Robust Control of
Quadrotor Unmanned Air Vehicles. Proceedings of the 31st Chinese
Control Conference (CCC), Hefei, 25-27 July 2012, 4462-4467.
19. Tony, C. and Mackunisy, W. (2012) Robust Attitude Tracking Control
of a Quadrotor Helicopter in the Presence of Uncertainty. Proceedings
of the IEEE 51st Annual Conference on Decision and Control (CDC),
Maui, 10-13 December 2012, 937-942.
20. Satici, A., Poonawala, H. and Spong, M. (2013) Robust Optimal
Control of Quadrotor UAVs. IEEE Access, 1, 79-93. http://dx.doi.
org/10.1109/ACCESS.2013.2260794
21. Falkenberg, O., Witt, J., Pilz, U., Weltin, U. and Werner, H. (2012)
Model Identification and H1 Attitude Control for Quadrotor MAVs.
Intelligent Robotics and Applications, 460-471.
22. Raffo, G.V., Ortega, M.G. and Rubio, F.R. (2010) An Integral Predictive/
34 Unmanned Aerial Vehicles (UAV) and Drones

Nonlinear h? Control Structure for a Quadrotor Helicopter. Automatica,

46, 29-39. http://dx.doi.org/10.1016/j.automatica.2009.10.018
23. Roza, A. and Maggiore, M. (2012) Path Following Controller
for a Quadrotor Helicopter. Proceedings of the American Control
Conference (ACC), Montreal, 27-29 June 2012, 4655-4660. http://
dx.doi.org/10.1109/ACC.2012.6315061
24. Lee, D., Kim, H.J. and Sastry, S. (2009) Feedback Linearization
vs. Adaptive Sliding Mode Control for a Quadrotor Helicopter.
International Journal of Control, Automation and Systems, 7, 419-428.
http://dx.doi.org/10.1007/s12555-009-0311-8
25. Santos, M., Lopez, V. and Morata, F. (2010) Intelligent Fuzzy Controller
of a Quadrotor. Proceedings of the 2010 International Conference on
Intelligent Systems and Knowledge Engineering (ISKE), Hangzhou,
15-16 November 2010, 141-146.
26. Nicol, C., Macnab, C.J.B. and Ramirez-Serrano, A. (2008) Robust
Neural Network Control of a Quadrotor Helicopter. Proceedings of
the Canadian Conference on Electrical and Computer Engineering,
Niagara Falls, 4-7 May 2008, 1233-1237.
27. Dierks, T. and Jagannathan, S. (2010) Output Feedback Control of a
Quadrotor UAV Using Neural Networks. IEEE Transactions on Neural
Networks, 21, 50-66. http://dx.doi.org/10.1109/TNN.2009.2034145
28. Boudjedir, H., Yacef, F., Bouhali, O. and Rizoug, N. (2012) Adaptive
Neural Network for a Quadrotor Unmanned Aerial Vehicle. International
Journal in Foundations of Computer Science and Technology, 2, 1-13.
http://dx.doi.org/10.5121/ijfcst.2012.2401
29. Zeghlache, S., Saigaa, D., Kara, K., Harrag, A. and Bouguerra, A.
(2012) Backstepping Sliding Mode Controller Improved with Fuzzy
Logic: Application to the Quadrotor Helicopter. Archives of Control
Sciences, 22, 315-342.
30. Benallegue, A., Mokhtari, A. and Fridman, L. (2006) Feedback
Linearization and High Order Sliding Mode Observer for a Quadrotor
UAV. Proceedings of the International Workshop on Variable Structure
Systems, Alghero, 5-7 June 2006, 365-372.
31. Madani, T. and Benallegue, A. (2008) Adaptive Control via
Backstepping Technique and Neural Networks of a Quadrotor
Helicopter. Proceedings of the 17th World Congress of the International
Federation of Automatic Control, Seoul, July 6-11 2008, 6513-6518.
 A Review of Control Algorithms for Autonomous Quadrotors 35

32. Cetinsoy, E., Dikyar, S., Hancer, C., Oner, K.T., Sirimoglu, E., Unel, M.
and Aksit, M.F. (2012) Design and Construction of a Novel Quad Tilt-
Wing UAV. Mechatronics, 22, 723-745. http://dx.doi.org/10.1016/j.
mechatronics.2012.03.003
33. Ryll, M., Bulthoff, H. and Giordano, P. (2012) Modeling and Control
of a Quadrotor UAV with Tilting Propellers. Proceedings of the 2012
IEEE International Conference on Robotics and Automation (ICRA),
Saint Paul, 14-18 May 2012, 4606-4613.
34. Senkul, F. and Altug, E. (2013) Modeling and Control of a Novel Tilt—
Roll Rotor Quadrotor UAV. Proceedings of the 2013 International
Conference on Unmanned Aircraft Systems (ICUAS), Atlanta, 28-31
May 2013, 1071-1076.
 CHAPTER

3
Central Command Architecture for
High-Order Autonomous Unmanned
Aerial Systems

Larry M. Silverberg, Chad Bieber

Mechanical and Aerospace Engineering, North Carolina State University,
Raleigh, USA

ABSTRACT
This paper is the first in a two-part series that introduces an easy-to-implement
central command architecture for high-order autonomous unmanned aerial
systems. This paper discusses the development and the second paper
presents the flight test results. As shown in this paper, the central command
architecture consists of a central command block, an autonomous planning

Citation: Silverberg, L. and Bieber, C. (2014), “Central Command Architecture for

High-Order Autonomous Unmanned Aerial Systems”. Intelligent Information Manage-
ment, 6, 183-195. doi: 10.4236/iim.2014.64019.
Copyright: © 2014 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
38 Unmanned Aerial Vehicles (UAV) and Drones

block, and an autonomous flight controls block. The central command block
includes a staging process that converts an objective into tasks independent
of the vehicle (agent). The autonomous planning block contains a non-
iterative sequence of algorithms that govern routing, vehicle assignment,
and deconfliction. The autonomous flight controls block employs modern
controls principles, dividing the control input into a guidance part and a
regulation part. A novel feature of high-order central command, as this paper
shows, is the elimination of operator-directed vehicle tasking and the manner
in which deconfliction is treated. A detailed example illustrates different
features of the architecture.
Keywords: Unmanned Aerial Vehicles, Autonomy, Central Command,
High-Order Systems, Deconfliction

INTRODUCTION
Unmanned aerial vehicles (UAV) are gaining interest with relaxing
restrictions in civilian airspaces. Looking ahead, a systematic approach is
needed for autonomous unmanned aerial systems (AUAS) consisting of a
large number of vehicles (agents). This paper, which is the first in a series
of two papers, presents an easy-to-implement architecture for high-order
AUAS; the second paper presents flight test results. The different AUAS
approaches are referred to in this paper as command approaches with
central command at one extreme and individual command at the other. In
central command, the AUAS authority lies in one entity, and the vehicle
requires only limited information about its environment. Central command
is naturally suited to problems dictated by global objectives, such as search
and surveillance in precision agriculture, border security, law enforcement,
and wildlife and forestry services, to name a few. In individual command, the
authority lies in the individual vehicle. The vehicle requires local situational
awareness, communication with other vehicles, and understanding of system
objectives. Individual command is naturally suited to problems dictated
by local objectives that require coordination, like those found in air traffic
control problems and in coordinated pick and place problems.
When an AUAS operates multiple vehicles in close proximity some
method of preventing collisions becomes necessary. Collision avoidance,
or the real-time act of recognizing and maneuvering to avoid an impending
conflict, also called “see and avoid”, is a fundamental principle of current
flight operations and will be a requirement of future operations that share
manned airspace. In contrast, deconfliction, or the act of planning paths in
 Central Command Architecture for High-Order Autonomous ... 39

advance that do not conflict, is a central command level method of preventing

conflicts as events are planned. Collision avoidance and deconfliction are
not mutually exclusive; deconfliction prevents collisions between planned
or known events whereas collision avoidance prevents collisions between
unplanned events. Future systems will incorporate both methods. This paper
develops a robust method of deconfliction based on an assignment method
that produces non-crossing paths. This method is presented within an easy-
to-implement architecture that is scalable to high-order AUAS. The method
section introduces the basic architecture and the results section provides a
detailed example.
The requirements of an architecture under which high-order AUAS
become feasible is the subject of ongoing research. Bellingham, Tillerson,
et al., 2003 [1] separated their architecture into 1) routing, 2) vehicle
assignment of tasks, and 3) deconfliction, after which the vehicles receive
their directions. Cummings, Bruni, et al., 2007 [2] separated their architecture
into 1) central mission and payload management, 2) navigation, 3) vehicle
autopilots, and 4) flight controls. Shima and Rassmussen [3] describe a
pantheon of architectures across the three axes of command, information
availability and performance. Concerning user load, Adams, Humphrey, et
al., 2009 [4] studied the problems that arise with human attention when one
or more users give commands and monitor data. Donmez, Nehme, et al.,
2010 [5] showed that wait time increases with the number of vehicles in
AUAS that do not have full autonomous planning. These studies suggest that
full autonomy is a necessary requirement in high-order central command.
About computational requirements, routing and task assignment are
combinatorial problems that have prohibitive computational requirements
for high-order AUAS, necessitating heuristics. Gardiner, Ahmed, et al.,
2001 [6] reviewed real-time collision avoidance algorithms for multiple
UAVs and illustrated the computational complexities that limit the number
of vehicles and flight time. How, King, et al., 2004 [7] demonstrated a
simultaneous task assignment and deconfliction approach with 8 vehicles.
These studies suggest the use of heuristics and non-iterative algorithms to
reduce computational complexity. Pertaining to the question of the division
of labor between remote and on-board processing, it is natural to perform the
flight controls on-board to reduce communication throughput requirements
to remote transmission of off-board command decisions at perhaps 1 Hz
or less. Edwards and Silverberg, 2010 [8] exploited this division of labor
in an autonomous soaring flight demonstration. Concerning the choice
between acquiring on-board states from synthesized relative measurements
40 Unmanned Aerial Vehicles (UAV) and Drones

(using situational awareness) or from inertial measurements (like IMU and

GPS), Levedahl and Silverberg, 2005 [9] showed that acquiring on-board
states from inertial measurements is particularly well-suited to the central
command problem because of the sensitivity issues that arise in synthesized
relative measurements in high-order AUAS.

METHOD

The High-Order Central Command (HOCC) Architecture

The HOCC architecture is described by the block diagram shown in Figure
1. Drawing from the work of Bellingham, Tillerson, Richards, and Howe
[1] , the HOCC architecture is divided into central command, autonomous
planning, and autonomous flight controls. The central command block directs
the execution of objectives. It contains a staging process that converts the
high-level objective into assignable tasks. The autonomous planning block
is an efficient, non-iterative sequence of three processes called Routing
(R), Vehicle Assignment (V), and Deconfliction (D). A unique feature of
the architecture introduced in this paper, as shown below, is that while
the processes have internal iterations, they never require going back to a
previous process (external iterations).

Figure 1. Block diagram for the HOCC architecture.

 Central Command Architecture for High-Order Autonomous ... 41

The complexity of the individual processes used, namely the Visibility

Graph method, the A* method, the Hungarian method, are well known.
For example, the number of calculations in A* is on the order of nmE
log(V) where n is number of vehicles, m is number of tasks, E is number
of obstacle edges, and V is number of obstacle vertices and the number of
calculations in the Hungarian method is on the order of n4. The elimination
of external iterations, as described below, leads to a robust, computationally
efficient algorithm. Throughout this paper, we only consider vehicles that
are assignable to every task (homogeneity).

Staging
As shown in Figure 1, a user is provided information from an operational
area and other field parameters, and data Cj (j = 1, 2, …) generated by AUAV
payload Pj (j = 1, 2, …). Using this and other information, the user commands
an objective. In absence of fully autonomous planning algorithms that
remove the vehicle from the calculus, user load would be, at the very least,
proportional to the number of vehicles, which is prohibitive in high-order
systems. Thus a critical feature of the HOCC architecture introduced in
this paper is the decoupling of the objective from the vehicles and, instead,
the expression of the commands in terms of a relatively small number of
parameters. Operator commanded dynamic re-tasking of individual vehicles
no longer applies.
The staging process converts an objective into a set of tasks. This
paper considers spatial staging, such as passing over a point (for imagery
or delivery), a line scan (following a line segment) which could be part
of covering a wider region, and loitering (orbiting a point) which could
arise when a vehicle is waiting for further instructions or to monitor a point
of interest. The spatial tasks have starting points, so the conversion of an
objective into a set of tasks determines a set of spatial points that vehicles
need to reach. During the staging process the candidate routes are unknown
and the assignment of vehicles to tasks is unknown.

Routing
Autonomous planning is the second block of the HOCC architecture and
routing is its first process. Support tasks, such as launching or recovering
and refueling, are conducted by the administrator (A) who operates in the
background. The routing process itself is performed in two parts, mapping
the environment and choosing a path [10] . The first part produces a graph of
42 Unmanned Aerial Vehicles (UAV) and Drones

a vehicle-task route segment. In this paper, the Visibility Graph method [11]
is employed. Briefly, this method produces a graph of possible connections
between vehicles and obstacle vertices, tasks and vertices, and between the
vertices themselves, representing a graph of possible routes through the
environment. Note that search areas are regarded as obstacles to prevent
routes from passing through them, which can otherwise lead to conflicts.
After determining the graph between a vehicle point and a task, the second
part is to find the shortest route between them. This part is performed by the
A* method [12] . This method does not first find the different paths from
which the shortest path could be found. Instead, it searches in the direction
of the task, and can leave portions of the graph unsearched. The shortest
vehicle-task routes are collected into a matrix of vehicle-task route lengths
used in vehicle assignment. New routing is calculated each time a vehicle
completes a task and when a new objective is initiated.

Vehicle Assignment
A frequent goal in AUAS path planning, because of limited fuel and flight
duration requirements, is to determine vehicle routes by minimizing distance
of travel of individual vehicles. An important variation on this goal, which is
introduced in this paper, is the minimization of total distance of travel of all
of the vehicles (a min-sum assignment). The minimization of total distance
of travel leads to an importing routing principle: The routes determined by
minimizing total distance of travel do not cross. The mathematical proof of
this routing principle is presented in the Appendix. Although non-crossing is
guaranteed, the paths can still touch obstacles. Indeed, vehicle-task segments
can share vertices of obstacles, as shown in Figure 2 below.
As shown, two vehicle task routes share vertex C. The vehicle at A travels
to C and then to A*. The vehicle at B travels to C and then to B*. Notice that
the total distance of travel after the shared vertex, CA* + CB*, is the same
whether A travels to A* or to B*. Indeed, the solution to the minimization
problem is indeterminate when two routes share a vertex. The presence of
this indeterminacy, however, presents an opportunity. It allows the algorithm
to ensure that the vehicles do not conflict after passing the shared vertex C,
such as a task at A* requiring a vehicle to loiter for a short time, thereby
staying very near the path of the vehicle proceeding to B*. Assume that the
vehicles begin at the same time and that vehicle B reaches C before vehicle
A. The algorithm selects the vehicle that first reaches vertex C to travel to
the farthest point, in this case, the vehicle at B travels to B*. The inspection
of a potential conflict associated with a separation distance falling below
 Central Command Architecture for High-Order Autonomous ... 43

a minimal level is thus reduced to checking separation distances while the

vehicles are traveling along the two segments AC and BC before the shared
vertex. The desire to ensure that vehicle separation distances exceed required
minimums makes total distance of travel an important cost function in AUAS
path planning. Thus, the min-sum assignment enables rapid deconfliction of
planned paths by checking only shared vertices.
The assignments that minimize the total distance of travel are determined
in this paper by the Hungarian method [13] . It is likely that an objective
will have more tasks than the system has vehicles, or that a new objective
might be added. This leads to repeated application of the assignment
method. The assignments are updated each time a task is completed or when
a new objective arrives. Therefore, the number of assignments is equal to
the total number of tasks plus the number of objectives. The computational
cost of the Hungarian method is smaller than the computational cost of the
Visibility method and the A* method when obstacle vertices outnumber
tasks or vehicles.

Figure 2. Shared vertex.

Deconfliction
As described above and proven in the Appendix, potential conflicts in this
HOCC architecture are treated mainly by avoiding them in the first place (in
contrast with collision avoidance). As stated earlier, this was accomplished
by utilizing the non-crossing property of a min-sum assignment. Non-
crossing is a more powerful statement than deconfliction. A deconflicted
route means two vehicles must not occupy the same space at the same time,
but the non-crossing property shows that the entire length of the paths do
not cross at any time. This decouples vehicle speed from the path planning
and is useful in solving the potential conflicts in the vicinity of obstacle
vertices. By loosening any initial assumptions about vehicle speed, shared
44 Unmanned Aerial Vehicles (UAV) and Drones

vertices can be deconflicted by small speed adjustments without changing

the path of the vehicle. Because the assigned path is known not to intersect
with other paths, this small speed adjustment cannot cause other routing
conflicts, eliminating the need to re-check the final solution for conflicts
The elimination of route intersections leaves potential conflicts only in the
neighborhood of boundaries of obstacles that are shared by more than one
vehicle. When a potential conflict is identified, the first step is to determine
whether or not a conflict arises. The simplest action is to eliminate a conflict
by adjusting the speed of one or more vehicles without changing any
routes. Minor speed adjustments around the cruise speed are almost always
feasible. In extreme cases timing can be corrected by placing vehicles in
holding patterns, and is scalable to high-order AUAS when route segments
are sufficiently long. Other remedies, like route adjustment [9] , become a
last resort.

Autonomous Flight Controls

The autonomous planning block leads to vehicle assignments. The vehicle
assignments represent low frequency and bandwidth information that
can be performed off-board. The autonomous flight controls, in contrast,
represent higher frequency and bandwidth processes that are ideally suited
for on-board implementation. Best practices in autonomous flight controls
are reviewed below. The first step is to convert assigned routes to guidance
paths (G). At this stage routes are refined to account for turning radii and
other properties to enable vehicles to follow the routes precisely. Indeed,
slopes and curvatures of connected line segments are discontinuous at
route vertices and can’t be followed precisely. Circular arcs between line
segments [14] produce paths whose directions of travel (first derivatives) are
continuous. Further refinement produces paths whose turning radii (second
derivatives) are continuous, which allows time for the AUAV to change its
flight direction through roll. The next step is to regulate the guidance paths.
The k-th vehicle regulates its guidance path rk independent and without
awareness of other vehicles. An external disturbance uDk, like a wind load,
causes deviations from the actual flight path xk. The control surfaces are
adjusted by a guidance input uNk and a regulation input uRk. The guidance
input serves to fly the vehicle along its guidance path in the absence of
external disturbances and the regulation part compensates for errors. The
primary errors are caused by external disturbances but others errors originate
from prescribing non-smooth guidance paths, transducer error, and model
 Central Command Architecture for High-Order Autonomous ... 45

errors in the plant Lk. The guidance state is

received while following the actual state . Let

uk represent the control input vector. The equations governing the motion
of the vehicle, the autonomous navigator, the autonomous regulator and the
form of the control input are

where KGk is the guidance operator, KRk is the regulator operator, and
where the control input is the sum of a guidance input uGk, a regulator input
uRk, and a disturbance input uDk.
This is called the modern control form [15] . Note that the operator
notation above does not specify the realization, that is, whether the system
is represented by differential equations, difference equations, algebraic
equations, or a hybrid. Under ideal conditions (KGk = Lk and the guidance
input is smooth), the guidance input causes the vehicle to follow the
guidance state vector exactly. It follows that the input-output relationship
for the vehicle, the error, and the characteristic equation are:

where I is the identity operator and ek is the state error vector. The
characteristic equation is independent of the guidance state vector. Therefore,
the regulator gain matrix GRk determines the vehicle’s stability characteristics
(settling time, peak-overshoot, and steady-state error) independent of the
guidance state vector rk.
These best practices pertaining to autonomous flight controls are
summarized as an AUAS Flight Controls Principle: Assume that a realizable
guidance path has been produced and that a vehicle has the control authority
capable of maintaining the guidance path within certain limits placed on
settling time, peak overshoot, and steady-state error.
Then autonomous guidance and autonomous regulation can be designed
by separate and distinct methods.
Following best practices in autonomous flight controls lead to vehicle
paths that more closely follow their guidance paths, which is of particular
interest in autonomous systems that are not instrumented with on-board
collision avoidance systems.
46 Unmanned Aerial Vehicles (UAV) and Drones

RESULTS

Problem Statement
Four vehicles loiter in an air space that contains a no-fly zone (obstacle), as
shown in Figure 3. At time t = 0, an AUAS user commands the AUAS to
search area A. At time t = T, while the vehicles are searching area A, the user
gives a second command to search area B. The AUAS, once completing its
two objectives, is commanded to return to the original loiter points.

HOCC Solution
The scenario begins with the AUAS user commanding the first objective
to survey area A. The command to survey area B is made later. The staging
process converts the objective to search area A into 5 line segment tasks
after which the vehicles are tasked to return to their original loiter points
(see Figure 4). Each line segment is regarded as a task and a vehicle can
follow any line segment from the left or from the right.

Figure 3. Problem statement.

Figure 4. Part of the graph between a vehicle and the end point of a task.
 Central Command Architecture for High-Order Autonomous ... 47

Figure 4 shows the first step in determining the shortest route between any
one vehicle point and the end points associated with a task. There are two end
points associated with each task since the vehicle is allowed here to follow
a line segment from the left or the right. Consider the vehicle point and the
pair of end points associated with the bottom line segment, as shown. The
graph connects solid dots that designate allowable points and the squares that
designate obstacle vertices (points designated by circles are not allowed).
This figure only shows some of the possible connections in the graph. The
first step in determining the best route, when employing the visibility graph
method, is to produce a graph that connects the vehicle point to the two
end points. The number of different routes along the segments, as shown,
is a combinatorial problem, and even enumerating all of the possibilities
is computationally expensive in high-order AUAS. As an alternative, the
visibility graph method is used, recognizing that it does not consider all of
the possibilities.
After determining the possible graphs between a vehicle point and a
task, the next step is to find the shortest route between them. The shortest
vehicle-task routes between one of the vehicles and each of the five tasks
are shown in Figure 5. There are comparable shortest vehicle-task routes
between each of the other vehicles and the five tasks, as well (not shown).
It happens that the best end points are each on the right side of the search
area for the vehicle shown. You will see in Figure 6 that there is a different
vehicle for which its shortest vehicle task route is to an end point located on
the left side of the search area.
Once the shortest vehicle-task routes are determined, vehicles need to be
assigned to tasks. This is done by minimizing the total distance of travel of
the vehicles (see Figure 6).
In Figure 7, the vehicles are shown on their way to their first task when
a second objective arrives. The staging process turns the second search area
into 5 new line segment tasks. The vehicles are shown at the instant that the
new objective arrives.
As shown, two tasks are currently being executed, and their end-points
are no longer available to be assigned. The points that are available are
shown as solid dots. When no longer available, the end-points are indicated
by circles. Also, notice that past routes are indicated by thin lines and that
the proposed routes are indicated by thick lines. As shown, there will be a
total of ten tasks and two objectives so the total number of time intervals
will be twelve.
48 Unmanned Aerial Vehicles (UAV) and Drones

Figure 8 is shown at the same instant as Figure 7 but the newly calculated
routes are now shown. As shown, only the assignment of the green vehicle
changes and heads to the new search area. Note that vehicles in the process
of executing a task are assigned to new tasks from the end point of the
task currently being executed. When the new objective arrived, the blue and
yellow vehicles were performing tasks, so their proposed paths beyond their
current tasks were also included in the calculation. The proposed route of
the yellow vehicle is to the left end point of the task just below its current
task and the proposed route of the blue vehicle is to the right end of the task
just above its current task. The red vehicle, after the new calculation, is
assigned to the same end point as in the previous time interval.

Figure 5. Five shortest vehicle-task paths for one of the vehicles during the
1st time interval.

Figure 6. Routes over the 1st time interval.

Figure 7. A new objective during the 1st time interval initiates the 2nd time
interval.
 Central Command Architecture for High-Order Autonomous ... 49

Figure 8. Routes over 2nd time interval.

The yellow vehicle is the first vehicle to complete a task. The completion
of its task triggers the start of the third time interval. Figure 9 shows the
vehicles at this instant, the past routes (thin lines), and the routes proposed
over this time interval (thick lines). Notice that, because the red vehicle is
executing its task, its next assignment is being considered. As the remaining
tasks in the original search area are assigned, it is easy to see that the blue
vehicle has a shorter route to the new area than the red or yellow vehicles.
Thus the red vehicle is assigned the task previously allotted to blue, and
blue’s assignment changes to the new search area. The assignments given
to the green and the yellow vehicles are the same as in the previous time
interval.
In Figure 10, the yellow vehicle has just completed its 2nd task. The red
and blue vehicles completed their first tasks, so a total of 4 tasks have been
completed at this instant. Indeed, Figure 10 shows the start of the 6th time
interval. As stated earlier, routes never cross but they can share obstacle
vertices (squares). Indeed, the blue, red, and yellow vehicles are approaching
the top right vertex. The presence of shared vertices prompts actions in the
deconfliction block. The first action is to check whether there is a conflict
in the neighborhood of the shared vertex. The checking of this potential
conflict reveals that the blue vehicle reaches the vertex first, then the yellow
vehicle and finally the red vehicle. Moreover, the separation distances were
all determined to be larger than the required minimum. Recall to avoid
potential conflicts, that the vehicle that first reaches the shared vertex is
assigned the route to the farthest end point. Thus, the proposed route of
the blue vehicle is to the right-most task, the proposed route of the yellow
vehicle is to the next right-most task, and the proposed route of the red
vehicle is to the third right-most task.
50 Unmanned Aerial Vehicles (UAV) and Drones

In Figure 11, the green vehicle has just completed its task, triggering the
9th time interval. The green vehicle has been assigned the task of following
the middle task in search area B, completing the available tasks and leaving
the red vehicle unassigned. This unassigned vehicle is assigned to a loiter
point to await further tasks.
In Figure 12, the green vehicle has just completed a task, triggering the
10th time interval. During this time interval, the bottom left vertex of the
2nd search area is a shared vertex. The green vehicle arrives at the vertex
before the blue vehicle so the green vehicle returns to the loiter point that is
farther from the vertex.
In Figure 13, the yellow vehicle completes the last task, triggering the
12th time interval. The red vehicle is already at a loiter point, the green vehicle
is on route to a loiter point leaving the yellow and blue vehicles that happen
to be tasked to pass the bottom left vertex in search area B. The blue vehicle
reaches the vertex first, so it is routed to the farthest remaining vertex.

Figure 9. Routes over the 3rd time interval.

Figure 10. Routes over the 6th time interval.

 Central Command Architecture for High-Order Autonomous ... 51

Figure 11. Routes over the 9th time interval.

Figure 14 shows the complete routes of all four vehicles, with the vehicles
having returned to the loiter points.

SUMMARY AND CONCLUSIONS

This paper presented an easy-to-implement central command architecture that
is well suited to a large number of unmanned aerial vehicles―overcoming
present limitations pertaining to operator task load, computational complexity
and vehicle deconfliction. The architecture consists of a central command
block, an autonomous planning block, and an autonomous flight controls
block. Building on results found in the literature pertaining to

Figure 12. Routes over the 10th time interval.

Figure 13. 12th (last) time interval.

52 Unmanned Aerial Vehicles (UAV) and Drones

Figure 14. The end.

Larry M. Silverberg, Chad Bieber AUAS, the central command block is
fully autonomous thereby overcoming user load limitations associated
with a large number of vehicles. The autonomous planning part is a non-
iterative sequence of algorithms that govern routing, vehicle assignment,
and deconfliction. The non-iterative nature of the algorithms overcomes
computational limitations. The routing exploits a new routing principle that
naturally prevents vehicle routes from crossing. An example illustrates the
different features of the architecture. This paper is the first in a series of two
papers; the second will report on flight test results.

APPENDIX: THE SHORTEST LINES BETWEEN TWO

POINT SETS DO NOT CROSS
Definition: Let two curved line segments intersect. The intersection point is
a crossing point if the corresponding tangent lines are not parallel and the
point is located in the interior of the line segments (is not an end point of a
line segment).
Definition: A planar domain D in R2 refers to a subset of R2 that can be
bounded externally by a closed curve and internally by closed curves.
Definition: Refer to the set of points as point set A(n), and to
the set of points as point set B(m), in which points Qi and qi are
contained in a planar domain D and The curved line segments in
D that connect each point in A(n) to a unique point in B(m) compose the
connecting set C(n, m) of A(n) and B(m). Each curved line segment is called
a connecting line segment.
 Central Command Architecture for High-Order Autonomous ... 53

Definition: The connecting line segments in a connecting set C(n, m) of

A(n) and B(m) of the shortest total length are called minimal
connecting line segments.

Lemma: Line Segments

Consider the following three well-known results:
1). The shortest length L of a curved line segment is produced
by the straight line segment between the two end points
(see Figure A1(a)). This is more simply stated as “The shortest
distance between two points is a straight line.”
2). The length of any one side of a triangle is smaller than the total
length of the other two sides (see Figure A1(b)). This result
follows from the Pythagorean Theorem.
3). Any point inside a triangle can be connected to any two of its
vertices to form a pair of connected line segments. The length of
these line segments is larger than the length of the side that connects
the two vertices and is smaller than the length of the other two sides
(see Figure A1(c)). This result follows from the second result.

Theorem: The Shortest Lines between Two Point Sets Do Not

Cross
Minimal connecting line segments in any connecting set C(n, m) do not
cross. This is more simply stated as “The shortest lines between two point
sets do not cross.”
Proof: Figure A2 illustrates the theorem for a planar domain that is
bounded internally by rectangular-shaped closed curves. The connecting
line segments are colored and the minimizing line segments are black. The
actual proof begins next.
Begin the proof by ordering the points in A(n) and B(m) so that the
minimal connecting line segments in C(n,m) are . The shortest
total length is Next, remove connecting line
segment Q1q1 from C(n, m), remove Q1 from A(n), and remove q1 from B(m)
to produce the connecting set C(n 1, m 1) of A(n 1) and B(m 1). It is now
shown that are minimal connecting line segments in C(n 1, m -1).
In other words, let’s show that the following key equation holds:
54 Unmanned Aerial Vehicles (UAV) and Drones

Figure A1. Three well-known results.

To show this, assume that are not minimal connecting line

segments in C(n ? 1, m ? 1), in which
case there exists a total length L* that is smaller than Lmin(n 1, m 1). This
leads to the contradictory result:
Lmin(n, m) = |Q1q1| + L* < |Q1q1| + Lmin(n 1, m 1) = Lmin(n, m). Indeed,
are minimal connecting
line segments in C(n 1, m -1).
The key equation can be invoked in a sequence that begins with
n connecting line segments, and ends with any two. Furthermore, n line
segments do not cross if any combination of two does not cross. Thus, it
remains to prove the theorem for n = 2 connecting line segments. The case
of n = 1 line segment should be proven, too; in this case the crossing of the
curved line is with itself. Now, consider the case of n = 2.
Two crossing line segments are shown in Figure A3(a) and a view of the
infinitesimal region surrounding the crossing point is shown in Figure A3(b)
along with two new short-circuiting line segments. Say that the two straight
line segments Aa and Bb shown in Figure A3(b) contribute to the total length
L by an infinitesimal amount dL. As shown, by result (1) in the Lemma, the
total length of the short-circuiting line segments can be used to eliminate the
crossing point and switch the assignment from Aa and Bb to the assignment
of Ab and Ba while shortening dL and therefore shortening L. But L was
the smallest total length, which is contradictory. It follows that the two line
segments could not have crossed. The two line segments in Figure A3(a) can
be regarded as one line segment, so one line segment could not have crossed
on itself, either.
 Central Command Architecture for High-Order Autonomous ... 55

Figure A2. Illustration (n = m = 19 case).

Figure A3. Crossing line segments.

56 Unmanned Aerial Vehicles (UAV) and Drones

REFERENCES
1. Bellingham, J., Tillerson, M., Richards, A. and How, J.P. (2003) Multi-
Task Allocation and Path Planning for Cooperative UAVs. In: Butenko,
S., Murphey, R. and Pardalos, P.M., Eds., Cooperative Control: Models,
Applications and Algorithms, Kluwer Academic Publishers, Boston,
23-39.   
2. Cummings, M.L., Bruni, S., Mercier, S. and Mitchell, P.J. (2007)
Automation Architecture for Single Operator, Multiple UAV Command
and Control. The International Command and Control Journal, 1, 1-24.   
3. Shima, T. and Rassmussen, S. (Eds.) (2009) UAV Cooperative
Decision and Control. Society for Industrial and Applied Mathematics,
Philadelphia. http://dx.doi.org/10.1137/1.9780898718584   
4. Adams, J.A., Humphrey, C.M., Goodrich, M.A., Cooper, J.L., Morse,
B.S., Engh, C. and Rasmussen, N. (2009) Cognitive Task Analysis for
Developing Unmanned Aerial Vehicle Wilderness Search Support.
Journal of Cognitive Engineering and Decision Making, 3, 1-26. http://
dx.doi.org/10.1518/155534309X431926   
5. Donmez, B.D., Nehme, C. and Cummings, M.L. (2010) Modeling
Workload Impact in Multiple Unmanned Vehicle Supervisory
Control. IEEE Transactions on Systems, Man, and Cybernetics, Part
A: Systems and Humans, 40, 11801190. http://dx.doi.org/10.1109/
TSMCA.2010.2046731   
6. Gardiner, B., Ahmad, W., Cooper, T., Haveard, M., Holt, J. and Biaz, S.
(2011) Collision Avoidance Techniques for Unmanned Aerial Vehicles.
Auburn University, Auburn.   
7. How, J., Ellis King, E. and Kuwata, Y. (2004) Flight Demonstrations
of Cooperative Control for UAV Teams. AIAA 3rd “Unmanned
Unlimited” Technical Conference, Workshop and Exhibit, Chicago,
20-23 September 2004.   
8. Edwards, D. and Silverberg, L.M. (2010) Autonomous Soaring: The
Montague Cross-Country Challenge. Journal of Aircraft, 47, 1763-
1769. http://dx.doi.org/10.2514/1.C000287   
9. Levedahl, B. and Silverberg, L.M. (2005) Autonomous Coordination
of Aircraft Formations Using Direct and NearestNeighbor Approaches.
Journal of Aircraft, 42, 469-477. http://dx.doi.org/10.2514/1.6868   
10. Tsourdos, A., White, B. and Shanmugavel, M. (2011) Cooperative Path
Planning of Unmanned Aerial Vehicles. Wiley, New York.   
 Central Command Architecture for High-Order Autonomous ... 57

11. Lozano-Perez, T. and Wesley, M.A. (1979) An Algorithm for Planning

Collision Free Paths among Polyhedral Obstacles. Communications of
the ACM, 22, 560-570. http://dx.doi.org/10.1145/359156.359164   
12. Hart, P.E., Nilsson, N.J. and Raphael, B. (1968) A Formal Basis for the
Heuristic Determination of Minimum Cost Paths. IEEE Transactions
on Systems Science and Cybernetics, 4, 100-107. http://dx.doi.
org/10.1109/TSSC.1968.300136   
13. Kuhn, H.W. (1955) The Hungarian Method for the Assignment
Problem. Naval Research Logistics Quarterly, 2, 8397.   
14. Dubins, L.E. (1957) On Curves of Minimal Length with a Constraint on
Average Curvature, and with Prescribed Initial and Terminal Positions
and Tangents. American Journal of Mathematics, 79, 497-516. http://
dx.doi.org/10.2307/2372560   
15. Levedahl, B. and Silverberg, L.M. (2009) Control of Underwater
Vehicles in Full Unsteady Flow. IEEE Journal of Oceanic Engineering,
34, 656-668.   
 CHAPTER

4
Quadcopter Design for Payload
Delivery

Gordon Ononiwu, Ondoma Onojo, Oliver Ozioko, Onyebuchi

Nosiri
Department of Electrical/Electronic Engineering, Federal University of
Technology, Owerri, Nigeria

ABSTRACT
This paper presents the design and implementation of a quadcopter capable
of payload delivery. A quadcopter is a unique unmanned aerial vehicle
which has the capability of vertical take-off and landing. In this design, the
quadcopter was controlled wirelessly from a ground control station using

Citation: Ononiwu, G. , Onojo, O. , Ozioko, O. and Nosiri, O. (2016), “Quadcop-

ter Design for Payload Delivery”. Journal of Computer and Communications, 4, 1-12.
doi: 10.4236/jcc.2016.410001.
Copyright: © 2016 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
60 Unmanned Aerial Vehicles (UAV) and Drones

radio frequency. It was modeled mathematically considering its attitude and

altitude, and a simulation carried out in MATLAB by designing a proportional
Integral Derivative (PID) controller was applied to a mathematical model.
The PID controller parameters were then applied to the real system. Finally,
the output of the simulation and the prototype were compared both in the
presence and absence of disturbances. The results showed that the quadcopter
was stable and able to compensate for the external disturbances.
Keywords: Quadcopter, PID, Controller, Payload, Delivery, Unmanned,
Aerial, Vehicle

INTRODUCTION
Quadcopters are unique unmanned aerial vehicles which have a vertical
take-off and landing ability [1]. In the past, it was primarily used in the
military, where it was deployed in hostile territory, to reduce pilot losses.
The recent reduction in cost and feature size of semi-conductor logic has
led to the application of quadcopters in many other areas like weather
monitoring, forest fire detection, traffic control, cargo transport, emergency
search and rescue, communication relaying, etc. Quadcopters operate in
fixed rotor propulsion mode, an advantage over the traditional helicopter.
They have four rotors which are arranged in such a way that rotors on
transverse ends rotate in the same direction, while the other two operate
in the opposite direction [1]. Changing the parameters of the attitude (roll,
pitch and yaw), and altitude determines the behavior and position of the
quadcopter. The throttle on each rotor changes its attitude. The UAVs can
be broadly classified into two categories: fixed wing versus rotary wing,
each with their own strengths and weaknesses. For instance, the fixed-wing
UAVs usually have high speed, and heavy payload, but they must maintain
a continuous forward motion to remain in the air [2].
The applications for quadcopters have recently grown significantly, and
related works are presented here. [3] presented an altitude controller design
for UAVs in real time applications. The author described a controller design
method for the hovering control of UAVs and automatic vertical take-off
systems. In 2012, the attitude and altitude dynamics of an outdoor quadrotor
were controlled using two different structures of proportional integral
derivative (PID) controllers [4]. The tests showed that the two structures
were successful. [5] presents a Quadcopter for surveillance monitoring. In
[6], the design and control of VertiKUL, a Vertical Take-Off and Landing
(VTOL) transitioning tail-sitter Unmanned Aerial Vehicle (UAV) was
 Quadcopter Design for Payload Delivery 61

presented. The UAV is capable of hover flight and forward flight, and its
main application is for parcel delivery.

METHODOLOGY
In this work, aquadcopter was first modeled mathematically, after which a
simulation was carried out in MATLAB by designing a PID controller, which
was applied to the mathematical model. The PID controller parameters were
then applied to the real system. Finally, the output of the simulation and the
real system were compared. In this paper, attitude and altitude dynamics
of a quadcopter were considered. Therefore, roll, pitch, yaw, and altitude
dynamics were modelled.

Mathematical Model of Attitude Dynamics

One of the aims of the work is to stabilize the quadcopter in the hovering
condition. The following assumptions were made [4]-[7]: i) System structure
is supposed to be rigid and symmetrical. ii) Earth fixed reference frame is
assumed to be inertial. iii)The rotors are rigid, i.e. no blade flapping occurs.
Attitude of the quadcopter is defined by roll, pitch and yaw angles;
namely φ , θ , and ψ , respectively. Angular velocity components in body
reference frame are p, q, and r, respectively. F1, F2, F3, and F4 are the thrust
forces from the four motors, which are generated by propellers. Axes and
states are represented in Figure 1. It is possible to obtain a rotation matrix
with the following equation:
(2.1)
where R is the matrix transformation dynamics of the model quadcopter
moving from landing position to a fixed position in space.
Therefore, substituting Rx , Ry , and Rz into Equation (2.1), we get the
system matrix:

(2.2)
The angular velocity of the quadcopter can be written as:

(2.3)
62 Unmanned Aerial Vehicles (UAV) and Drones

It is possible to write state equations of the Euler angles in terms of angular

velocity components, p, q, and r by using rotation matrices.
(2.4)
(2.5)
(2.6)
Inertia matrix can be written below as:

(2.7)

Figure 1. Schematic representation of the system.

Inertia matrix, I is assumed to be a diagonal matrix (i.e. the mass
products of inertia terms are assumed to be 0). Thrust forces generated by
each motor-propeller pair are denoted by F1, F2, F3 and F4. Therefore,
resultant moments about roll, pitch and yaw axes are given below:
 Quadcopter Design for Payload Delivery 63

(2.8)

(2.9)
(2.10)
Therefore,

(2.11)
Hence, state vector and state equations considering the yaw, roll and the
pitch dynamics of the platform are expressed below:

(2.12)

(2.13)
Also, by applying the force and moment balance laws, we have the
following quadcopter motion formulation:
64 Unmanned Aerial Vehicles (UAV) and Drones

(2.14)

(2.15)

(2.16)
where Ki is the drag coefficient (assume zero since drag is negligible at
low speed).
The angle movement of Quadcopter is given by

(2.17)

(2.18)

Controller Input
Quadcopter has four controller input U1, U2, U3, and U4.
(2.19)
(2.20)
(2.21)

(2.22)
where,
Th1 = thrust generated by each motor
l = force to moment scaling factor
Ii = moment of inertia with respect to the axes.
The second derivative of Euler angles are:
(2.23)
(2.24)
(2.25)
 Quadcopter Design for Payload Delivery 65

Motor and Propeller Model

Brushless motors are used as the actuators of propellers of the quadcopter.
Four brushless motors in the system are driven via Electronic Speed
Controllers (ESC). ESC’s converts PWM signals into a three-phased signal,
which rotates the motor continuously. Motor and propeller unit models are
identified experimentally. These models are algebraic, steady state ones.
Thrust force generated due to the rotation of propellers can be calculated
with the equation given below:
(2.26)
where b is thrust factor which is determined experimentally.

Mathematical Model of Altitude Dynamics

Total thrust force and weights are the acting forces in z direction. Translational
acceleration of the quadcopterin z-axis can be obtained, using the following
formula;

(2.27)
This equation reveals position in z-axis and velocity in z-axis as two
states.

The PID Controller Design

The PID is a popular way of controlling the quadcopter, and one widely
used strategy to control a quadcopter with a PID is to create an individual
closed-loop system for each independent axis of each sensor. That is, we
would have the same structure depicted in Figure 2 for each sensor axis
where y ( t) would be the sensor output (angular velocity or acceleration of
each axis) and u ( t) the desired value to that axis. This way we would have
three different PID controllers for the gyroscope (one for each axis) and the
input to each control loop would be an angular velocity reference. That is,
making all references equal to zero would make the quadcopter preserve all
its rotation angles. Other three distinct PIDs would then be used to control
the accelerometer axes. Therefore, setting zero to x and z axes would make
the quadcopter hover (all weight force would be concentrated in y axis).
Another widely used strategy is to mix the accelerometer and gyroscope
signals to find a single angle value for each axis. That is, combining the
gyroscope information with the accelerometer information, one can find a
66 Unmanned Aerial Vehicles (UAV) and Drones

unique angle value for each axis that can represent an absolute angle of
rotation or a variation with respect to the previous time step. Therefore,
one can use a total of only three closed-loop systems instead of six. Also,
there are three multiple ways of combining those signals, and the control
system efficiency will directly depend on which way is chosen. Since the
sensors can have errors (e.g. noise and drift), a method that minimizes the
main problems of each sensor is chosen and the best features of each can be
combined.

Figure 2. A typical PID controller internal blocks.

Therefore, the control signals generated by each closed-loop structure
must be combined before sending to the motors. A commonly adopted way
of combining those signals is given below:

(2.28)

(2.29)

(2.30)

(2.31)
where, mi = signal sent to the ith motor, T = throttle signal and
control signals (for pitch, roll and yaw axes respectively)
It can be seen in the equations above that those signals are used in a
differential mode due to the plant symmetry. Thus, another important detail
is that the signals of each term inside each equation depends on each axis
reference (e.g. whether positive pitch rotation is clockwise or counter-
clockwise) as well as where each motor is positioned. So one must know
the motor configuration (whether plus + or cross ×) before writing the
equations. Finally, each motor will be individually controlled in an open-
loop configuration which can already lead us to a flyable quadcopter.
 Quadcopter Design for Payload Delivery 67

SIMULATION
The quadcopter design details were used to produce a model to be constructed
in the simulation window. Some effects (like aerodynamic effects such as
blade flapping) were ignored or dramatically simplified, thus, a limitation
to the model. The system model contains five major blocks (systems) each
containing its own subsystem or unit blocks that describe the overall system
dynamics of our quadcopter. The system blocks include: i) Path Command
block; ii) Position Controller block; iii) Attitude Controller block; iv)
Quadcopter Control mixing block; v) Quadcopter Dynamics block.

Path Command and Position Controller

The path command block is used to set the desired position of quadcopter in
terms of x, y and z coordinates as well as the yaw (for directioning on the x-y
plane). These parameters serve as the set points for the position controller.
These variables are loaded as Matlab codes into the simulation through the
MATLAB workspace. The position controller takes inputs from the Path
command as its set point, and also receives the current position (state) of the
quadcopter as feedback. It then compares this feedback with the set point to
give the error.
The position error in the body frame is computed, and maps the body
frame desired velocity. This desired velocity is then mapped to a desired
attitude. The essence of this rotation is because the Inertial-frame velocity is
used as the state to be controlled (Path command input).

The Attitude Controller

The attitude controller, like the position controller, takes feedback from the
output of the system as one of its inputs, but takes its second input from
the position controller. This block contains the full proportional integral
derivative (PID) controller which does the final attitude correction. The
feedback to this block contains the current attitude, and angular velocity
of the quadcopter in x, y and z axes. The controller essentially compares
the current state with the set point, and attempts to bring these various
parameters to the set point. The integral part of the PID ensures that the
parameters remain at set point once achieved.
Figure 3(a) and Figure 3(b) are the Simulink blocks for Pitch and Yaw,
including their corresponding attitude correction outputs.
68 Unmanned Aerial Vehicles (UAV) and Drones

Figure 3. (a) Simulink block for the pitch (Theta) correction; (b) Simulink
block for the yaw (Psi) correction.

Quadcopter Control Mixing

There are two quadcopter configurations available, the quad + configuration
and the quad × configuration. The quadcopter control mixing block does the
control signal mixing depending on the quadcopter configuration used (plus
+ or cross ×). It gets the corrected parameters for Pitch (θ ), Roll (ψ ), Yaw
(φ ) and Altitude (Z) from the attitude controller and generates the necessary
throttle commands to actuate each motor (m1, m2, m3, m4).
(3.1)

(3.2)

(3.3)

. (3.4)
 Quadcopter Design for Payload Delivery 69

Quadcopter Dynamics
The quadcopter dynamics block contains two major subsystem blocks: The
motor dynamics block, and The State Equations block. The Primary function
of the Motor dynamics block is to convert the percentage throttle command
to the motor RPM.
The motor dynamics block contains the state space model;
(3.5)
(3.6)
where,

The four RPM values generated are fed into the State equations block which
contains a level two S-function block, where the flight dynamic state equations
are loaded. This S-function is written in MATLAB. The S-function block
holds the initial conditions (Table 1) used for the simulation. Its parameters are
the quadcopter model functions and the input initial conditions. The outputs,
describes the physical behavior of the quadcopter.
The main function of the S-function block is to use various parameters
fed into it, to calculate the quadcopter outputs (Angular velocities, linear
velocities, position (along the x, y and z axes) and attitude (yaw, pitch and
roll)) which are then used for the simulation plots and 3D animation. Also, a
disturbance block was added as input to the S-function block to simulate for
response to external forces which might act on the quadcopter body during
flight and how it can compensate and recover from these disturbances whilst
still trying to maintain flight stability (set point).
The data used for the quadcopter simulation was based on the design
model presented. The individual parameters of the components that make up
the quadcopter were carefully calculated. Some parameters such as Torque,
Thrust and Time Constants where estimated from already measured data of a
similar system. This estimation is due to the unavailability of necessary test
bench (and instruments) required to carefully monitor and measure these
parameters. Since our design model is a quadcopter for payload delivery,
our simulation focused on the following: 1) Load test (maximum allowable
load on-board); 2) Hover test (with load and without load); 3) Disturbance
test (with load and without load).
70 Unmanned Aerial Vehicles (UAV) and Drones

The PID Control gains were carefully tuned and the appropriate gain
value was chosen based on the simulation results obtained. The PID gains
used in the Attitude controller are presented in Table 1 while the PD gains of
the position controller are presented in Table 2.

SIMULATION RESULTS AND DISCUSSION

The simulation was run for 60 seconds and several graphs were
generated. Figure 4 shows the result of the simulation with no external
disturbance added. The Path command was such that the quadcopter will
move a diamond-shaped path. P, Q and R are angular velocities about the
X, Y and Z axes respectively; Phi, theta and Psi are the Roll, Pitch and Yaw
angles;
U, V and W are the translational velocities along the X, Y and Z axes
respectively; X, Y, and Z define the position of the quadcopter along the X,
Y and Z axes respectively.

Table 1. Attitude controller PID gains.

Table 2. The position controller PD gains.

 Quadcopter Design for Payload Delivery 71

Figure 4. Result of the simulation without disturbance.

Hover Test
The hover simulation of a quadcopter in flight describes its ability to float
at/about a fixed point in air. Hover test was done at 10 feet and the results
are given below.
The graphs shown in Figure 4 represent a stable system. At first, there
was an overshoot, but with time the controller sets each of these parameters
to set point.

Disturbance Test
Disturbance was introduced to the quadcopter flight simulation which could
represent the action of external force like wind on the body of the quadcopter
while in flight. The graph in Figure 5 represents the system without any
disturbance. However, as can be seen from the graph in Figure 6, when a
disturbance was introduced, the quadcopter was able to compensate properly.
Therefore, the PID controller logic is functional and it can be implemented
for the actual system.
The effect of a disturbance to the flight path was further tested with
regards to linear and angular velocity is presented in Figure 7. The system
acknowledged the disturbance but was able to compensate adequately, and
return to a stable flight.
72 Unmanned Aerial Vehicles (UAV) and Drones

Load Test
Since the quadcopter design model is for payload delivery, the simulation
shows how an increase in the gross mass of the quadcopter affects its motor
speed. The result is presented in Figure 8 and Figure 9. Again, the system
was able to maintain a stable flight.

Figure 5. Motor throttle vs motor speed.

Figure 6. Motor throttle vs motor speed (disturbance test).

CONCLUSION AND FUTURE WORK

A stable 1.26 kg quadcopter for payload delivery was designed, simulated
and implemented. The design and simulation, which utilized a PID controller,
was able to yield results that showed good compensation in the presence of
 Quadcopter Design for Payload Delivery 73

external disturbances. The Quad X structural model was used to enable the
Quadcopter to gain more speed over a short period of operation, as timely
delivery of payload is an important factor. Future study would be to equip the
quadcopter with a solar powered system, and to encrypt the communication
between the quadcopter and the ground station.

Figure 7. Simulation result graphs (disturbance test).

Figure 8. Motor throttle vs motor speed (load test).

74 Unmanned Aerial Vehicles (UAV) and Drones

Figure 9. Simulation result graphs (load test).

CONFLICTS OF INTEREST
The authors declare no conflicts of interest.
 Quadcopter Design for Payload Delivery 75

REFERENCES
1. Gupte, S., Mohandas, P.I.T. and Conrad, J.M. (2012) A Survey of
Quadrotor Unmanned Aerial Vehicles. 2012 Proceedings of IEEE,
Orlando, FL, 15-18 March 2012, 1-6. http://dx.doi.org/10.1109/
secon.2012.6196930
2. Zeng, Y., Zhang, R. and Lim, T.J. (2016) Wireless Communications
with Unmanned Aerial Vehicles: Opportunities and Challenges.
3. Mustapa, M.Z. (2015) Altitude Controller Design for Quadcopter UAV.
Jurnal Teknologi.
4. Guclu, A. (2012) Attitude and Altitude Control of an Outdoor
Quadrotor. The Graduate School of Natural and Applied Science,
Atilim University, 76.
5. Ononiwu, G.C., Onojo, O.J., Chukwuchekwa, N. and Isu, G.O.
(2015) UAV Design for Security Monitoring. International Journal of
Emerging Technology & Research, 2, 16-24.
6. Hochstenbach, M. and Notteboom, C. (2015) Design and Control of
an Unmanned Aerial Vehicle for Autonomous Parcel Delivery with
Transition from Vertical Take-off to Forward Flight. International Journal
of Micro Air Vehicles, 7, 395-405. http://dx.doi.org/10.1260/1756-
8293.7.4.395
7. Min, B.-C., Hong, J.-H. and Matson, E.T. (2011) Adaptive Robust
Control (ARC) for an Altitude Control of a Quadrotor Type UAV
Carrying an Unknown Payloads. 2011 11th International Conference
on Control, Automation and Systems, Korea, 26-29 October 2011,
1147-1151.
 SECTION 2
TRAJECTORY CONTROL
TECHNIQUES
 CHAPTER

5
Advanced UAV Trajectory Generation:
Planning and Guidance

Antonio Barrientos, Pedro Gutiérrez and Julián Colorado

Universidad Politécnica de Madrid – (Robotics and Cybernetics Group),
Spain

INTRODUCTION
As technology and legislation move forward (JAA & Eurocontrol, 2004)
remotely controlled, semi-autonomous or autonomous Unmanned Aerial
Systems (UAS) will play a significant role in providing services and
enhancing safety and security of the military and civilian community at
large (e.g. surveillance and monitoring) (Coifman et al., 2004). The potential

Citation: Antonio Barrientos, Pedro Gutierrez and Julian Colorado (January 1st 2009).
“Advanced UAV Trajectory Generation: Planning and Guidance”, Aerial Vehicles,
Thanh Mung Lam, IntechOpen, DOI: 10.5772/6467.
Copyright: © 2009 The Author(s). Licensee IntechOpen. This chapter is distributed
under the terms of the Creative Commons Attribution-NonCommercial-ShareAlike-3.0
License
80 Unmanned Aerial Vehicles (UAV) and Drones

market for UAVs is, however, much bigger than just surveillance. UAVs
are ideal for risk assessment and neutralization in dangerous areas such as
war zones and regions stricken by disaster, including volcanic eruptions,
wildfires, floods, and even terrorist acts. As they become more autonomous,
UAVs will take on additional roles, such as air-to-air combat and even
planetary science exploration (Held et al., 2005).
As the operational capabilities of UAVs are developed there is a
perceived need for a significant increase in their level of autonomy,
performance, reliability and integration with a controlled airspace full of
manned vehicles (military and civilian). As a consequence researchers
working with advanced UAVs have moved their focus from system
modeling and low-level control to mission planning, supervision and
collision avoidance, going from vehicle constraints to mission constraints
(Barrientos et al., 2006). This mission-based approach is most useful for
commercial applications where the vehicle must accomplish tasks with a
high level of performance and maneuverability. These tasks require flexible
and powerful trajectory-generation and guidance capabilities, features
lacking in many of the current commercial UAS. For this reason, the
purpose of this work is to extend the capabilities of commercially available
autopilots for UAVs. Civil systems typically use basic trajectory-generation
algorithms, capable only of linear waypoint navigation (Rysdyk, 2003),
with a minimum or non-existent control over the trajectory. These systems
are highly constrained when maneuverability is a mission requirement. On
the other hand, military researchers have developed algorithms for high-
performance 3D path planning and obstacle avoidance (Price, 2006), but
these are highly proprietary technologies that operate with different mission
constraints (target acquisition, threat avoidance and situational awareness)
so they cannot be used in civil scenarios.
This chapter presents a robust Trajectory Generation and Guidance
Module (TG2M), a software tool capable of generating complex six-degrees-
of-freedom trajectories in 3D space with velocity, acceleration, orientation
and time constraints. The TG2M is an extension module to the Aerial
Vehicle Control Language (AVCL), a software architecture and interpreted
language specification that address the issues of mission definition, testing,
validation and supervision for UAVs (Barrientos et al., 2006). The AVCL
platform incorporates a 3D visual simulator environment that uses a
Geographic Information System (GIS) as the world’s data model. The GIS
backend means all objects are geo-referenced and that several official and
commercial databases may be used for mission planning (roads, airports,
 Advanced UAV Trajectory Generation: Planning and Guidance 81

power lines, crop fields, etc.). The language specification contains a wide
set of instructions that allow the off/on-line supervision and the creation of
vehicle-independent missions.
The TG2M is designed to model 3D cartesian paths with parametric
constraints. The system uses two types of mathematical descriptors for
trajectories: analytical functions and polynomial interpolation. The two
main contributions of the module are its geometrical representation of the
trajectory and its parametric definition. Simple maneuvers like lines and
circumference arcs are created with analytical functions that constrain the
geometry of the desired path; then the parametric constraints are applied.
These constraints are typically kinematic: constant velocity and acceleration,
trapezoidal curves to accelerate or stop, etc. More complex maneuvers are
described with polynomial interpolation and fitted to the critical control
path points, meeting desired position, time and velocity constraints. These
polynomial functions are based on third and fourth order splines with fixed
boundary conditions (for example initial and final velocities), which join all
control points with a continuous and smooth path (Jaramillo-Botero et al.,
2004).
The section 2 of this chapter presents some of the current techniques
extracted from the military aerial survey applications, showing complex
mission-planning tools capable of addressing the mission-specific constraints
required. Within those approaches, this section also introduces a brief
description of the AVCL architecture, describing its components, modules,
and the main features provisioned, addressing a novel way of human-mission
planning definition and testing. The section 3 introduces the AVCL built-in
TG2M framework, describing the different techniques used for the trajectory
planning and guidance of UAVs, as well as the mathematical treatment of
these methods (analytical functions and polynomial interpolation). On the
other hand, simulation-based results (see section 4) using a mini-helicopter
simulator embedded into the AVCL environment will show the capabilities
of the TG2M while flying aggressive and simple maneuvers. Last but
not least, “final observations” in section 5 includes comments about the
TG2M framework and the upcoming additions to the TG2M under current
development.

MOTION-PLANNING METHODOLOGIES
A planning algorithm should provide feasible and flyable optimal trajectories
that connect starting with target points, which should be compared and
82 Unmanned Aerial Vehicles (UAV) and Drones

valued using specific criteria. These criteria are generally connected to the
following major concerns, which arise during a plan generation procedure:
feasibility and optimality. The first concern asks for the production of a plan
to safely “move” the UAV to its target state, without taking into account the
quality of the produced plan. The second concern asks for the production
of optimal, yet feasible, paths, with optimality defined in various ways
according to the problem under consideration (LaValle, 2006). Even in
simple problems searching for optimality is not a trivial task and in most
cases results in excessive computation time, not always available in real-
world applications. Therefore, in most cases we search for suboptimal or
just feasible solutions. The simplest way to model an UAV path is by using
straight-line segments that connect a number of waypoints, either in 2D
or 3D space (Moitra et al., 2003, Zheng et al., 2005). This approach takes
into account the fact that in typical UAV missions the shortest paths tend to
resemble straight lines that connect waypoints with starting and target points
and the vertices of obstacle polygons. Although waypoints can be efficiently
used for navigating a flying vehicle, straight-line segments connecting the
corresponding waypoints cannot efficiently represent the real path that will
be followed by the vehicle due to the own kinematics of the traced path. As a
result, these simplified paths cannot be used for an accurate simulation of the
movement of the UAV in an optimization procedure, unless a large number
of waypoints is adopted. In that case the number of design variables in the
optimization procedure explodes, along with the computation time. This is
why this section presents some background based on the state-of-the-art
mission/path planning for UAVs. The purpose (apart from the state-of-the-
art survey) is to compare to the AVCL architecture in order to observe how
its TG2M embedded framework (see section 3 for details) is a complement
of some lacking approaches found in this specialized literature.

Background
Researchers at top research centers and universities (e.g. JPL at Caltech,
Carnegie Mellon, NASA, ESA, among others) are dealing with the
development of new strategies that allow high-level mission UAV planning
within the desired flight performance. As examples of these approaches
we present the NASA Ames Research Center, the Georgia Institute of
Technology, and the University of Illinois.
For the past decade NASA has focused on developing an efficient and
robust solution to Obstacle Field Navigation (OFN), allowing a fast planning
of smooth and flyable paths around both known and unknown obstacles
 Advanced UAV Trajectory Generation: Planning and Guidance 83

(Herwitz, 2007). Evolutionary algorithms have been used as a viable

candidate to solve path-planning problems effectively, providing feasible
solutions within a short time. Given a number of UAVs that are launched
from different and known initial locations, the algorithm must create 2-D
trajectories with a smooth velocity distribution along each trajectory and
with the goal of reaching a predetermined target location, while ensuring
collision avoidance and satisfying specific route and coordination constraints
and objectives. B-Splines curves are used in order to model both the 2-D
trajectories and the velocity distribution along each flight path. A flying
laboratory for autonomous systems, based on the Yamaha RMAX helicopter
is being developed, which it incorporates the OFN planner and one all-
digital camera system with a state-of-the-art tracking and passive ranging
capabilities. Machine stereo-vision is being used to determine safe landing
areas and monocular vision is used to track the landing location without
access to GPS. A Ground Control Station (GCS) is being integrated into the
simulation environment to investigate the issues related to high altitude and
long endurance flights.
As an alternative approach other researchers do not focus on generating
complex trajectories profiles; their aim is to develop robust mission
management, capable of successfully integrating the available onboard
hardware (e.g. cameras, sensors, etc). The Georgia Tech’s UAV Lab developed
the GTMax architecture (Alison et al., 2003) based on a Yamaha R-Max mini-
helicopter, is an example of this alternative approach. The GTMax system
is capable of fully autonomous flight with decentralized software modules
for trajectory generation, offline simulation, supervision, guidance and
control. The trajectory module generates lineal waypoints between targets
with initial and final velocity parameterization. Commands to the helicopter
take the form of different types of waypoints. All trajectories generated are
assumed to be physically feasible by the helicopter, the kinematic model
used for the trajectory generation uses specifiable limits on the maximum
speed and acceleration the aircraft may have during a maneuver. From a
high-level mission management perspective the GTMax architecture allows
the UAV to autonomously locate targets with an identifying symbol within
a designated search area. Once the correct target is identified the mission
coordination and flight path generation is done at a centralized location on
the ground, and the commands are transmitted to the UAV via a wireless
datalink. The GTMax GCS interfaces with the primary flight computer and
displays vehicle status, the object tracker information and the flight plans
generated by the mission planner. The Vision Monitoring Station receives
84 Unmanned Aerial Vehicles (UAV) and Drones

streaming video from the camera and the output of the image processor.
This allows the operator to monitor the efficiency of the image processing
as well as visually document the results of the search in the final phase of
the mission.
For other projects the main goal is the unification of high-level mission
commands and development of generic languages that support online UAV
mission planning without constraining the system to a single vehicle. The
University of Illinois at Urbana-Champaign (Frazzoli, 2002) presents a new
framework for UAV motion planning and coordination. This framework
is based on the definition of a modelling language for UAV trajectories,
based on the interconnection of a finite number of suitably defined motion
primitives, or maneuvers. Once the language is defined, algorithms for the
efficient computation of feasible and optimal motion plans are established,
allowing the UAV to fulfill the desired path.
If we analyze the UAS described above, almost all of them share one
important limitation: their software architecture is tightly coupled to one
vehicle and the capabilities of its lowlevel controller. Civil applications require
open and extendable software architectures capable of talking to vehicles
from different suppliers. The AVCL addresses those limitations, allowing to
model different vehicles into a single common language (vehicleindependent
missions). In the same fashion the described vehicles show that complex
and simple maneuvers could be a suitable solution depending on the kind
of mission to fulfill. For this reason the AVCL is extended with the TG2M
framework, capable of generating simple and complex 3D paths with the
necessary vehicle constrains. The next sub-section introduces the AVCL
architecture and some of the features provisioned.

The Aerial Vehicle Control Language - AVCL

The AVCL is not just a language capable of describing the missions and
capabilities of an heterogeneous group of vehicles, it is part of a bigger
framework that includes its interpreter, a definition of a base-vehicle, and a
Mission Planner that uses GIS as the datamodel for the world (Barrientos et
al., 2006). The Mission Planner (MP) is not tied to a particular set of vehicles,
sensors or commands. At any given time new functionality can be loaded and
displayed to the human operator as new options and commands. This means
that the MP tool is to be extended through Vehicle and Command Libraries
without recompiling, and those new capabilities and better vehicles can be
added easily. The Mission Planner is a great tool for simulation and direct
 Advanced UAV Trajectory Generation: Planning and Guidance 85

comparison of various trajectory trackers, UAV models and controllers,

because it can display N missions at the same time.
When considered just as a language the AVCL concept is the abstraction
layer that allows the human supervisor to create missions that are vehicle
and payload independent, promoting code reuse. At the same time the AVCL
statements and commands hide device specific software/hardware, and serve
as mission definition and storage. As an example of the code used to define
operations within a mission:

Compared to previous vehicle programming languages the AVCL and its

interpreter provide several advantages: intuitive handling of different systems
of units; its use of the objectoriented-programming paradigm; facilities for
inter-vehicle communications; run-time definition of relations between
vehicles, sensors and other equipment; it may be extended easily through
C, C++ or C# code; the interpreter is a light-weight application written in
C++, therefore it may be deployed in many SW/HW architectures. Before the
development of the TG2M module the AVCL framework relied on a simpler
guidance module to connect waypoints with straight-line segments, and while
the language could describe complex maneuvers and mission constraints the
framework lacked the capacity to fly a vehicle through complex paths.

Figure 1. The AVCL simplified diagram for mission planning

86 Unmanned Aerial Vehicles (UAV) and Drones

TRAJECTORY GENERATION AND GUIDANCE

MODULE - TG2M
The TG2M is designed to model 3D cartesian paths with parametric
constraints. The system uses two types of mathematical descriptors for
trajectories: polynomial interpolation and analytical functions. Complex
maneuvers are described with polynomial interpolation based on third and
fourth order splines with fixed boundary conditions (e.g. initial and final
velocities user definition), which join all control points with a continuous
and smooth path. Likewise, simple maneuvers like lines and circumference
are created with analytical functions that constrain the geometry of the
desired path. These constraints are typically kinematic: constant velocity
and acceleration, trapezoidal curves to accelerate or stop, etc.

Figure 2. The TG2M framework

As shown in Fig. 2, two main modules compose the TG2M framework:
the geometrical trajectory generation and the online closed-loop guidance.
All the parameters and fundamental data (e.g. the desired maximum speed,
etc) are provided by the user via the AVCL interpreter.

Polynomial Interpolation Using Splines

The fundamental idea behind the spline interpolation is based on the
definition of smooth curves through a number of points. These curves are
represented by a polynomial function (normally of third-grade) defined by
 Advanced UAV Trajectory Generation: Planning and Guidance 87

some coefficients that determine the spline used to interpolate the numerical
data points. These coefficients bend the line so that it passes through each
of the data points without any erratic behavior or breaks in continuity. The
essential idea is to define a third-degree polynomial function S (t) of the
form:

(1)
This third-degree polynomial needs to conform to the following conditions
in order to interpolate the desired knot-points as depicted in Fig. 3:

Figure 3. Knot-points to interpolate using 3D splines (with free boundary con-

ditions)
• S (t), S′ (t) and S′′ (t) will be continuous on the interval [t0, tn].
• Since the curve S(t) must be continuous across its entire interval,
it can be concluded that each sub-function must joint at the data
points, so: .
Taking into account the set of conditions previously described, for each

i = 1, 2.....n − 1, , and letting , Eq. (1) is

re-writing as:

(2)
Also, to make the curve smooth across the interval, the first derivative
of the S (t)function must be equal to the derivative of the position reference
points; this yields:
88 Unmanned Aerial Vehicles (UAV) and Drones

(3)
Applying the same approach for the second derivative:

(4)
For the solution of the polynomial coefficients in Eq. (1), the system in
Eq. (5) must be solved as:

(5)
where the matrix A ℜnxn (n is the number of knot-points to interpolate)
corresponds to:

(6)
The term h ℜn in Eq. (6) is the time vector defined for each point (P0
−1

,P1,....Pn ). The bi coefficients in Eq. (1) are stacked in Y ℜnx 3 , which yields
the term f ℜnx 3 as:

(7)
From Eq. (2) and (3), the ai and ci coefficients are respectively obtained as:

(8)
 Advanced UAV Trajectory Generation: Planning and Guidance 89

Example 1. Simple UAV trajectory planning using 3D splines with fixed

boundary condition. Let’s define three knot-points as: P0 = [0,0,0], P1 =
[5,10,10], P2 = [0,10, 20] with the following time condition for each point:
t = [0,10, 20] (note that the time vector components are given in seconds).
Calculate a 3D spline considering zero initial and final velocities.
The natural 3D splines with free boundary conditions may generate
smooth paths but without control over the velocities over the knot-points.
Because of this, Eq. (6) and (7) must be complemented in order to generate
a 3D spline with the fixed boundary conditions, in this case, with zero initial
and final velocities. Re-writing those equations yield:

(9)
The same system A⋅Y = f must be solved with the new A ∈ℜ from Eq. nxn

(9) and f ∈ℜnx3 defined in Eq. (10). Note that the first derivative of the S (t)
function has been added in the first and last position of the f vector, allowing
the control of the initial and final velocities of the curve.

(10)
The first procedure is to obtain the time vector as follows:

, then the system: Y = A −1 f must be solved to obtain

the polynomial coefficients as:
90 Unmanned Aerial Vehicles (UAV) and Drones

Finally, the two polynomials for each (x, y, z) component are evaluated
from the time [0, 10] to [10, 20]. Figure 4 shows the results.

Figure 4. 3D spline with fixed boundary condition

The end of Example 1: Simple UAV trajectory planning using 3D splines.
In the previous example, the 3D splines with fixed boundary conditions
allowed to define a smooth curve across the knot-points. Nevertheless, two
basic problems must be taken into account: the 3D spline only allows the
user to establish the initial and the final velocities of the whole trajectory,
limiting the user to have total control over the other points. The second
problem is the smoothness of the acceleration curves (linear). To solve these
problems 4D splines must be used. These address the user total control over
the velocity profile across the whole trajectory and its results in additional
smoothness for position, velocity and even acceleration curves. This could
be more effective for UAVs tasks where the mission requires a strong control
of the vehicle acceleration and velocities at each defined knot-point.

Figure 5. Knot-points to interpolate with 4D splines

 Advanced UAV Trajectory Generation: Planning and Guidance 91

The fourth-degree polynomial is defined by:

(11)
The same 3D-spline conditions previously described also apply to the
4D-spline. To obtain a generalized solution of the system to solve ( A⋅Y = f
), we start from the three-point case as depicted in Fig. 5. The polynomials
for each trajectory segment as a function of time t are:

(12)
Taking the first and second derivatives (velocities and accelerations),
we obtain:

(13)
The second derivatives of Eq. (12) yield a set of accelerations. Equaling
the acceleration functions for the intermediate points (t1 for each case) and
setting to zero the initial and final acceleration of the path segment yields:

(14)
Equations (12), (13) and (14) conform the complete system in order to
obtain the ten polynomial coefficients. Solving A⋅Y = f , the matrix A ∈ℜ5
(n−1) x 5 (n−1)
corresponds to:
92 Unmanned Aerial Vehicles (UAV) and Drones

(15)
The polynomial coefficients are stacked into the Y ∈ℜ5(n−1) vector, and
with the f ∈ℜ5 (n−1) term, the total system is defined from Eq. (14) and (15):

(16)
Example 2. UAV trajectory planning using 4D splines with total velocity
control Lets define three knot-points as: P0 = [0,0,0], P1 = [5,7,5], P2 = [4,
2,10] in the time t = [0,4,8] (note that the time vector components are given
in seconds). Calculate the 4D spline considering the following velocity
profile: V0 = [0,0,0], V1 = [0.5,0.8,1], V2 = [1,1.5,1].
Solving the system: Y = A −1 f , we obtain the following numerical data:

Finally, the two polynomials for each (x, y, z) component are evaluated
from the time [0, 4] to [4, 8] seconds. Figure 6 shows the results:
 Advanced UAV Trajectory Generation: Planning and Guidance 93

Figure 6. 4D spline with total control of the knot-points velocities

The end of Example 2: UAV trajectory planning using 4D splines with
total velocity control

Simple Maneuvers with Analytical Functions

For simple maneuvers the TG2M framework also supports the definition
of straight-lines and circumferences via analytical functions that constrain
the geometry of the desired path. These constraints are typically kinematic:
constant velocity and acceleration, trapezoidal curves to accelerate or stop,
etc. This kind of parameterization is useful when the mission requires
the UAV stops at the desired end-point of the trajectory effectively, due
to the usercontrol of the acceleration slope tilt level. For both (lines and
circumferences) the desired set of velocities must fulfill the following
trapezoidal velocity profile:

Figure 7. Trapezoidal velocity profile used for simple UAV maneuvers

From Fig. 7, three fundamental segments compose the total function to
define the straightline. The two intermediate points of the trapezoidal curve:
P(τ) , P (T − τ ) are calculated as:
94 Unmanned Aerial Vehicles (UAV) and Drones

(17)
Once that the intermediate points have been defined, the three segments
that compose the total function are defined by Eq. (18). In the first section
(from P0 to P(τ )) the initial velocity is set Vi = 0 and progresses toward a
final velocity Vk . The second segment (from P(τ ) to P (T − τ )) is traced at
constant maximum velocity Vk . Finally the last segment (from P (T − τ )to
P T( )) drives the vehicle from Vk to zero velocity.

(18)
The same approach is applied to generate circumferences with the
trapezoidal profile used for the straight-lines. Three knot-points define the
circumference path and the objective is to find the trace angle across the
trajectory.

Figure 8. X-Y plane for circumference maneuver

The first step is to determine the center of the circle that joins the three knot-
points. The equation of the trajectory plane is defined by the cross product
between the vectors formed by points P0 , P1 and P2 , as shown in Eq. (19).
 Advanced UAV Trajectory Generation: Planning and Guidance 95

(19)
The center of the circle c (xc , yc ,z c ) is always equidistant to any point
over the circle, then:

(20)
To obtain the relation between the angle motion (θ ), the angular velocity
(ω ) and the tangential velocity ( vt ), constant velocity is assumed across the
path, yielding:

(21)
Likewise, the relation between the angle motion (θ ) and the angular
acceleration of the curve is given by:

(22)
If the known parameter is the total motion time (T) of the trajectory, we
set the acceleration term to the left-side of the equation, yielding:
96 Unmanned Aerial Vehicles (UAV) and Drones

(23)
Otherwise, if the constrained parameters are the initial and final velocity,
the acceleration function is given by:

(24)
Example 3. UAV straight-line trajectory with trapezoidal velocity profile
Lets defined a straight-line trajectory starting from P0 = [0,0, 20] to P (T )=
[10,0, 20] with 10 meters of displacement in the X coordinate at a constant
altitude of 20 meters. Define the 3D cartesian trajectory with a maximum
velocity of 1m/s taking into account a trapezoidal velocity profile with 30%
of acceleration time.
The distance vector to trace is: dist = P (T)− P0 = [10,0,0].
The total velocity is obtained from the norm of the vector dist: VT = dist
= 10.

The total motion time is calculated as: , where Vk is

the maximum velocity at 1m/s and pt is the percentage of acceleration

(30%=0.3). The acceleration motion is also obtained as: .

The number of the total points to generate is: n = 28 (the user defines this
parameter depending on the UAV controller data rate). In the same way, the
number of points in the first and last segment of the trapezoidal profile is
calculated as: k = 2(pt ⋅n) = 16, and for the constant velocity segment: kc =
n − 2k = 12 . Once that the preliminary data have been calculated, Eq. (16)
is used to obtain the intermediate points (see Fig. 7):
 Advanced UAV Trajectory Generation: Planning and Guidance 97

The first segment of the path (from P0 to P(τ )) is calculated using Eq.
(17):

The time vector t is obtained as: , which

is: t =
For the second segment (at constant velocity Vk ), the time vector is

calculated as: [0 0.0833 0.1667 0.25 0.3333 0.4167 0.5

0.5833 0.6667 0.75 0.83 0.9167]

Finally the last segment drives the vehicle from Vk to zero end velocity.
The time vector is:

yielding the following function for each coordinate of motion:

98 Unmanned Aerial Vehicles (UAV) and Drones

The results are presented in Fig. 9:

Figure 9. UAV straight-line motion with trapezoidal velocity profile

The end of Example 3: UAV straight-line trajectory with trapezoidal
velocity profile

UAV Guidance
The geometrical representation of a UAV trajectory has been presented
in the previous sub-section. Complex trajectories may be described using
third-fourth order degree splines, or simple maneuvers may be generated
using common lines and circumferences functions with some parametric
features defined by the end-user.
But in order to complete the generation of trajectories for a single UAV
a guidance module is required.
 Advanced UAV Trajectory Generation: Planning and Guidance 99

Figure 10. TG2M guidance scheme

Figure 11. Velocity error command

The AVCL simulation module contains a dynamic model of a mini-
helicopter (Valero, 2005) and its associated controllers: attitude, velocity and
position. The velocity controller is based on a Proportional-Integrative “PI”
design; it receives the velocity error command generated by the guidance
module shown in Fig. 10. These velocity error references are named Vxref
and Vyref, with respect to either the world or vehicle’s frame, and are
calculated with the UAV’s error position. In Fig. 11 the theoretical (or ideal)
UAV velocity vector is represented by the Vt term. Due to wind and other
perturbations during flight, a velocity error vector Ve must be considered in
order to set the final velocity references to send to the vehicle’s controller.
100 Unmanned Aerial Vehicles (UAV) and Drones

This vector is derived from the UAV position error between the current and
desired positions. Finally, the velocity references Vxref and Vyref are the
components of the vector Vt +Ve.
The guidance module must take into account that the helicopter’s orientation
(yaw) is independent from the direction of travel. The high-level modules
must generate a set of orientations across the vehicle’s trajectory. The built-
in AVCL control module (see Fig. 10) is capable of receiving velocity and
yaw angle orientation commands from the TG2M module and generating the
needed commands for the attitude controller that governs the vehicle’s roll
and pitch. The TG2M’s guidance module focuses on the generation of yaw
references, and use a simple Proportional “P” controller for smooth yaw
angle transition during the flight. Two methods are use to generate the yaw
angle references: for simple maneuvers the yaw angle is calculated using
simple trigonometric relations due to the path displacement. For complex
trajectories using splines we introduce a feasible method to calculate a
smooth yaw angle evolution. This method also shows how to calculated roll
and pitch references due to the vehicle trajectory and its velocity. For roll,
pitch and yaw angles calculation, the following frame of reference is used:

Figure 12. Frames of reference for UAV guidance and control

From Fig. 12, the following vectors can describe the motion of the UAV
vehicle in 6 DOF:

(25)
In Eq. (25), η1 denotes the position of the center of mass CM of the
vehicle and η2 its orientation described by the Euler angles with respect to
the inertial frame {I}. The vector v1 refers to the linear velocity and v2 to the
 Advanced UAV Trajectory Generation: Planning and Guidance 101

angular velocity of vehicle frame {B} with respect to inertial frame {I}. In
order to express the velocity quantities between both frames of references
(from {B} to {I} and vice versa), the following transformation matrix is
used:

(26)
The body-fixed angular velocities and the rate of the Euler angles are
related through:

(27)
The position and the magnitude of the velocity vector at a point P on the
trajectory are given by :

(28)
The method to define the Euler angles is based on the Frenet-Serret theory
(Angeles, 1997). To every point of the curve we can associate an orthonormal
triad of vectors (a set of unit vectors that are mutually orthogonal) namely
the tangent et , the normal en and the bionormal eb (see Fig. 13). The Frenet-
Serret theory says that by properly arranging these vectors in a matrix ∈ℜ3x 3
, we obtain a description of the curve orientation due to the position, velocity
and acceleration of the UAV while tracing out the path. The unit vectors are
then defined as:

(29)
In the definition of a frame associated with the curve the original
definition of the Frenet frame for counterclockwise rotating curves is used;
in the case of a clockwise rotating curve, the z-axis of the Frenet frame
points in the opposite direction upwards than the inertial {I} frame. So in
order to define small relative rotation angles for the orientation of a vehicle
102 Unmanned Aerial Vehicles (UAV) and Drones

rotating clockwise and having its zb axis pointing downwards, we define a

reference frame associated with the curve as previously, but rotated with
respect to the Frenet by an angle of 180 degrees about the x−axis of the
Frenet frame (see Fig. 13).

Figure 13. The inertial, the Frenet, the vehicle and the curve Frames
Collectively we denote the Frenet and the rotated frame as the “curve”
frame {C}. According to the notation of rotational transformations used in
robotics literature, we can express the coordinates of a vector given in the
curve frame {C} to the {I} frame with the matrix:

(30)
For a counterclockwise rotation: . Likewise,
the rotation of the {B} frame from the {C} frame to the reference {R} frame
can be expressed using customary aeronautical notation by considering the
sideslip angle β and angle of attack α, (Siouris, 2004):

(31)
The vector vR refers to the y-axis velocity component in the reference
frame and wR,uR to the z and x - axis respectively. The overall rotation is
composed by a rotation about body zB axis through the angle β , followed
by a rotation about the body yB through the angle α , which is expressed as:
 Advanced UAV Trajectory Generation: Planning and Guidance 103

(32)
Finally, the roll, pitch and yaw angles can be deduced as follows:

(33)
Where ri, j represent the components of the rotation matrix RIR ∈ℜ3x 3
. Computing the previous methodology, we use 3D splines with fixed
boundary conditions in order to generate a complicated path as shown in
Fig. 11. Seven knot-points have been used (distance are in meters): P=[0 0 0;
5 1 2; 10 5 5; 15 10 10; 10 15 15; 5 10 20; 0 8 20; 5 0 20]. Eq. (33) has been
used to obtain the UAV orientation with respect to the Inertial Frame as:

Figure 14. Roll, Pitch and Yaw angle references for UAV guidance
For complex maneuvers the Frenet theory allowed to define smooth yaw
references as well as roll and pitch angles if it is required. Nevertheless, the
computational cost of calculating those equations could decrease the system
performance if the number of knot-points of the path is large. For this reason,
normal maneuvers such as straight-lines use simple trigonometric theory
to obtain the UAV orientation. On the other hand, the end-user is able to
define the kind of orientation of the vehicle, this means that the vehicle is not
constrained to be oriented just by the trajectory direction, hence, it will be
104 Unmanned Aerial Vehicles (UAV) and Drones

able to trace out the trajectory oriented towards to any fixed point defined.
The yaw angle defined by the term ψ is given by:

(34)
Where xdiff , ydiff correspond to the difference between the target fixed point
and the current position of the UAV. In addition, depending of the motion
quadrant, the term ψ in Eq. (34) must be fixed, this means that for the {x-
,y+} and the {x- ,y-} quadrant, the yaw angle is ψ = ψ +π , otherwise, for the
{x+,y-} quadrant, ψ = ψ + 2π .
Figure 15 shows a circumference-arc generated using Eq. (24) as well as
the yaw evolution of the UAV oriented towards the center of the arc located
at [0,0] coordinate in the x-y plane using the previosly theory described in
Eq. (34).
This section has successfully introduced the mathematical treatment and
methods for the generation of complex trajectories and simple maneuvers
using the available theory reported in specialized literature (LaValle S.M,
2006), (Jaramillo-Botero et al., 2004). Geometrical trajectory generation
and some techniques for its parameterization based on polynomial splines
and function with trapezoidal velocity profile are an interest solution for this
problem, actually, some of these methodologies are used for complex UAV
trajectory definition nowadays. The novel solution presented in this book
is the integration of these methods into a powerful environment that allows
high-level user control of complex missions. The AVCL definitively brings
those features and the next section will introduce some tests using the AVCL
interpreter and the simulation environment.

Figure 15. Circumference 3D motion, trapezoidal velocity profile and yaw

angle evolution
 Advanced UAV Trajectory Generation: Planning and Guidance 105

TG2M SIMULATION
Results As shown in Fig. 1 the Mission Planner (MP) has two similar loops
for mission planning and simulation/supervision. The difference is that in
the Planning Loop the interpreter sends the projected waypoints back to
the MP’s Enhanced Reality, while in the Simulation Loop the interpreter
commands the simulated vehicle, which in turn sends the simulated positions
to the MP. Our research group has developed a Simulink-based model of
a UAV helicopter named Vampira, which includes a position controller
and is capable of real-time simulation. This simulator has been used with
the Mission Planning and Simulation tool to test the TG2M. For Mission
Supervision the AVCL commands would be sent to the real vehicle, and its
position plotted alongside the projected and/or simulated paths. The Vampira
helicopter was built within the framework of the project: “Guidance and
Control of an Unmanned Aerial Vehicle” DPI 2003 01767 (VAMPIRA),
granted by the Spain Ministry of Science and Technology, and it will be
used for the real-world tests of the built-in TG2M framework. Figure 16
shows the Vampira prototype, which includes: a GPS, Wi-Fi link, IMU, and
a PC104 computer for the low-level control (main rotor and tail servos).
The Vampira’s dynamics model has been obtained, identified and validated
in previous works (Valero, 2005), (del Cerro et al., 2004). This work takes
advantage from the AVCL simulation capabilities in order to validate the
TG2M framework theory for trajectory planning.

Figure 16. The Vampira’s Helicopter prototype

Three test scenarios showcase the TG2M validation process. These tests
involve the whole methodology previously presented in the other sections
of this chapter, as well as the numerical simulation results using the AVCL
environment and the embedded dynamics and control algorithms for the
Vampira’s helicopter. Two complex maneuvers are presented using 3D and
106 Unmanned Aerial Vehicles (UAV) and Drones

4D splines respectively and a simple last test using analytical function to

generate a parameterized circumference motion.
1). Semi-spiral using 3D splines for the velocity profile generation
and the Frenet theory for UAV orientation: In this first test, we
used a 3D spline to joint three knot control points: (P0(0, 0, 0),
P1(3, 5, 10), P2(6, -7, 20)) at the desire time (given in seconds)
for each point: (t(0, 10, 20)) and the desire initial and final speed
(given in m/s): (V0(0, 0, 0), Vf(0, -0.2, 0.4)):

Figure 17. The AVCL simulation environment: Vampira’s helicopter executing

a semi-spiral motion using 3D splines

Figure 18. Test1: Cartesian UAV position and velocities with respect to the
Inertial Frame
The UAV started from initial point located at (0, 0, 0) coordinate and
finished its trajectory at (6, -7, 20). Visual simulation depicted in Fig 17,
showed smooth motion across the trajectory due to the 3D spline approach.
Nonetheless, 3D splines just allow the user to define the initial and final
 Advanced UAV Trajectory Generation: Planning and Guidance 107

velocities of the motion, lacking of velocity control for the rest of the
knotpoints.

Figure 19. Test1: UAV orientation (Euler angles evolution)

To solve this problem, the following test introduces a more complex
trajectory generation using 4D splines, addressing total user control of
the UAV velocity profile. 2). Complex trajectory using 4D splines for the
velocity profile generation and the Frenet theory for UAV orientation.
This trajectory includes different kind of maneuvers joined into a single
polynomial function (take-off, circumference-type motion and slow down in
spiral-type motion). This test includes UAV long-endurance to high altitude
(150 meters above ground) and a maximum easting displacement about of
60 meters. The following knot-control points (given in meters) have been
defined: (P0(0, 0, 0), P1(0, 0, 20), P2(0, 0, 40), P3(0, 0, 60), P4(10, 2, 80) ,
P5(20, 4, 110) , P6(25, -7, 130) , P7(30, -10, 150), P8(35, -5, 140) , P9(30,
16, 125) , P10(20, 5, 130) , P11(33, -10, 145) , P12(40, -5, 135) , P13(55,
-6, 125)):

Figure 20. Test2: smooth 4D spline for complex maneuve

108 Unmanned Aerial Vehicles (UAV) and Drones

Figure 21. Test3: Arc-type motion using analytical functions

The advantage about using 4D splines relies in the possibility of defining
feasible paths that matches with the knot-control points defined (with less
match error percentage than the 3D polynomial splines). In addition, the
user is able to define the set of velocities for each of the knot-points during
the motion. The set of velocities (given in m/s) are: (V0(0, 0, 0), V1(0, 0,
0.8), V2(0, 0, 1), V3(0, 0, 1.2), V4(0.5, 1, 1.4) , V5(0, 0.5, 1.7) , V6(2, 1.5,
2.5), V7(3, 2.2, 3), V8(2, 1.2, 2) , V9(1, 0.5, 1.5) , V10(-0.5, -1, 0.8) , V11(-
3, -2, 0.4) , V12(-1, -0.5, 0.8), V13(0, 0, 0)).
3). Simple arc-type maneuver with trapezoidal velocity profile
parameterization: for the analytical AVCL feature of trajectory planning,
the TG2M module supports straight-lines and circumferences motions. An
arc defined by: P0(0, 0, 2), P1(10, 5.5, 2), P2(20, 0, 2)) with a maximum
velocity of 0.5m/s is tested using the AVCL interpreter that allows the user
to define the trapezoidal velocity profile configuration. For this case, the
acceleration slopes of the curves (see Fig. 21) have been set to the 30% of
the total motion.

Figure 22. Test3: UAV orientation (Euler angles evolution)

 Advanced UAV Trajectory Generation: Planning and Guidance 109

FINAL OBSERVATIONS
For modeling continuous cartesian trajectories in the AVCL, several analytical
functions and polynomial interpolation methods are available; all of which
can be used in any combination. The TG2M module handles the definition
of trajectories using knot control points as well as the incorporation of path
constraints. It also supports the definition of complex tasks that require
the construction of trajectories made up of different primitive paths in any
combination of analytical and interpolated functions. The user-designed
spatial trajectories can be visualized in three dimensions on the display
window or plotted versus time using the embedded plotting library.
Simulation results have shown that the TG2M module works perfectly
for the definition and testing of wide kind of smooth trajectories, allowing
the user a high-level control of the mission due to the AVCL interpreter.
The three different scenarios used for testing, allowed verifying that the
mathematical framework used for the trajectory generation and guidance
was really working during simulation flight. Percentage errors during
maneuver execution were minimal, maintaining the UAV at the desired
velocity limits and within the established path. We also incorporated velocity
error fixing during flight. For high altitude tests, the velocity of the wind
plays a mandatory role as a main disturbance external force. The TG2M
module includes wind perturbation compensation. The Guidance module
fixes the velocity commands in real-time flight maneuver, decreasing the
error position tracking. For the three scenarios tests, the AVCL simulation
environment includes normal wind conditions during simulation flight,
introducing small perturbations into the UAV equations of motion. As shown
in the obtained results, those perturbations were compensated, allowing the
UAV to follow the desired trajectory within the less error as possible.
The Frenet-Serret formulas included for the UAV orientation also
presented a good approach in order to obtain smooth UAV rotation rate
during flight. The use of simple trigonometric theory to obtain and define
the UAV orientation profile (Yaw angle) is not convenient for complex
maneuvers. Splines sometimes require a lot of know-points for feasible
trajectory guidance, hence, using these polynomial equations, the Frenet
approach allowed smooth angle changes between knot-points, which it had
not been obtained with the simple trigonometric angle calculation.
110 Unmanned Aerial Vehicles (UAV) and Drones

REFERENCES
1. JAA & Eurocontrol,.A concept for the European Regulations for Civil
Unmanned Aerial Vehicles. UAV Task-Force Final Report. 2004
2. Coifman, B., McCord, M., Mishalani, M., Redmill, K., Surface
Transportation Surveillance from Unmanned Aerial Vehicles. Proc. of
the 83rd Annual Meeting of the Transportation Research Board, 2004.
3. Held, Jason M; Brown, Shaun and Sukkarieh, Salah. Systems
for Planetary Exploration. I5th NSSA Australian Space Science
Conference, pp. 212-222, ISBN: 0864593740. RMIT University,
Melbourne, Australia, from 14 to 16 September 2005.
4. Gutiérrez, P., Barrientos, A., del Cerro, J., San Martin., R. Mission
Planning and Simulation of Unmanned Aerial Vehicles with a GIS-based
Framework; AIAA Guidance, Navigation and Control Conference and
Exhibit. Denver, EEUU, 2006.
5. Rysdyk, R. UAV path following for constant line-of-sight. In 2th AIAA
Unmanned Unlimited. Conf. and Workshop and Exhibit, San Diego,
CA, 2003.
6. Price, I.C, Evolving self organizing behavior for homogeneous and
heterogeneous swarms of UAVs and UCAVS. PhD Thesis, Graduate
School of Engineering and Management, Air Force Institute of
Technology (AU), Wright-Patterson AFB, OH, March 2006.
7. Herwitz, S. Developing Military COEs UAV applications. UAV
Applications Center – NASA Ames Research Center, Moffett Field,
CA. 2007.
8. Alison A. P., Bryan G., Suresh K. K., Adrian A. K., Henrik B. C.
and Eric N. J. Ongoing Development of an Autonomous Aerial
Reconnaissance System at Georgia Tech . UAV Laboratory, School of
Aerospace Engineering. Georgia Institute of Technology, 2003
9. Jaramillo-Botero A., Correa J.F., and Osorio I.J., Trajectory planning in
ROBOMOSP, Robotics and Automation Group GAR, Univ. Javeriana,
Cali, Colombia, Tech. Rep. GAR-TR-10-2004, 2004.
10. LaValle, S.M.: Planning Algorithms. Cambridge University Press
(2006)
11. Moitra, A., Mattheyses, R.M., Hoebel, L.J., Szczerba, R.J., Yamrom,
B.: Multivehicle 5reconnaissance route and sensor planning. IEEE
Transactions on Aerospace and Electronic Systems, 37 (2003).
12. Zheng, C., Li, L., Xu, F., Sun, F., Ding, M.: Evolutionary Route Planner
 Advanced UAV Trajectory Generation: Planning and Guidance 111

for Unmanned Air Vehicles. IEEE Transactions on Robotics 21 (2005)

609–620
13. Frazzoli, E. Maneuver-based motion planning and coordination
for single and multiple UAVs. AIAA’s 1st technical conference and
workshop on unmanned aerospace vehicles. University of Illinois at
Urbana-Champaign, il 61801. S. Portsmouth, Virginia. May 2002.
14. Angeles J., Fundamentals of Robotic Mechanical Systems. Theory,
Methods, and Algorithms. Springer, New York, 1997.
15. Siouris, G. M. Missile Guidance and Control Systems. Springer, New
York, 2004.
16. Valero, Julio. Modelo dinámico y sistema de supervisión de vuelo de
un helicóptero autónomo. Proyecto de fin de carrera. ETSIIM-UPM.
2005.
 CHAPTER

6
Modelling and Control Prototyping of
Unmanned Helicopters

Jaime del-Cerro, Antonio Barrientos and Alexander Martínez

Universidad Politécnica de Madrid – Robotics and Cybernetics Group, Spain

INTRODUCTION
The idea of using UAV’s (Unmanned Aerial Vehicles) in civilian applications
has created new opportunities for companies dedicated to inspections,
surveillance or aerial photography amongst others. Nevertheless, the main
drawback for using this kind of vehicles in civilian applications is the
enormous cost, lack of safety and emerging legislation. The reduction in
the cost of sensors such as Global Positioning System receivers (GPS) or

Citation: Jaime del-Cerro, Antonio Barrientos and Alexander Martinez (January 1st
2009). “Modelling and Control Prototyping of Unmanned Helicopters”, Aerial Vehi-
cles, Thanh Mung Lam, IntechOpen, DOI: 10.5772/6468.
Copyright: © 2009 The Author(s). Licensee IntechOpen. This chapter is distributed
under the terms of the Creative Commons Attribution-NonCommercial-ShareAlike-3.0
License
114 Unmanned Aerial Vehicles (UAV) and Drones

nonstrategic Inertial Measurement Units (IMU), the low cost of computer

systems and the existence of inexpensive radio controlled helicopters have
contributed to creating a market of small aerial vehicles within an acceptable
range for a wide range of applications.
On the other hand, the lack of safety is mainly caused by two main points:
Mechanical and control robustness. The first one is due to the platform being
used in building the UAV in order to reduce the cost of the system, which is
usually a radio controlled helicopter that requires meticulous maintenance
by experts.
The second is due to the complexity of the helicopter dynamics since it
is affected by variations in flying altitude, weather conditions and changes
in vehicle’s configuration (for example: weight, payload or fuel quantity).
These scenarios disrupt the modeling process and, consequently, affect
the systematic development of control systems, resulting to tedious and
critical heuristic adjustment procedures.
Researchers around the World propose several modeling techniques or
strategies for dynamic modeling of helicopters.
Some works on helicopter modeling such as (Godbole et al or Mahony
et al,2000), (Gavrilets et al, 2001), (Metler et al and La Civita et al, 2002),
and (Castillo et al, 2007) show broad approaches that have been done in this
field of engineering. The lack of an identification procedure in some cases
and the reduced field of application in others, make sometimes difficult to
use them.
In this chapter, not only a modeling is described, but also the identification
procedure that has been successfully tested.
The proposed model has been defined by using a hybrid (analytical
and heuristic) algorithm based on the knowledge of flight dynamic and by
resolving some critical aspects by means of heuristic equations that allow
real time simulations to be performed without convergence problems.
Evolutionary algorithms have been used for identification of parameters.
The proposed model has been validated in the different phases of the
aircraft flight: hovering, lateral, longitudinal or vertical using a Benzin
Trainer by Vario which relies on a 1.5 m of main rotor diameter and a
payload of about five kilograms.
 Modelling and Control Prototyping of Unmanned Helicopters 115

Figure 1. Benzin trainer by Vario with GNC (Guidance Navigation & Control)
system onboard
A full structure of control has been developed and tested by using a
proposed fast design method based on Matlab-Simulink for simulation and
adjustment in order to demonstrate the usefulness of the developed tool. The
proposed architecture describes low level control (servo actuators level),
attitude control, position, velocity and maneuvers. Thus there are commands
such as straight flying by maintaining a constant speed or maneuvers such
as flying in circles i.e.
The results have confirmed that hybrid model with evolutionary
algorithms for identification provides outstanding results upon modeling a
helicopter. Real time simulation allows using fast prototyping method by
obtaining direct code for onboard controllers and consequently, reducing the
risk of bugs in its programming.

HELICOPTERS FLIGHT PRINCIPLES

The required thrust for flying helicopters (elevation and advance) is only
provided by the main rotor. In order to obtain a comprehensive knowledge
about the forces involved in the main rotor, an exhaustive explanation would
be required. The forces involved in the main rotor generate twisting and
bending in the blades, as well as elevation forces and different kinds of
aerodynamic resistance.

Figure 2. Basic Helicopter Thrust analysis

116 Unmanned Aerial Vehicles (UAV) and Drones

In a first approach, only a sustentation and a resistance force isolated from

any other influence could be taken into account. The thrust (T) generated by
the air pressure against the blade has an inclination in relation to the rotation
plane. This force can be divided in vertical force (FA) and resistance (FR)
that is applied in the horizontal plane against rotation direction (Figure 2).
Helicopters rely on mechanisms to modify the attack angle of the blade
in the main rotor. It allows controlling the movement of the fuselage through
the inclination of the rotation plane. The fact is that the attack angle of the
blade is not constant neither in time nor space. It continuously changes while
the blade is rotating as azimuth angle indicates. It can be assumed that the
attack angle is the addition of two components: The first is an average attack
angle during one complete rotation of the blade, called collective angle. The
second component depends on the azimuth angle. When the azimuth angle
is 0º or 180º, the blade has the roll cyclic angle. When 90 º or 270 º, the blade
has the pitch cyclic angle (Figure 3).

Figure 3. Attack angle during a blade revolution

Using these three signals (collective, roll and pitch cyclic), a pilot is able
to control the main rotor. In addition to these signals, the pilot also controls
the attack angle of tail rotor blades and the engine throttle.
Typically, radio-controlled helicopters rely upon a commercial control
system to maintain the speed of the main rotor constant. The vertical control
of the helicopter is done by changing the collective attack angle in the main
rotor.
On other hand, the mission of the tail rotor is to compensate the torque
that main rotor creates on the helicopter fuselage. The compensation torque
can be adjusted by changing the attack angle of its blades. The tail rotor
typically requires values between 5 to 30 per cent of the total power of the
helicopter.
 Modelling and Control Prototyping of Unmanned Helicopters 117

A lot of physical principles such as ground effect, downwash, flapping or

atmospheric effects have to be considered in an in-depth study of helicopter
flight dynamics. Taking into account the application scope for the proposed
model, which is a small helicopter with small capabilities, no change of air
density is considered. Moreover, the blades of the proposed model are also
considered as solid bodies.

MODEL DESCRIPTION
Mathematical models of helicopter dynamics can be either based on
analytic or empirical equations. The former applies to the physical laws
of aerodynamics, and the latter tries to define the observed behavior with
simpler mathematical expressions.
The proposed model is a hybrid analytic-empirical model that harnesses
the advantages of both: high-velocity and simplicity of the empirical method,
as well as fidelity and the physical meaning of the analytic equations.

Inputs and Outputs

The proposed model tries to replicate the behavior of a small radio controlled
helicopter, therefore the inputs and outputs have been selected as the controls
that the pilot relies when using a commercial emitter. Cyclic (Roll and Pitch)
and collective controls have been considered as inputs.

Table 1. Model inputs and outputs

Radio controlled helicopters usually rely on electronic systems based

on gyroscopes to tail stabilization. Based on this fact, the yaw angle is
controlled by giving rotational speed commands. Thus, the yaw rate has
been considered as one of the inputs to the model.
It is also common that helicopters rely on a main rotor speed hardware
controller. In such manner, the rotor maintains the speed and consequently,
the vertical control is then performed by the changing of the collective angle.
118 Unmanned Aerial Vehicles (UAV) and Drones

The assumption in considering constant speed reduces the complexity of

the model and maintains its quality. Furthermore, the use of this hardware
controller decreases the number of inputs since no throttle command is
required.
Accelerations and rotational speeds are the outputs of the model. Table 1
summarizes the inputs and outputs of the model.

Model Block Diagram

A block diagram of the proposed model is described in Figure 5. A brief
definition of every part is also shown in the following sections.

Main Rotor
It is modeled with analytic equations, derived form research on full-scale
helicopters [Heffley 2000], which can be used to model small helicopters
without fly-bars. The iterative algorithm computes thrust and induced wind
speed while taking into account geometrical parameters, speed of the blades,
commanded collective, roll and pitch cyclic.
One of the most important concepts to be known is the relationship
among the Torque, Power and Speed. Such response arises from the induced
velocity of the air passing through the disk of the rotor.

Figure 4. Outputs of the model

 Modelling and Control Prototyping of Unmanned Helicopters 119

The airflow goes downward, and due to the action-reaction principle, it

generates a vertical force that holds the helicopter in the air. The engine
provides the torque to make the blades rotate and create the airflow.
Although there are many factors that make difficult to exactly
determination of the relative speed between the helicopter and the airflow
through the disk that the rotor creates when rotating, it is possible to work
with a first order approach. In this way, it is possible to model using the
momentum classic theory and estimating the force and induced velocity
using an feedback aerodynamic block. Nevertheless, calculus turns to be
difficult because the feedback is highly non-linear.
Another aspect regarding the induced speed is its influence on the
surrounding surfaces, thus it can be affected by the ailerons and others
fuselage parts and it changes depending on the speed and direction of the
flight. In this model, torque and induced speed have been modeled assuming
a uniform distribution of the air passing through the rotor disk.

Figure 5. Model Block Diagram

The computation of the torque and induced speed is based on the classic
momentum theory but using a recursive scheme that allows to reach a fast
convergence. The equations used to model the main rotor have two groups
of inputs, as Figure 6 shown.
The collective step (θCol) and the blade torsion (θTwist), compose the
first branch. The rotor axis inclination (Is) and the cyclic roll and pitch the
second one.
The attack angles a1 and b1 are derived from the cyclic roll and pitch
references. The wind speed relative to the helicopter is present through its
three components: ua, va, and wa.
120 Unmanned Aerial Vehicles (UAV) and Drones

Figure 6. Main rotor block diagram

The equations corresponding to this part are:

(1)

(2)
The output of the block is the rotor’s thrust (F). R and Ω represent the
radius of the rotor and its angular speed respectively; ρ is the air density, and
a, b and c are geometrical blade factors. The relationship between thrust and
angular rates has been derived from observation; therefore the model is also
empirical.

(3)

(4)

(5)
When the helicopter flies close to the ground (distances less than 1.25
times the diameter of the rotor) the ground effect turns to be very important.
This effect has been modeled using the parameter η defined in (6) where h is
the distance from the helicopter to the ground.

(6)
 Modelling and Control Prototyping of Unmanned Helicopters 121

In these cases, the thrust is modified using (7) where Th’ is the resulting
thrust after correcting Th. Values of To, T1 and T2 have been calculated for
no creating a discontinuity when h is 1.25 times the diameter of the rotor.

(7)
By other hand, the commanded Roll and Pitch cyclic have been considered
as reference for rotations in x and y axes considering the helicopter frame.
In addition to this, a coupling effect has been considered for simulating real
coupling in the helicopter.

Figure 7. Roll and Pitch dynamic

As it was mentioned in the last section, RC helicopters usually rely upon
commercial speed controllers in the main rotor. These devices have been
modeled using an ideal dynamic response. On the other hand, engine has
been modeled as a first order system. Therefore, variations in the speed of
the rotor have been considered only due to changes in the collective angle
as Figure 8 shows.

Figure 8. Engine model

122 Unmanned Aerial Vehicles (UAV) and Drones

Tail Rotor
The algorithm for estimating the thrust provided by tail rotor is similar to
the main one but only the pitch angle of the blades has been considerated
as input. This signal is provided by the hardware controller that is in charge
of stabilization of the tail. A PI classical controller has been used to model
the controller and the sensor has been considered as no dynamic as Figure
9 shows.

Figure 9. Tail rotor model

Fuselage
In a first step, all the forces (main and tail rotor, gravity, friction and
aerodynamic resistances) have to be taking into account for computing the
movement of the fuselage of the helicopter. After that, accelerations and
velocities can be estimated.
Forces due to the aerodynamic frictions are estimated using the relative
velocity of the helicopter with respect to the wind applying (8), where Ar is
the equivalent area and vw is the wind velocity.

(8)
The resulting equations are summarized in (9).

(9)
Where the prefix Fus means aerodynamic forces on fuselage and T the
main rotor thrust. The prefix M means Torque where suffix d denotes main
 Modelling and Control Prototyping of Unmanned Helicopters 123

rotor, the t denotes tail rotor and the v is the effect of the air in the tail of the
helicopter. The prefix G means the gravity components.
Once forces and torques have been calculated, accelerations can be
obtained and therefore velocities. Once the velocity referred to helicopter
frame (vh) is calculated, a transformation is required in order to obtain an
absolute value by using (10).

(10)
Flybars are important elements to take into account in the dynamic
model. In this work, they have been modeled by using empiric simple
models, and the stabilization effect that they produced on the helicopter has
been simulated using (11). A more realistic model of flybars can be obtained
in [Mettler-2003].

(11)

Identification of the parameters

Once the mathematic equations have been obtained, a procedure to give
values to the parameters that appear into the model is required. Some of the
parameters can be easily obtained by simple measurements or weighs but
some of them turn to be very difficult to obtain or estimate.
Table 2 describes the parameter list to be identified.
Some methods have been studied for performing the parameters
identification, such as multi-variable systems procedures (VARMAX), but
they are difficult to apply because there is an iterative process into the model.
Due to this, the selected method was evolutionary algorithms, after trying
unsuccessfully stochastic methods.
Genetic algorithms may be considered as the search for a sub-optimal
solution of a specific costbased problem.
The parameters are codified as chromosomes that are ranked with a
fitness function, and the best fits are reproduced through genetic operators:
crossover and mutation. This process continues until the fitness of the best
chromosome reaches a preset threshold.
124 Unmanned Aerial Vehicles (UAV) and Drones

Table 2. Parameter to identify list

The main steps of the process for identification using GA’s have been:

Parameters codification
The parameters are codified with real numbers, as it is more intuitive format
than large binary chains. Different crossover operations have been studied:
• The new chromosome is a random combination of two chains.

• The new chromosome is a random combination of random genes

with values in the range defined by the ascendant genes.
 Modelling and Control Prototyping of Unmanned Helicopters 125

The first operator transmits the positive characteristics of its ascendants

while the second one generates diversity in the population as new values
appear. In addition to the crossover operator there is a mutation algorithm.
The probability of mutation for the chromosomes is 0.01, and the mutation
is defined as the multiplication of random genes by a factor between 0.5 and
1.5.
When the genetic algorithms falls into a local minimum, (it is detected
because there is no a substantial improvement of the fitness in the best
chromosome during a large number of iterations), the probability of
mutation have to be increased to 0.1. This improves mutated populations
with increased probability of escaping from the local minimum.

Initial population
The initial population is created randomly with an initial set of parameters
of a stable model multiplied by random numbers selected by a Monte-Carlo
algorithm with a normal distribution with zero mean and standard deviation
of 1.
The genetic algorithm has been tested with different population sizes,
between thirty and sixty elements. Test results showed that the bigger
population did not lead to significantly better results, but increased the
computation time. Using 20 seconds of flight data and a population of 30
chromosomes, it took one hour to process 250 iterations of the genetic
algorithm on a Pentium IV processor. Empirical data suggests that 100
iterations are enough to find a sub-optimal set of parameters.

Fitness function
The fitness function takes into consideration the following state variables:
roll, pitch, yaw and speed. Each group of parameters (chromosome)
represents a model, which can be used to compute simulated flight data
from the recorded input commands. The difference between simulated and
real data is determined with a weighted least-squares method and used as the
fitness function of the genetic algorithm.
In order to reduce the effect of the error propagation to the velocity
due to the estimated parameters that have influence in attitude, the global
126 Unmanned Aerial Vehicles (UAV) and Drones

process has been decomposed in two steps: Attitude and velocity parameters
identification.
The first only identifies the dynamic response of the helicopter attitude,
and the genetic algorithm modifies only the parameters related to the
attitude. Once the attitude-related parameters have been set, the second
process is executed, and only the velocity-related parameters are changed.
This algorithm uses the real attitude data instead of the model data so as not
to accumulate simulation errors. Using two separate processes for attitude
and velocity yields significantly better results than using one process to
identify all parameters at the same time.
The parameters related to the attitude process are: TwstTr, IZ, HTr,
YuuVt, YuvVt, Corr1, Corr2, Tproll, Tppitch, Kroll, Kpitch, Kyaw, DTr,
DVt, YMaxVt, KGyro, KdGyro, Krp, Kpr, OffsetRoll, OffsetPitch and
OffsetYaw. The parameters related to the vehicle’s speed are TwstMr, KCol,
XuuFus, YvvFus, ZwwFus and OffsetCol.
The fitness functions are based on a weighted mean square error equation,
calculated by comparing the real and simulated responses for the different
variables (position, velocity, Euler angles and angular rates) applying a
weighting factor.

Figure 10. Probability function for selection of elements

The general process is described bellow:
• To create an initial population of 30 elements.
• To perform simulations for every chromosome;
• Computation of the fitness function.
• Classification the population using the fitness function as
the index. The ten best elements are preserved. A Monte-
 Modelling and Control Prototyping of Unmanned Helicopters 127

Carlo algorithm with the density function shown in Figure 10.

Probability function for selection of elements is used to determine
which pairs will be combined to generate 20 more chromosomes.
The 10 ‘better’ elements are more likely (97%) to be combined
with the crossover operators.
• To repeat from step 2 for a preset number of iterations, or until a
preset value for the fitness function of the best element is reached.

Data acquisition
The data acquisition procedure is shown in Figure 11. Helicopter is manually
piloted using the conventional RC emitter. The pilot is in charge to make the
helicopter performs different maneuvers trying to excite all the parameters
of the model. For identification data is essential to have translational flights
(longitudinal and lateral) and vertical displacements. All the commands
provided by the pilot are gathered by a computer trough to a USB port by
using a hardware signal converter while onboard computer is performing
data fusion from sensors and sending the attitude and velocity estimation to
the ground computer using a WIFI link.
In this manner inputs and outputs of the model are stored in files to
perform the parameters identification.

Figure 11. Data acquisition architecture

IDENTIFICATION RESULTS
The identification algorithm was executed several times with different
initial populations and, as expected, the sub-optimal parameters differ.
Some of these differences can be explained by the effect of local minimum.
Although the mutation algorithm reduces the probability of staying inside a
local minimum, sometimes the mutation factor is not big enough to escape
from it.
128 Unmanned Aerial Vehicles (UAV) and Drones

The evolution of the error index obtained with a least-squares method

in different cases is shown in Figure 12. The left graph shows the quick
convergence of the algorithm in 50 iterations. On the other hand the right
graph shows an algorithm that fell in a local minimum and had an overall
lower convergence speed. Around the 50th step a mutation was able to escape
the local minimum, and the same behavior is observed in the 160th step.

Figure 12. Error evolution for two different cases

The result may be used as the initial population for a second execution
of both processes. In fact, this has been done three times to obtain the best
solution. The analysis of cases where mutation was not able to make the
algorithm escaped from local minimum, led to the change of the mutation
probability from 0.1 to 1 when detected.
On the other hand, not all the parameters were identified at the same
time, actually two iterative processes were used to identify all parameters.
The first process used 100 steps to identify the parameters related to the
modeling of the helicopter’s attitude, beginning with a random variation
of a valid solution. The second process preserved the best element of the
previous process’ population, and randomly mutated the rest. After 100
steps the parameters related to the helicopter’s speed were identified.
The simulated attitude data plotted as a blue line against real flight data
in red, is shown for roll, pitch, yaw angles in Figure 13. They give you an
idea about how simulations follow real tendencies even for small attitude
variations.
 Modelling and Control Prototyping of Unmanned Helicopters 129

Figure 13. Roll, Pitch and Yaw real vs simulated

On the other hand, the results obtained for the velocity analysis are shown
in Figure 14.

Figure 14. Velocity Simulation Analysis

The quality of the simulation results is the proof of a successful identification
process both for attitude and speed.
The simulated roll and yaw fit accurately the registered helicopter
response to input commands. The simulated pitch presents some problems
due to the fact that the flight data did not have a big dynamic range for this
signal. Nevertheless the error is always below three degrees.
For the simulation of the vehicle’s velocity good performance was
obtained for both frontal and lateral movement. The unsuccessful modeling
of vertical velocity can be linked to the sensors available onboard the
130 Unmanned Aerial Vehicles (UAV) and Drones

helicopter. Vertical velocity is registered only by the differential GPS, and

statistical studies of the noise for this sensor show a standard deviation
of 0.18 meters per second. In other words, the registered signal cannot be
distinguished from the noise because the registered flight did not have a
maneuver with significant vertical speed; therefore modeling this variable
is not possible.
It is important to analyze the values of the obtained parameters since
the parameters are identified using a genetic algorithm. The dispersion of
the value of the parameters for five different optimization processes was
analyzed. Most of the parameters converge to a defined value, which is
coherent with the parameter’s physical meaning, but sometimes, some
parameters were not converged to specific values, usually when no complete
set of flights was used. Thus, if no vertical flights were performed, the
parameters regarding vertical movements turned to be with a great dispersion
in the obtained values.
This conclusion can be extended to different optimization processes
for different groups of flight data. In other words, several flights should be
recorded, each with an emphasis on the behaviors associated to a group of
parameters. With enough flight data it is trivial to identify all the parameters
of the helicopter model.

CONTROL PROTOTYPING

Figure 15. Control Developing template

In order to develop and test control algorithms in a feasible way, the
proposed model has been encapsulated into a Matlab-Simulink S-Function,
as a part of a prototyping template as Figure 15 shows. Others modules have
been used for performing realistic simulations, thus a sensor model of GNC
 Modelling and Control Prototyping of Unmanned Helicopters 131

systems has also been created. Auxiliary modules for automaticmanual

switching have been required for testing purposes.
The proposed control architecture is based on a hierarchic scheme with
five control levels as Figure 16 shows. The lower level is composed by
hardware controllers: speed of the rotor and yaw rate. The upper level is
the attitude control. This is the most critical level for reaching the stability
of the helicopter during the flight. Several techniques have been tested on it
(classical PI or fuzzy controllers). The next one is the velocity and altitude
level. The fourth is the maneuver level, which is responsible for performing
a set of pre- established simple maneuvers, and the highest level is the
mission control.
Following sections will briefly describe these control levels from the highest
to the lowest one.

Figure 16. Control Architecture

Mission control
In this level, the operator designs the mission of the helicopter using GIS
(Geographic Information System) software.
These tools allow visualizating the 3D map of the area by using Digital
Terrain Model files and describing a mission using a specific language.
(Gutiérrez et al -06). The output of this level is a list of maneouvers (parametric
commands) to be performed by the helicopter.

Maneuver Control
This control level is in charge of performing parametric manoeuvres such as
132 Unmanned Aerial Vehicles (UAV) and Drones

flight maintaining a specific velocity during a period of time, hovering with

a fix of changing yaw, forward, backward or sideward flights and circles
among others. The output of this level are velocity commands.
Internally, this level performs a prediction of the position of the
helicopter, applying acceleration profiles (Figure 17).
Velocity and heading references are computed by using this theoretical
position and the desired velocity in addition to the real position and velocity
obtained from sensors.

Figure 17. Maneuvers Control Scheme

A vertical control generates references for altitude control and Yaw control.
In manoeuvres that no high precision of the position is required, i.e. forwared
flights, the position error is not taken into account, because the real objective
of the control is maintain the speed. This system allow manage a weighting
criteria for defining if the main objetive is the trajectory or the speed profile.
The manoeuvres that helicopter is able to perform are an upgradeable list.

Velocity Control
The velocity control is in charge of making the helicopter maintains the
velocity that maneouvers control level computes. The velocity can be referred
to a ENU (East-North-Up) frame or defined as a module and a direction.
Therefore, a coordinate transformation is required to obtain lateral and
frontal velocities (vl, vf) that are used to provide Roll and Pitch references.
This transformation is very sensitive to the yaw angle estimation.
 Modelling and Control Prototyping of Unmanned Helicopters 133

Figure 18. Velocity control structure

Control maneuver has also to provide with the yaw reference due to the
capability of helicopters for flying with derive (different bearing and
heading).
Vertical velocity is isolated from the horizontal because it is controlled
by using the collective command.

Figure 19. Velocity control example

Concerning to the control algorithms used to test the control architecture,
two main problems have been detected and solved: The first one is based on
the fact that the speed is very sensitive to the wind and payload. Due to this,
a gain-scheduller scheme with a strong integration efect has been requiered
in PI algorithm. (Figure 19)
134 Unmanned Aerial Vehicles (UAV) and Drones

The second one arises from the solution of the first problem. Thus, when
operator performs a manual-automatic swiching by using channel nine of
the emitter, integral action needs to be reset for reducing the effect of the
error integration when manual mode is active.

Attitude Control
The attitude control is in charge of providing with commands to the servos (or
hardware controllers). Several control techiques were tested in this level, but
probably fuzzy controllers turned out to have the best performance. Figure
20 shows the proposed architecture and Figure 21 the control surfaces used
by fuzzy controller.
As it can be observed, the Yaw have been isolated from the multivariable
Roll-Pitch controler with a reasonable quality results.

Figure 20. Attitude control structure

Once the control structure has been designed, the first step is to perform
simulations in order to analyze if the control fulfills all the established
requirements. Considering the special characteristics of the system to
control, they main requirements to be taken into account are:
• No (small) oscillations for attitude control. In order to transmit a
confidence in the control system to the operator during the tests.
• No permanent errors in velocity control. This is an important
concept to consider for avoiding delays in long distance missions.
• Adjustable fitness functions for maneuvers control. This allows
characterizing the mission where position or velocity is the most
important reference to keep.
The operator relies upon all the tools that Matlab-Simulink provides and
an additional 3D visualization of the helicopter as Figure 22 shows for test
performing.
 Modelling and Control Prototyping of Unmanned Helicopters 135

Figure 21. Control surfaces for aAttitude level

The controller block has to be isolated for automatic C codification by
using the embedded systems coder of Matlab-Simulink when simulation
results are satisfactory.Then, the obtained code can be loaded into onboard
computer for real testing (Del-Cerro 2006).

Figure 22. Simulation framework

136 Unmanned Aerial Vehicles (UAV) and Drones

Figure 23. Roll and pitch Control results during real flight

CONTROL RESULTS
Figure 23 shows the results obtained for attitude control during an eighty-
seconds real flight with the Vario helicopter.
Lines in red show the real attitude of the helicopter while green ones are
the references provided by the velocity control level.
Dotted blue line indicates when the helicopter is in manual (value of
1) or automatic (value of 0). It can be observed that, when helicopter is in
manual mode, the red line does not follow the references of the automatic
control, because the pilot is controlling the helicopter. Figure 24 shows the
obtained results for velocity control. The same criteria that in last graphic
has been used, thus red line indicates real helicopter response and green
means references from maneuvers control level. Dotted blue line is used
to show when the helicopter flies autonomously or remote piloted. Small
delays can be observed.
 Modelling and Control Prototyping of Unmanned Helicopters 137

Figure 24. Velocity control results

Figure 25 shows two examples of the maneuvers control performance.
The right side graph shows the position when helicopter is moving in a
straight line. It can be observed that maximum error in transversal coordinates
is less than one meter. This result can be considered as very good due to is
not a scalable factor. The precision of the position control reduces the error
when trajectory is close to the end by weighting the error position against
the velocity response.
On the other hand, the left side shows a square trajectory. The speed
reference performs smooth movementes and therefore the position control is
less strong that in the first case. Even in this case, position errors are smaller
than two meters in the worst case.
In general, a high precision flight control has been developed, because
sub-metric precision has been reached during the tests with winds slower
than twenty kilometers per hour.
138 Unmanned Aerial Vehicles (UAV) and Drones

Figure 25. Maneuvers control results

FINAL OBSERVATIONS AND FUTURE WORK

The first part of this chapter describes a procedure for modeling and
identification a small helicopter. Model has been derived from literature, but
changes introduced in the original model have contributed to improve the
stability of the model with no reduction of the precision.
The proposed identification methodology based on evolutionary
algorithms is a generic procedure, having a wide range of applications. The
identified model allows performing realistic simulations, and the results
have been validated.
The model has been used for designing real controllers on a real
helicopter by using a fast prototyping method detailed in section 4.
The model has confirmed a robust behavior when changes in the flight
conditions happen. It does work for hover and both frontal and lateral non-
aggressive flight. The accuracy and convenience of a parametric model
depends largely on the quality of its parameters, and the identification
process often requires good or deep knowledge of the model and the modeled
phenomena. The proposed identification algorithm does away with the
complexity of model tuning: it only requires good-quality flight data within
the planned simulation envelope. The genetic algorithm has been capable
of finding adequate values for the model’s 28 parameters, values that are
coherent with their physical meaning, and that yields an accurate model.
It is not the objective of this chapter to study what the best control technique
is. The aim is to demonstrate that the proposed model is good enough to
perform simulations valid for control designing and implementation using
 Modelling and Control Prototyping of Unmanned Helicopters 139

an automatic coding tool. Future work will focus on the following three
objectives:
• To develop a model of the engine in order to improve the model
behavior for aggressive flight maneuvers.
• To compare this model with others in the literature in a wide
range of test cases.
• To validate the proposed model with different helicopters.
140 Unmanned Aerial Vehicles (UAV) and Drones

REFERENCES
1. Castillo P, Lozano R, Dzul A. (2005) Modelling and Control of
Mini-Flying Machines. Springer Verlag London Limited 2005.
ISBN:1852339578.
2. Del-Cerro J, Barrientos A, Artieda J, Lillo E, Gutiérrez P, San- Martín
R, (2006) Embedded Control System Architecture applied to an
Unmanned Aerial Vehicle International Conference on Mechatronics.
Budapest-Hungary
3. Gavrilets, V., E. Frazzoli, B. Mettler. (2001) Aggresive Maneuvering of
Small Autonomous Helicopters: A. Human Centerred Approach. The
International Journal of Robotics Research. Vol 20 No 10 , October
2001. pp 795-807
4. Godbole et al (2000) Active Multi-Model Control for Dynamic
maneuver Optimization of Unmanned Air Vehicles. Proceedings of the
2000 IEEE International conference on Robotics and Automation. San
Francisco, CA. April 2000.
5. Gutiérrez, P., Barrientos, A., del Cerro, J., San Martin., R. (2006)
Mission Planning and Simulation of Unmanned Aerial Vehicles with
a GIS-based Framework; AIAA Guidance, Navigation and Control
Conference and Exhibit. Denver, EEUU, 2006.
6. Heffley R (1988). Minimum complexity helicopter simulation math
model. Nasa center for AerosSpace Information . 88N29819.
7. La Civita M, Messner W, Kanade T. Modeling of Small-Scale
Helicopters with Integrated First-Principles and System-Identification
Techniques. American Helicopter Society 587h Annual Forum,
Montreal, Canada, June 11-13,2002.
8. Mahony-R Lozano-R, Exact Path Tracking (2000) Control for an
Autonomous helicopter in Hover Manoeuvres. Proceedings of the
2000 IEEE International Conference on Robotics & Automation. San
Francisco, CA. APRIL 2000. P1245-1250
9. Mettler B.F., Tischler M.B. and Kanade T.(2002) System Identification
Modeling of a ModelScale Helicopter. T. Journal of the American
Helicopter Society, 2002, 47/1: p. 50-63. (2002).
10. Mettler B, Kanade T, Tischler M. (2003) System Identification Modeling
of a Model-Scale Helicopter. Internal Report CMU-RI-TR-00-03.
 CHAPTER

7
Contour Based Path Planning with
B-Spline Trajectory Generation for
Unmanned Aerial Vehicles (UAVs) over
Hostile Terrain

Ee-May Kan1 , Meng-Hiot Lim1 , Swee-Ping Yeo2 , Jiun-Sien Ho3 ,

Zhenhai Shao4
1
Nanyang Technological University, Singapore City, Singapore
2
National University of Singapore, Singapore City, Singapore
3
Temasek Polytechnic, Singapore City, Singapore
4
University of Electronic Science Technology of China, Chengdu, China.

ABSTRACT
This research focuses on trajectory generation algorithms that take into
account the stealthiness of autonomous UAVs; generating stealthy paths

Citation: E. Kan, M. Lim, S. Yeo, J. Ho and Z. Shao, “Contour Based Path Planning
with B-Spline Trajectory Generation for Unmanned Aerial Vehicles (UAVs) over Hos-
tile Terrain,” Journal of Intelligent Learning Systems and Applications, Vol. 3 No. 3,
2011, pp. 122-130. doi: 10.4236/jilsa.2011.33014.
Copyright: © 2011 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
142 Unmanned Aerial Vehicles (UAV) and Drones

through a region laden with enemy radars. The algorithm is employed to

estimate the risk cost of the navigational space and generate an optimized path
based on the user-specified threshold altitude value. Thus the generated path
is represented with a set of low-radar risk waypoints being the coordinates
of its control points. The radar-aware path planner is then approximated
using cubic B-splines by considering the least radar risk to the destination.
Simulated results are presented, illustrating the potential benefits of such
algorithms.
Keywords: Unmanned Aerial Vehicles (UAVs), Radar, Path Planning,
B-Splines

INTRODUCTION
Unmanned aerial vehicles (UAVs) are increasingly being used in real-world
applications [1-3]. They are typically surveillance and reconnaissance
vehicles operated remotely by a human operator from a ground control
station; they have no on-board guidance capabilities that give them some
level of autonomy, for example, to re-plan a trajectory in the event of a
change in the environment or mission. With such rudimentary capabilities,
only simple tasks can be accomplished and the operation is also limited
to simple or uncomplicated situations, typically in well-characterized
environments. It is useful to endow UAVs with more advanced guidance
capabilities, in particular capabilities that increase the vehicle’s autonomy
to allow for more complex missions or tasks. Currently operating UAVs
use rudimentary guidance technologies, such as following pre-planned or
manually provided waypoints. Over the years, advances in software and
computing technology have fuelled the development of a variety of new
guidance methodologies for UAVs. The availability of various guidance
technologies and understanding of operational scenarios create opportunities
towards refining and implementing such advanced guidance concepts.
The basic difficulties include partially known and changing environment,
extraneous factors, such as threats, evolving mission elements, tight timing
and positioning requirements. Moreover, these must be tackled, while at
the same time, explicitly accounting for the actual vehicle manoeuvring
capabilities, including dynamics and flight-envelope constraints. The term
flight envelope refers to the parameters within which an aerial vehicle may
be safely flown under varying, though expected, wind speed, wing loading,
wind shear, visibility and other flight conditions without resorting to
extreme control measures such as abnormal spin, or stall recovery, or crash
 Contour Based Path Planning with B-Spline Trajectory Generation ... 143

landing. These aerial vehicles are usually mobilized to carry out critical
missions in high-risk environments, particularly in situations where it may
be hazardous for human operators. However, such missions demand a high
level of stealth, which has a direct implication on the safety and success of
the mission. Therefore it is important to minimize the risk of detection or
more specifically, the probability of detection of the UAVs by enemy radars.
There has been extensive research in the area of path planning especially
in the artificial intelligence, and optimization with most restricted to two-
dimensional (2D) paths [4,5]. Different conventional approaches have been
developed to solve the path planning problem, such as the cell decomposition,
Djikstra’s algorithm, road map and potential field [6-9]. Sleumer and
Tschichold [6] present an algorithm for the automatic generation of a map
that describes the operation environment of a building consisting of line
segments, representing the walls as an input. The algorithm is based on
the exact cell decomposition approach and generates a connectivity graph
based on cell indices. However, for successful implementation of exact
cell decomposition, it is required that the geometry of each cell be simple.
Moreover, it is important to test the adjacency of two cells to find a path
crossing the portion of the boundary shared by the two cells. Both exact cell
decomposition and approximate cell decomposition methods are accurate,
yet may be computationally expensive when used in a terrain environment
since numerous obstacles at varying elevations could be found. They are only
effective when the environment contains a countable number of obstacles.
Another technique used to effectively represent the alternate paths
generated by a path planner is to use Bsplines. B-splines allow a parametric
curved path to be described by only a few control points rather than the
entire curve segmented which could be as many as thousands of sections,
depending on the length and curvature of the path. Recent methods include
modelling the trajectory using splines based on elastic band theory [10,11],
and interactive manipulation of control points for spline trajectory generation
[12,13]. A series of cubic splines was employed in [14] to connect the straight
line segments in a near-optimal manner, whereas the algorithm presented in
[15] yields extremal trajectories that transition between straight-line path
segments smoothly in a time-optimal fashion.
Previous efforts [16,17] have concentrated on path planning over a
hostile radar zones based on the amount of energy received by enemy radar at
several locations. In this paper, we investigate on the number and placement
of control points within a hostile region by considering the complexity of
144 Unmanned Aerial Vehicles (UAV) and Drones

the terrain. We take into account the path optimality within a hostile region
by considering the complexity of the terrain. We propose an efficient
algorithm to identify the waypoints based on the topographic information
of the enemy terrain. The waypoint generation process is based on the user-
specified threshold flying altitude. The threshold altitude to avoid radar
detection and terrain collisions is determined based on the knowledge of the
enemy terrain. The navigational path between the waypoints is considered
when minimizing the exposure to a radar source. Additionally, the generated
trajectory is of minimal length, subject to the stealthy constraint and satisfies
the aircraft’s dynamic constraints.
Details on the formulation of the problem are presented in this paper.
We model this problem as described in the following section and adopt
a heuristic-based approach to solve it. The rest of this paper is organized
as follows: Section 2 gives a description of the problem model used for
measuring stealth in this work and illustrates the calculation of the radar
risks. Section 3 demonstrates the details of the problem formulation and
Section 4 presents the solution approach of the route planner, substantiated
with simulated results. The conclusion and future work are presented in
Section 5.

MODELLING
In this section, the model we used throughout this work is presented. The
integrated model of the UAV and radar has the following features. The Radar
Cross Section (RCS) of the UAV depends on both the aspect and bank angles
whereas the turn rate of the UAV is determined by its bank angle. Hence, the
RCS and aircraft dynamics are coupled through the aspect and bank angles.

Measuring Stealth Based on Aircraft Model

The bank-to-turn aircraft is assumed to move in a horizontal plane at a
constant altitude according to the equations [18],

(1)
where x and y are the Cartesian coordinates of the aircraft, ϕ is the
heading angle as shown in Figure 1, v is the constant speed, u is the input
signal and the acceleration is normal to the flight path vector, followed by
U is the maximum allowable lateral acceleration. Figures 2 and 3 show the
dependence of RCS on aspect and bank angles.
 Contour Based Path Planning with B-Spline Trajectory Generation ... 145

(2)

Figure 1. Aircraft position, velocity, azimuth, heading, and aspect angles.

Figure 2. RCS as a function of aspect angle.

Figure 3. RCS as a function of bank angle.

146 Unmanned Aerial Vehicles (UAV) and Drones

be the azimuth, aspect, and elevation angles, respectively, where z is the

aircraft altitude. Let the bank angle μ be given by

(3)
where g is the acceleration of gravity. We model the RCS of the aircraft
as function of the aspect angle λ, the elevation angle φ, and the bank angle
μ, so that
(4)
The RCS obtained from Equation (4) is used as an estimate value in
Equation (6) for measuring stealth in this work. As an example, real aircraft
RCS measurements as functions of aspect and bank angles are shown in
Figures 3 and 4 [18]. The following section illustrates the calculation of the
cost associated to radar risk that will be undertaken by the UAV.

Radar Model
The radar model in this work will be presented in terms of its inputs (aircraft
range and RCS) and output (an estimate of the probability that an aircraft
can be tracked for an interval of time). For the sake of simplicity, we assume
that the radar is located at ( x , y, z) of the navigational space. Let

(5)

Figure 4. Square cylinder for RCS modelling.

be the aircraft range from the enemy radar to the aircraft. The radar detects
the aircraft by receiving a sequence of radio frequency pulses reflected from
 Contour Based Path Planning with B-Spline Trajectory Generation ... 147

it at fixed observation times. To generate a stealthy path for the UAV, we have
to take into account the energy reflected to enemy radar site as the enemy
radar tracks the location, speed, and direction of an aircraft. The range at
which radar can detect an object is related to the power transmitted by the
radar, the fidelity of the radar antenna, the wavelength of the radar signal,
and the RCS of the aircraft [19]. The RCS is the area a target would have
to occupy to produce the amount of reflected power that is detected back at
the radar. RCS is integral to the development of radar stealth technology,
particularly in applications involving aircraft. For example, a stealth aircraft
which is designed to be undetectable will have design features that give it a
low RCS. Low RCS offers advantages in detection range reduction as well
as increasing the effectiveness of decoys against radar-seeking threats.
The size of a target’s image on radar is measured by the RCS and
denoted by the symbol σ and expressed in square meters. The azimuth and
range detected by the radar serve as the inputs to a tracking system, typically
based on one or more Kalman filters.
he purpose of the tracking system is to provide a predicted aircraft
position and velocity so that a decision can be made to launch a missile and
guide it to intercept. RCS modelling and reduction is crucial in designing
stealth weapon systems. We make use of Finite-Difference Time-Domain
(FDTD) Method technique in RCS modelling. The numerical technique is
general, geometry-free, and can be used arbitrarily for any target, within the
limitations of computer memory and speed. Figure 4 shows the geometric
surface for the RCS simulation with the FDTD algorithm.
The FDTD-based package was supplied in [20] for the RCS prediction
of various simple canonical targets. The complete RCS signature of a target
is described by the frequency response of the target illuminated by plane
waves from all possible physical angles. Realistic targets are frequently very
large and are often largely made up of highly conductive materials.
The FDTD-based RCS prediction procedure is as follows:
• The target, located in the computational space, is illuminated by a
Gaussian-pulse plane wave from a given direction;
• The scattered fields are simulated inside the computational
volume until all transients dissipate;
• The time-domain far fields in the spherical coordinates are then
extrapolated by using equivalent current via the near-field to far-
field routine, over a closed, virtual surface surrounding the target.
148 Unmanned Aerial Vehicles (UAV) and Drones

For either vertical or horizontal planes, co-polarized or cross-polarized RCS

behaviour is computed after the simulation for a given frequency range.
Figure 5 illustrates the RCS behaviour with respect to the azimuth angle
based on the RCS modelling described above.
Subsequently the computation of the energy received by a radar is given
by the radar range equation [21]

(6)
where S is the signal energy received by the radar, Pavg is the average
power transmitted by the radar, G is the gain of the radar antenna, σ is the
radar cross section of the target, Ae is the effective area of the radar antenna,
tot is the time the radar antenna is pointed at the target, and R is the range to
the target. Every radar has a minimum signal energy that it can detect, Smin.
This minimum signal energy determines the maximum range (Rmax) at which
the radar can detect a given target.

(7)

Figure 5. RCS behaviour with respect to the azimuth angle.

According to Equation (7), the distance at which a target can be detected
for given radar configuration varies with the fourth root of its RCS. Figure 6
gives some understanding of just how little radar power is typically reflected
back from the target and received by the radar. In this case, the target presents
the same aspect to the radar at ranges from 1 to 50 miles. At a range of 50
miles, the relative power received by the radar is only 1.6 × 10–5% of the
strength at one mile. This diagram graphically illustrates how significant
the effect of energy dissipation is with distance, and how sensitive radars
 Contour Based Path Planning with B-Spline Trajectory Generation ... 149

must be to detect targets at even short ranges. Hence the cost associated
to radar risk is based on a UAV’s exposure to enemy radar. In this work,
the cost of radar risk involves the amount of energy received by enemy
radar at several locations based on the distance between the UAV and the
radar. We assume that the UAV’s radar signature is uniform in all directions
and is proportional to 1/R4 (where R is the distance between the UAV and
the enemy radar); the UAVs are assigned simplified flight dynamics and
performance constraints in two-dimensions; and all tracking radars are given
simplified detection properties. Based on the assumptions stated above, we
classify the navigational space into different levels of risk as described in
the following chapter. The next step is to compute a stealthy path, steering
the UAV clear and around known radar locations. The generated path should
carry minimal radar risks, and adhere to physical and motion constraints.
Also, the distance between the UAV and the radars should be maximized in
order to minimize the energy received at the radar.

PROBLEM FORMULATION
To describe the problem, consider the contour based navigational space
shown in Figure 7. The objective is to find a path which minimizes the
UAV’s exposure to radar sites (small triangles) from the current UAV
position (circle) to the target location (star). In the present study, it is
assumed that the hostile radar zones are known beforehand. The contours
are used to denote elevation and depth on maps. From these contours, a
sense of the general terrain can be determined. The details of the radar zone
can be acquired using satellite data or from surveillance data. The start point
and the end point of the flight are known and it is required to find a feasible
path connecting these two points. The threshold altitude value is determined
by the user which assumes knowledge of the enemy terrain. By flying at
a userspecified threshold altitude value, the aircraft can take advantage of
terrain masking and avoid detection by enemy radars. The aircraft works by
transmitting a radar signal towards the ground area and the signal returns
can then be analysed to see how the terrain ahead varies, which can then be
used by the aircraft’s autopilot to specify the threshold altitude value. The
specified threshold allows the aircraft to fly at reasonably constant height
above the earth so as to avoid the radar detection. In this work, the path
of the UAV is smoothened using cubic Bsplines, which is controlled by
manipulating a number of control points. The generated path should carry
minimal radar risks, and adhere to physical and motion constraints.
150 Unmanned Aerial Vehicles (UAV) and Drones

Figure 6. Reduction in the strength of target echoes with range.

Figure 7. Contour based navigational space with known radar sites.

Given a single radar located at the origin and a single aircraft travelling
from initial point to the destination, Pachter [22] showed that an objective
function for minimizing cost, is

(8)
where v is the constant speed of the aircraft and l is the path length.
A closed form solution to this problem was obtained using the calculus of
variations, with the limitation that the aircraft traverses an angle with respect
 Contour Based Path Planning with B-Spline Trajectory Generation ... 151

to the radar of less than 60˚. Beyond this limit, the optimal path length is
infinite but the cost remains finite. The objective cost function Equation (8),
is augmented to include the rest of the radar at the navigational space, giving

(9)
The cost associated with radar risk involves the amount of energy αi ,
received by enemy radar based on the distance from the UAV to the radar.
We first look for appropriate waypoints that the UAV should traverse in
planning a path which minimizes the radar exposure.

GENERATION OF NODES
The procedure starts by requesting from the user the detection altitude or
maximum elevation desired which assumes knowledge of the enemy terrain.
Nodes below the detection altitude are extracted and used as base in the
generation of straight line segments. The nodes above the detection altitude
are discarded, given that at those elevations the UAVs could be detected.
Appropriate value for the parameter is based on the user’s knowledge of: the
terrain; the operation of the UAVs; the purpose, logistics, and safety of the
mission. This means that the user’s input is a determinant in the quality of
the resulting path. As shown in Figure 8, the initial nodes can be searched by
specifying the detection altitude for the UAV. The next step is to make use
of heuristic search algorithm to compute a stealthy path, steering the UAV
clear and around known radar locations.

IMPLEMENTATION

Generation of Nodes
This search algorithm works by starting at UAV’s present position (red
circle). It adds the heuristic function value and determines the expansion
order of nodes. We make use of a distance-plus-cost heuristic function, f(n)
to determine the order in which the search visits nodes in the space. We
assume that the UAV’s radar signature is uniform in all directions and is
proportional to 1/R4 where R is the aircraft range. We modify the evaluation
function by adding the cost associated with radar risk to the f(n) as follows:
152 Unmanned Aerial Vehicles (UAV) and Drones

(10)
where w is a value in the range [0, 1]; g(n) is the pathcost function which
shows the intensity between the current node and node n; h(n) is the distance
value which is estimated from node n to the goal node; ci is relative to Ri
and m is the number of enemy radars located at the navigational space. The
h(n) plays a significant role in the search process when w value is smaller;
the number of vertices to be explored would be decreased. The choice of w
between 0 and 1 gives the flexibility to place weight on exposure to threats or
fuel expenditure depending on the particular mission scenario. For example
when w = 0, the search process would proceed from the starting location to
the target location by exploring the least number of vertices. However the
generated path is not stealthy and may cause UAV to enter high radar risk
region. A w value closer to 0 would result in the shortest paths, with little
regard for the exposure to enemy radar, while a value closer to 1 would result
in paths that avoid threat exposure at the expense of longer path length. For
this case, the cost weighting parameter w was chosen to give a reasonable
tradeoff between proximity of the path to threats and the path length. The
evaluation function, (10) is tested empirically and compared in terms of
the number of nodes visited. During the test, the cost weighting parameter
w is varied from 0.0 to 1.0 and the test result is demonstrated in Table 1.
The experimentally defined value for the w ranges from [0.1, 0.3] in which
the generated path is optimal and computationally inexpensive, which is in
line with our algorithm objectives. The generated path is neither the safest
possible path nor the shortest possible path, but represents a compromise
between the two objectives.

Figure 8. Generated waypoints based on user-specified threshold altitude value.

 Contour Based Path Planning with B-Spline Trajectory Generation ... 153

The algorithm traverses various paths from the starting location to the goal.
At the same time, it compares the risk cost of all the vertices. Starting with
the initial node, it maintains a priority queue of nodes to be traversed, known
as the openQueue (see Listing 1). The lower f(n) for a given node n, the
higher its priority. At each step of the algorithm, the node with the lowest
f(n) value is removed from the queue, the f and h values of its neighbors.

Table 1. Test results.

The algorithm traverses various paths from the starting location to the
goal. At the same time, it compares the risk cost of all the vertices. Starting
with the initial node, it maintains a priority queue of nodes to be traversed,
known as the openQueue (see Listing 1). The lower f(n) for a given node n,
the higher its priority. At each step of the algorithm, the node with the lowest
f(n) value is removed from the queue, the f and h values of its neighbors are
updated accordingly, and these neighbors with the least radar cost are added
to the openQueue. The conditions of adding a node to the openQueue are as
follows:
• The node is not in high radar risk region;
• The node is movable and does not exist in the openQueue;
• The distance between the starting point and the current node is
shorter than the earlier estimated distance;
• The distance between the starting point and the current node does
not violate the pre-specified detection altitude.
If the four conditions are satisfied, then the current node will be added
to the openQueue. The algorithm continues until a goal node has a lower f
154 Unmanned Aerial Vehicles (UAV) and Drones

value than any node in the queue. If the actual optimal path is desired, the
algorithm may also update each neighbor with its immediate predecessor in
the best path found so far; this information can then be used to reconstruct the
path by working backwards from the goal node. The output of this stage is a
set of straight line segments connecting a subset of the nodes resulting from
the waypoint generation process, as long as the link does not violate the pre-
specified detection altitude. The path planning algorithm is implemented in
a MATLAB environment and a sample of the result can be seen in Figure 9.
The pseudo-code for the algorithm is as shown in Listing 1.
The next step is to modify the generated path by using a series of cubic
splines to optimize the path by moving away from high radar risks and
improving the navigability of the path for the UAV.

Figure 9. Generated path with least radar risk based on user-specified detection
altitude.

Trajectory Generation
As a first step, the path is parameterized in both x and y directions. An initial
spline knot is fixed at the UAV starting location, while the final spline knot
is fixed at the first edge of the path. We begin by splitting each edge of the
path into three segments, each one described by a different B-spline curve
[23]. In this work, we calculate the radar risk cost at several locations along
each edge and take the length of the edge into account. The radar risk cost
 Contour Based Path Planning with B-Spline Trajectory Generation ... 155

was calculated at three points along each edge: Li/6, Li/2, and 5Li/6, where
Li is the length of edge i. The radar risk cost for each edge may be calculated
by

(11)
where N is the number of enemy radars; M is the number of discrete
point and d1/6,i,j is the distance from the 1/6th point on the i th edge to the j
th enemy radar. The goal now is to find the (x, y) locations which minimize
the exposure to the enemy radar for the middle knots. We apply the graph
search algorithm to look for the middle knots with minimal risk cost for each
edge of the path. We determine the interpolating curve based on tangent
vectors and curvature vectors for each pair of points. The interpolation
problem is solved by assuming p = 3, which produces the C2 cubic spline
[24]. The parameters ûk are used to determine the knot vector u as

(12)
Since the number of control points, m + 1, and the number of the knots,
nknot + 1, are related by nknot = m + 4 and, as can be easily deduced from (12),
nknot = n + 6, the unknown variables pj are n + 3. Once the control points pj, j
= 0, , n + 2 are known, and given the knot vector u, the B-spline is completely
defined. Once this step has been completed, the procedure switches to the
algorithm (as shown in Listing 2) to determine how to connect the knots.
Note that some experimentally defined knots are added by the algorithm in
order to smoothen the spline curve.
156 Unmanned Aerial Vehicles (UAV) and Drones

Function search_path (start, goal)

closedQueue: = the empty queue/*The set of nodes
already evaluated*/
openQueue: = queue containing the initial node/
*The set of tentative nodes to be evaluated*/
g[start]: = 0/*Distance from start along optimal
path*/
h[start]: = estimate_of_distance (start, goal)
f[start]:= h[start]/*Estimated total distance from
start to goal through y*/
while openQueue is not empty
n: = the node in openQueue having the lowest
f[] value
if n = goal
return reconstruct_path (came_from, goal)
remove n from openQueue
add n to closedQueue
for each y in neighbor_nodes (n)
if y in closedQueue
continue
tentative_g: = g [n] + dist_between (n, y) +
risk_cost
/*Compares the risk cost*/
tentative_is_better:= false
if y not in openQueue
add y to openQueue
h[y]: = estimate_of_distance (y, goal)
tentative_is_better:= true
elseif tentative_g < g[y]
tentative_is_better: = true
if tentative_is_better = true
came_from[y]: = n
g[y]: = tentative_g
f[y]: = g[y] + h[y]
return failure
function reconstruct_path (came_from,current_node)
if came_from[current_node] is set p = reconstruct_path (came_from, came_
from[current_node])
return (p + current_node)
else
return the empty path

Listing 1. Pseudo-code for the path planning algorithm.

 Contour Based Path Planning with B-Spline Trajectory Generation ... 157

Function spline_path (start, goal)

locate_points()/*get coordinates (x, y) of middle
knots*/
m_points: = the vector list/* to store control
points*/
m_nodes: = the vector list/* to store small nodes*/
m_steps: = number of steps/*steps for spline*/
add_points (x, y)/*store (x, y) in a vector list*/
while (x, y) is not goal
new_controlpoint_added ()
/*spline curve based on the control points and
nodes*/
for int i = 0 to m_steps
compute blending function

for t ∈[0,1]
get_point (double u, const GLVector & P0,
const GLVector & P1, const GLVector & P2, const
GLVector & P3)
/*return coordinate (x, y) for particular t*/
add_node (const GLVector & node)/*add
small nodes in between the control points and we
make use of 100 small nodes in this work*/
end for
end while
render_spline()/*display generated curve*/
return the spline path

Listing 2. Pseudo-code for generating a spline path.

It is worth noticing that the interpolation by means of cubic B-splines
guarantees the C2 continuity of the geometric path. Owing to the continuity
of the curve derivatives, it is not possible to obtain a path with sharp corners
that is not flyable by the UAVs. An illustration of the generated solution is
shown in Figure 10.
158 Unmanned Aerial Vehicles (UAV) and Drones

CONCLUSIONS AND FUTURE WORKS

This paper presents a contour based path planner to generate an optimal
path for UAVs over hostile terrain. The proposed method is flexible given
that it allows the user to input a threshold altitude value to avoid radar
detection and terrain collisions. The key strengths of the method are: 1) the
ability to plot a safe path to the target while minimizing exposure to the
enemy radar; 2) the generated paths are smoothened using cubic splines
to improve navigability and 3) computational efficiency is achieved. Using
this planning approach, the ability to generate feasible paths has been
demonstrated through simulations. The information provided in this paper
is significant for the planning of an air reconnaissance mission. However the
procedure has limitations and is sensitive to some of the input parameters,
leaving room for improvements in future research based on computational
intelligence approaches [25-28].

Figure 10. Optimal path based on heuristic search algorithm and user-specified
threshold altitude.

ACKNOWLEDGEMENTS
The authors gratefully acknowledge the funding support from Temasek
Defence Systems Institute, Singapore.
 Contour Based Path Planning with B-Spline Trajectory Generation ... 159

REFERENCES
1. A. Agarwal, M. H. Lim, M. J. Er and T. N. Nguyen, “Rectilinear
Workspace Partitioning for Parallel Coverage Using Multiple UAVs,”
Advanced Robotics, Vol. 21, No. 1, 2007, pp. 105-120.
2. K. K. Lim, Y. S. Ong, M. H. Lim and A. Agarwal, “Hybrid Ant Colony
Algorithm for Path Planning in Sparse Graphs,” Soft Computing
Journal, Vol. 12, No. 10, 2008, pp. 981-994.
3. C. W. Yeu, M. H. Lim, G. Huang, A. Agarwal and Y. S. Ong, “A
New Machine Learning Paradigm for Terrain Reconstruction,” IEEE
Geoscience and Remote Sensing Letters, Vol. 3, No. 3, 2006, pp. 981-
994. doi:10.1109/LGRS.2006.873687
4. L. Davis, “Warp Speed: Path Planning for Star Trek: Armada,” AAAI
Spring Symposium, AAAI Press, Menlo Park, 2000.
5. E. Frazzoli, M. Dahleh and E. Feron, “Real-Time Motion Planning
for Agile Autonomous Vehicles,” Journal of Guidance, Control and
Dynamics, Vol. 25, No. 1, 2002, pp. 116-129. doi:10.2514/2.4856
6. N. H. Sleumer and N. Tschichold-Gürman, “Exact Cell Decomposition
of Arrangements Used for Path Planning in Robotics,” Technical
Reports 329, ETH Zürich, Institute of Theoretical Computer Science,
1999.
7. J. C. Latombe, “Robot Motion Planning,” Kulwer Academic Publishers,
Boston, 1991.
8. T. Lozano-Pyrez and M. A. Wesley, “An Algorithm for Planning
Collision-Free Paths among Polyhedral Obstacles,” Communications
of the ACM, Vo1. 22, No. 10, 1979, pp. 565-570.
9. M. Jun, “Path Planning for Unmanned Aerial Vehicles in Uncertain
and Adversarial Environments,” In: S. Butenko, R. Murphey and
P. Pardalos, Eds., Cooperative Control: Models, Applicarions and
Algorithms, Kluwer, 2003, pp. 95-111.
10. J. Hilgert, K. Hirsch, T. Bertram and M. Hiller, “Emergency Path
Planning for AutonomousVehicles Using Elastic Band Theory,”
Advanced Intelligent Mechatronics, Vol. 2, 2003, pp. 1390-1395.
11. T. Sattel and T. Brandt, “Ground Vehicle Guidance Along Collision-
Free Trajectories Using Elastic Bands,” Proceedings of 2005 American
160 Unmanned Aerial Vehicles (UAV) and Drones

Control Conference, Portla, 2005, pp. 4991-4999.

12. J. Hwang, R. C. Arkin and D. Kwon, “Mobile Robots at Your Fingertip:
Bezier Curve On-Line Trajectory Generation for Supervisory Control,”
Proceedings of Intelligent Robots and Systems (IROS 2003), Las
Vegas, Vol. 2, 2003, pp. 1444-1449.
13. J. Aleotti, S. Caselli and G. Maccherozzi, “Trajectory Reconstruction
with NURBS Curves for Robot Programming by Demonstration,”
Proceedings of Computational Intelligence in Robotics and Automation,
Barcelona, 2005, pp. 73-78.
14. K. B. Judd and T. W. McLain, “Spline Based Path Planning for
Unmanned Air Vehicles,” AIAA Guidance, Navigation, and Control
Conference and Exhibit, Montreal, 2001.
15. E. P. Anderson, R. W. Beard, and T. W. McLain, “RealTime Dynamic
Trajectory Smoothing for Unmanned Air Vehicles,” IEEE Transactions
on Control Systems Technology, Vol. 13, No. 3, 2005, pp. 471-477.
doi:10.1109/TCST.2004.839555
16. E. M. Kan, M. H. Lim, S. P. Yeo, S. H. Shao and J. S. Ho, “Radar Aware
Path Planning for UAVs,” IEEE Symposium on Intelligent Systems
and Applications, Trabzon, June 2009.
17. E. M. Kan, S. P. Yeo, C. S.Tan, S. H. Shao and J. S. Ho, “Stealth
Path Planning for UAVs in Hostile Radar Zones,” Proceedings of the
11th IASTED International Conference on Control and Applications,
Cambridge, July 2009, pp. 54-60.
18. F. W. Moore, “Radar Cross-Section Reduction via Route Planning and
Intelligent Control,” IEEE Transactions on Control Systems Technology,
Vol. 10, No. 5, 2002, pp. 696-700. doi:10.1109/TCST.2002.801879
19. U. F. Knott, “Radar Cross Section Measurements,” Van Nostrand
Reinhold, New York, 1993.
20. L. Sevgi, “Complex Electromagnetic Problems and Numerical
Simulation Approaches,” IEEE Press/John Wiley & Sons, New York,
2003.
21. E. F. Knott, J. F. Shaeffer, and M. T. Tuley, “Radar Cross Section,” 2nd
Edition, Artech House, Norwood, 1993, p. 231.
22. M. Pachter, D. R. Jacques and J. M. Hebert, “Minimizing Radar
Exposure in Air Vehicle Path Planning,” Proceedings of the 41st Israel
Annual Conference on Aerospace Sciences, Tel-Aviv, February 2001.
23. M. G. Cox, “The Numerical Evaluation of B-Splines,” IMA Journal of
 Contour Based Path Planning with B-Spline Trajectory Generation ... 161

Applied Mathematics, Vol. 10, No. 2, 1972, pp. 134-149. doi:10.1093/

imamat/10.2.134
24. R. H. Bartels, J. C. Beatty and B. A. Barsky, “An Introduction to
Splines for Use in Computer Graphics and Geometric Modeling,”
Morgan Kaufmann Publishers, Massachusetts, 1987.
25. Q. Cao, M. H. Lim, J. H. Li, Y. S. Ong and W. L. Ng, “A Context
Switchable Fuzzy Inference Chip,” IEEE Transactions on Fuzzy
Systems, Vol. 14, No. 4, 2006, pp. 552- 567. doi:10.1109/
TFUZZ.2006.876735
26. M. H. Lim and Y Takefuji, “Implementing Fuzzy RuleBased Systems
on Silicon Chips,” IEEE Expert, Vol. 5, No. 1, 1990, pp. 31-45.
doi:10.1109/64.50855
27. R. Meuth, M. H. Lim, Y. S. Ong and D. C. Wunsh, “A Proposition on
Memes and Meta-Memes in Computing for Higher-Order Learning,”
Memetic Computing, Vol. 1, No. 2, 2009, pp. 85-100. doi:10.1007/
s12293-009-0011-1
28. M. H. Lim, Q. Cao, J. H. Li and W. L. Ng, “Evolvable Hardware Using
Context Switchable Fuzzy Inference Processor,” IEEE Proceedings:
Computers and Digital Techniques, Vol. 151, No. 4, 2004, pp. 301-311.
doi:10.1049/ip-cdt:20040666
 CHAPTER

8
Trajectory Tracking of Quadrotor Aerial
Robot Using Improved Dynamic
Inversion Method

Lifeng Wang, Yichong He, Zhixiang Zhang, Congkui He

Field Bus Technology & Automation Lab, North China University of
Technology, Beijing, China

ABSTRACT
This paper presents trajectory tracking control works concerning quadrotor
aerial robot with rigid cross structure. The quadrotor consists of four
propellers which are two paired clockwise rotate and anticlockwise rotate.
A nonlinear dynamic model of the quadrotor is provided, and a controller
based on the improved dynamic inverse is synthesized for the purpose of
stabilization and trajectory tracking. The proposed control strategy has been
tested in simulation that can balance the deviation of model inaccuracy well.

Citation: L. Wang, Y. He, Z. Zhang and C. He, “Trajectory Tracking of Quadrotor

Aerial Robot Using Improved Dynamic Inversion Method,” Intelligent Control and
Automation, Vol. 4 No. 4, 2013, pp. 343-348. doi: 10.4236/ica.2013.44040.
Copyright: © 2013 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
164 Unmanned Aerial Vehicles (UAV) and Drones

Keywords: Aerial Robot; Quadrotor; Nonlinear Model; Dynamic Inversion;

Robust Control

INTRODUCTION
Since the beginning of 20th century, many research efforts have been done to
create effective flying machines with improved performance and capabilities.
The quadrotor aerial robot is emerging as a kind of popular platform for
unmanned aerial vehicle (UAV) research due to its simple structure,
hovering ability and vertical take-off and landing (VTOL) capability. It is
especially suitable for reconnaissance and surveillance missions in complex
and tough environment. Because of quadrotor’s complexity of dynamic and
uncertainty of parameters, it is usually chosen as a testbed for validating the
advanced control algorithm of nonlinear control.
The quadrotor is described as aero robot in [1,2]. The core problems
of quadrotor are attitude control and trajectory tracking control. The flight
control unit and its control programming code are priority to its structure
when treated as a testbed for control algorithm. With the emerging of open
code project for autonomous control, it’s more convenient to transplant the
advanced control algorithms from modeling and simulation to validating
test, which also provide great help to research enthusiast or aerial robot
hobbyist. In [3], the differential flat control and nonlinear inverse control
are proposed to process the nonlinear trajectory control for differential
flatness property system. In [4], an real-time nonlinear optimal control,
which is called technique, is employed to solve the resultant
challenging problem considering the full nonlinear dynamics without
gaining scheduling techniques and timescale separations, which leads to a
closedform suboptimal control law. In [5,6], the back-stepping procedure is
synthesized for the purposes of stabilization and trajectory tracking, which
is generally robustness applied to the system of parameters uncertainty and
unmodeled dynamic to some extent. The various types of aerial robot UAV
based on quadrotor concept are put forward in [5-8], using the superiority of
fix-wing plant and rotor-wing helicopter to enhance the VTOL ability and
flying range. So, in general, the dynamic inverse method is more suitable
for various types of aerial robots due to its start from original model. In
[9], the dynamic inverse control with zero dynamic for non-linear affine
model system is proposed. In [10], the hierarchical flight control is synthesis
for rotorcraft-based UAV. In present paper, the hierarchical structure of
improved dynamic inverse method for quadrotor trajectory tracking is
 Trajectory Tracking of Quadrotor Aerial Robot Using Improved ... 165

proposed, which can balance the deviation of model inaccuracy to enhance

the robustness of control system.
In this paper, we first describe the mechanism of quadrotor aerial robot in
section 2. Next, we establish the dynamic model of quadrotor obtained from
NewtonEuler formula in section 3. In secion 4, a hierarchical structure of
improved dynamic inversion method is proposed. Furthermore, robustness
with respect to parameters uncertainty and unmodeled dynamic has been
observed in simulation. Finally, a trajectory tracking example is simulated
and remarked.

QUADROTOR MECHANISM
The typical quadrotor mechanism and coordinate system is briefly showed
in Figure 1, where four rotors are cross structure, providing the three
required moments (roll, pith and yaw) and lifting force. For describing the
flight condition of quadrotor exactly, we can respectively define flat-earth
frame E and rigid body frame B according to right-hand rule.
The aerodynamic forces and moments of propeller depend on various
parameters, such as the angle of attack, aerofoil and inflow ratio. We
assumed that the aerodynamic forces Fi and moments Ti of four propellers
are proportional to the square of angular velocity . Which is given by:

(1)
where i =1, 2,3, 4 . The , kt, kf is respectively aerodynamic force and anti-
torque coefficient. Since the forces and moments are derived by modifying
the speed of each rotor, then the lifting force and three moments can be
given as follows:
Lifting force (the sum of four rotor forces):

(2)
Roll torque (right-hand rule):

(3)
Pitch torque (right –hand rule):
166 Unmanned Aerial Vehicles (UAV) and Drones

(4)
Yaw torque (right –hand rule):

Figure 1. Quadroter UAV coordinate frame systems.

(5)
where L is the distance from motor position to the center of quadrotor
mass. Rewrite the matrix form of Equations (2)-(5) as follows:

(6)

DYNAMIC MODEL OF QUADROTOR

The frame E stands for earth coordinate system, and the body frame B stands
for the rigid quadrotor coordinate system. The body coordinate B is attached
to the center of mass of the quadrotor. The Euler angles which denote the
rotation of quadrotor in the flat-earth frame coordinate E is the variables,
,which respecttively represents the roll, pitch and yaw angles
of the quadrotor in the body-fixed coordinates.
 Trajectory Tracking of Quadrotor Aerial Robot Using Improved ... 167

Six DOF Model of Rigid Body

The dynamics of a rigid body under external forces applied to the center of
mass expressed in the body fixed frame are in Newton-Euler formalism:

(7)

where the is the velocity in the body frame,

is the angular velocity in the body frame. denote
the external propeller forces acted on the rotors, the weight forces of mass center, the
elastic force of landing gear only on the ground, respectively. are
external torques in the body frame. are the inertia matrix of quadrotor, m is
the mass of quadrotor.
The 3 × 3 symmetrical inertia matrix I is defined as:

(8)
and the cross product operator × is defined as :

(9)
The external propeller forces Fb acted on the rotors is defined as:

(10)
The weight forces of mass center is as follows:

(11)
where s and c are shorthand form of cosine and sine.
The elastic force of landing gear on the ground is modeling the
characteristics of taking off situation for quadrotor, so it disappears while
flying or leaving the ground. It is modeling as follows:
168 Unmanned Aerial Vehicles (UAV) and Drones

(12)
where kl and zl is elastic coefficient and elastic deformation on ground.
And the function denotes:

(13)
Then the Equation (7) can be expanded as follows:

(14)

(15)

Navigation Equations
To obtain the translational and rotational motions from the body frame to
the flat earth coordinate system, the well-known navigation equations are
described as follows:

(16)

where the vector stand for the positions of quadrotor in flat

earth coordinate, the vector stand for the attitude angles of body
frame in the flat earth coordinate. The matrix R and S is translational and rotational
transform matrix. The Equation (16) can be extended as follows:
 Trajectory Tracking of Quadrotor Aerial Robot Using Improved ... 169

(17)

(18)
where s, c and t stand for sine, cosine and tangent, respectively.

Motor Dynamics
The revolving speed of propeller driven by the brushless DC-motor is
proportional to the input voltage which is regulated by speed control unit.
Input voltage signals come from the flight control system in the light of
flight task. Assumed that the load anti-torque is proportional to the square
of revolving speed and the inductance of brushless DC-motor is omitted, the
dynamic model of DC-motor is described in following formula:

(19)
where the coefficient, is constant in dynamic model, is the
regulated signal pwm from the flight control unit and the ith is the index of DC-motor.

IMPROVED DYNAMIC INVERSE CONTROLLER

The above airframe dynamics of typical quadrotor involves Equations (14),
(15), (17)-(19), including 6 equations in Equations (14) and (15) stand
for airframe rigid-body dynamics, 6 equations in Equations (17) and (18)
stand for navigation transformation, 4 equations in Equations (19) stand for
motor dynamics. In these equations describe the dynamics of quadrotor, the
total state variables include 16 state variables and 4 control input variables.
The inherent characteristics of strong coupling in pitch-roll-yaw and
underactuation are shown in state equations. For solving the coupling in
pitch-roll-yaw, the dynamic inverse controller is applied to the controller.
In parallel, the hierarchical structure of slow-loop and fast-loop is adopted
to solve the underactuation. Nevertheless, the accurate model which is
necessary in dynamic inverse method is hardly to acquire, so it’s important
to develop an improved dynamic inverse model to enhance the robustness
of controller. The overall controller depicted in Figure 2 has fast loop as
attitude control loop and slow loop as trajectory tracking loop. The details
of improving dynamic inverse of hierarchical structure will be described in
following process.
170 Unmanned Aerial Vehicles (UAV) and Drones

Slow Loop Control

The desired trajectory can be divided into sequence of segments with
waypoints. So quadrotor tracks from one waypoint to the next waypoint
with desired velocity and acceleration constraints, then the segments are
connected

Figure 2. Improved dynamic inverse controller of hierarchical structure for

quadrotor.
to approximate the continuous trajectory well. The slow loop frequency
which depends on the response of translational motion and position sensor
is about 4 Hz.
A given trajectory command signal include position command
and orientation command . We assumed that the pitch and
roll angle are small enough to combine the formula (14) and (18). Then,
symbol inv1 can be denoted by the relationship of attitude angle and position
derivative as follows:

(20)
Trajectory control is considered to exponentially stabilize the waypoint
position error. Then the desired acceleration command would satisfy the
relationship as follows:

(21)
 Trajectory Tracking of Quadrotor Aerial Robot Using Improved ... 171

As depicted in Figure 2, , is the damping factor and natural frequency

of system, respectively. The factor and
. From Equations (20) and (21) above, the control problem of quadrotor
is addressed by decoupling the position control and attitude control, and
position control of slow loop provides the attitude set points for the attitude
controller of fast loop. We have testified that the desired system dynamics
could be perfect when damping factor is constant 0.4 - 0.8. Only if the
natural frequency of system required to aware, the controller factor can be
quantified as follows:

(22)

Fast Loop Control

Fast loop is attitude control loop which is required to response the change of
attitude angle quickly. Its frequency which settled by system dynamics and
control unit is 50 Hz.
Equation (17) can be transformed as follows:

(23)
where the matrix inverse stands for symbol inv2. We can acquire
by expected attitude angle command .
Equation (15) can be transformed as follows:

(24)
172 Unmanned Aerial Vehicles (UAV) and Drones

In Equation (24), the second item is intentionally omitted in order to

maintain zero dynamics while angular velocity error is zero. Thus, the
model error or unmodel dynamic characteristics in dynamic inverse method
can be balanced through improving the dynamic inverse model. Though the
proposed model of dynamic inverse is unaccurate, the zero dynamics can be
kept even if the model is not valid. We can acquire by expected
attitude angle command .

In above formula, the factors and

can be quantified in the same process. For

example:

(25)

The factor and is gained in the same formula.

Motor Speed Control

Equation (6) can be transform as follows:

(26)
where the square root stands for symbol inv3.

SIMULATION RESULTS
In this section, the demonstrating results are presented in order to observe
the performances of proposed control principle. We considered a case of a
rectangle shape trajectory-tracking problem. The parameters used for the
aircraft model and the controller are given in Table 1. These parameters are
referred to the quadrotor prototype designed by our lab, just refer Figure 3.
 Trajectory Tracking of Quadrotor Aerial Robot Using Improved ... 173

Table 1. Model parameters in demonstrator of quadrotor.

In the improving dynamic inverse controller, the parameters are defined

as follows: The natural frequency of slow loop in Equation (21) is guessed
as 1 Hz, the damping ratio is chosen as 0.72 according the experience in
simulation. So the proportional parameters in controller are defined in
Equation (22). In the same way, the natural frequency of pitch and roll
attitude in Equation (25) is guessed as 10 Hz and the damping ratio is chosen
as 0.7. But the natural frequency of yaw attitude is chosen as 2 Hz and
damping ratio is chosen as 0.7.
The rectangle shape trajectory is 10 m × 15 m field with an altitude
about 10 m altitude, the tracking simulation task is described as taking off
,tracking the rectangle shape trajectory, then landing. The results shown
in Figure 4 illustrate the trajectory tracking performance of control strategy
developed in this paper. The modeling difference between plant model and
dynamic inverse controller include two aspects: no motor dynamic model
in controller and 10% deviation of quadrotor body inertia and mass. But we
can see the proposed controller behaves perfectly in Figure 4, except for the
altitude due to quadrotor mass error.

CONCLUSION
The analysis presented here has shown the hover and
174 Unmanned Aerial Vehicles (UAV) and Drones

Figure 3. The demonstrator of simulating quadrotor.

Figure 4. Illustrate the trajectory tracking performance of improved dynamic

inverse strategy.
Lifeng Wang, Yichong He, Zhixiang Zhang, Congkui He trajectory
tracking capability of proposed controller. A nonlinear dynamic model of
the quadrotor is provided, and a controller based on the improved dynamic
inverse is synthesized for the purpose of stabilization and trajectory tracking.
The proposed control strategy can balance the deviation of model inaccuracy.
 Trajectory Tracking of Quadrotor Aerial Robot Using Improved ... 175

REFERENCES
1. T. Tomic, K. Schmid, P. Lutz, et al., “Towards a Fully Autonomous
UAV,” IEEE Robotics & Automation Magazine, Vol. 19, No. 3, 2012,
pp. 46-56.
2. H. Lim, J. Park, D. Lee and H. J. Kim, “Build Your Own Quadrotor:
Open-Source Projects on Unmanned Aerial Vehicles,” IEEE Robotics
& Automation Magazine, Vol. 19, No. 3, 2012, pp. 33-45.
3. A. Drouin, S. Simoes-Cunha, and A. C. Brandao-Ramos, “Differential
Flatness and Control of Nonlinear Systems,” Proceedings of the 30th
Chinese Control Conference, 22-24 July 2011, Yantai, pp. 643-648.   
4. M. Xin, Y. J. Xu and R. Hopkins, “Trajectory Control of Miniature
Helicopters Using a Unified Nonlinear Optimal Control Technique,”
Journal of Dynamic Systems, Measurement and Control, Vol. 133, No.
6, 2011, 14 p.   
5. K. T. Oner, E. Cetinsoy, M. Unel, et al., “Dynamic Model and Control of a
New Quadrotor Unmanned Aerial vehicle with Tilt-Wing Mechanism,”
World Academy of Science, Engineering and Technology, 2008.
6. Z. Fang and W. N. Gao, “Adaptive Integral Backstepping Control of
a Micro-Quadrotor,” The 2nd International Conference on Intelligent
Control and Information Processing, Harbin, 25-28 July 2011, pp. 910-
915.
7. I. F. Kendoul and R. Lozano, “Modeling and Control of a Small
Autonomous Aircraft Having Two Tilting Rotors,” IEEE Transactions
on Robotics, Vol. 22, No. 6, 2006, pp. 1297-1302.
8. K. T. Oner, E. Cetinsoy, E. Sirimoglu, et al., “Mathematical Modeling
and Vertical Flight Control of a Tilt-Wing UAV,” Turkish Journal of
Electrical Engineering & Computer Sciences, Vol. 20, No. 1, 2012, p.
149.
9. A. Das, K. Subbarao and F. Lewis, “Dynamic Inversion with Zero-
Dynamics Stabilisation for Quadrotor Control,” IET Control Theory
Applications, Vol. 3, No. 3, 2009, pp. 303-314. http://dx.doi.
org/10.1049/iet-cta:20080002   
10. H. Shim, “Hierarchical Flight Control System Synthesis for Rotorcraft-
Based Unmanned Aerial Vehicles,” Doctor of Philosophy Dissertation,
University Of California, Berkeley, 2000.   
 SECTION 3
QUADROTORS AND
SMALL-SIZE AERIAL
VEHICLES
 CHAPTER

9
Neural Network Control and Wireless
Sensor Network-based Localization of
Quadrotor UAV Formations

Travis Dierks and S. Jagannathan

Missouri University of Science and Technology, United States of America

INTRODUCTION
In recent years, quadrotor helicopters have become a popular unmanned
aerial vehicle (UAV) platform, and their control has been undertaken by
many researchers (Dierks & Jagannathan, 2008). However, a team of UAV’s
working together is often more effective than a single UAV in scenarios
like surveillance, search and rescue, and perimeter security. Therefore, the
formation control of UAV’s has been proposed in the literature.

Citation: Travis Dierks and S. Jagannathan (January 1st 2009). “Neural Network Con-
trol and Wireless Sensor Network-based Localization of Quadrotor UAV Formations”,
Aerial Vehicles, Thanh Mung Lam, IntechOpen, DOI: 10.5772/6477.
Copyright: © 2009 The Author(s). Licensee IntechOpen. This chapter is distributed
under the terms of the Creative Commons Attribution-NonCommercial-ShareAlike-3.0
License
180 Unmanned Aerial Vehicles (UAV) and Drones

Saffarian and Fahimi present a modified leader-follower framework and

propose a model predictive nonlinear control algorithm to achieve the
formation (Saffarian & Fahimi, 2008). Although the approach is verified via
numerical simulations, proof of convergence and stability is not provided.
In the work of Fierro et al., cylindrical coordinates and contributions from
wheeled mobile robot formation control (Desai et al., 1998) are considered
in the development of a leader-follower based formation control scheme for
aircrafts whereas the complete dynamics are assumed to be known (Fierro et
al., 2001). The work by Gu et al. proposes a solution to the leader-follower
formation control problem involving a linear inner loop and nonlinear
outer-loop control structure, and experimental results are provided (Gu et
al., 2006). The associated drawbacks are the need for a dynamic model and
the measured position and velocity of the leader has to be communicated to
its followers. Xie et al. present two nonlinear robust formation controllers
for UAV’s where the UAV’s are assumed to be flying at a constant altitude.
The first approach assumes that the velocities and accelerations of the leader
UAV are known while the second approach relaxes this assumption (Xie et
al., 2005). In both the designs, the dynamics of the UAV’s are assumed to
be available. Then, Galzi and Shtessel propose a robust formation controller
based on higher order sliding mode controllers in the presence of bounded
disturbances (Galzi & Shtessel, 2006).
In this work, we propose a new leader-follower formation control
framework for quadrotor UAV’s based on spherical coordinates where the
desired position of a follower UAV is specified using a desired separation, sd,
and a desired- angle of incidence,α d and bearing, βd . Then, a new control law
for leader-follower formation control is derived using neural networks (NN)
to learn the complete dynamics of the UAV online, including unmodeled
dynamics like aerodynamic friction in the presence of bounded disturbances.
Although a quadrotor UAV is underactuated, a novel NN virtual control input
scheme for leader follower formation control is proposed which allows all
six degrees of freedom of the UAV to be controlled using only four control
inputs. Finally, we extend a graph theory-based scheme for discovery,
localization and cooperative control. Discovery allows the UAV’s to form
into an ad hoc mobile sensor network whereas localization allows each UAV
to estimate its position and orientation relative to its neighbors and hence the
formation shape. This chapter is organized as follows. First, in Section 2,
the leader-follower formation control problem for UAV’s is introduced, and
required background information is presented. Then, the NN control law is
developed for the follower UAV’s as well as the formation leader, and the
 Neural Network Control and Wireless Sensor Network-based ... 181

stability of the overall formation is presented in Section 3. In Section 4, the

localization and routing scheme is introduced for UAV formation control
while Section 5 presents numerical simulations, and Section 6 provides
some concluding remarks.

BACKGROUND

Quadrotor UAV Dynamics

Consider a quadrotor UAV with six DOF defined in the inertial coordinate
frame , Ea , as where are
the position coordinates of the UAV and describe its
orientation referred to as roll, pitch, and yaw, respectively. The translational
and angular velocities are expressed in the body fixed frame attached to the
center of mass of the UAV, E b , and the dynamics of the UAV in the body
fixed frame can be written as (Dierks & Jagannathan, 2008)

(1)
where

and m is a positive scalar that represents the total mass of the UAV, J ∈ℜ3x3
represents the positive definite inertia matrix,
represents the translational velocity, represents
the angular velocity, , are the nonlinear aerodynamic
effects, provides the thrust along the z-direction, provides
the rotational torques, and , represents
unknown, but bounded disturbances such that for all time t , with τ
M
being a known positive constant, is an nxn identity matrix, and
represents an mxl matrix of all zeros. Furthermore,
represents the gravity vector defined as where
is a unit vector in the inertial coordinate frame, , and
is the general form of a skew symmetric matrix defined as in (Dierks &
Jagannathan, 2008). It is important to highlight for any vector
, and this property is commonly referred to as the skew symmetric
property (Lewis et al., 1999).
182 Unmanned Aerial Vehicles (UAV) and Drones

The matrix is the translational rotation matrix which is used

to relate a vector in the body fixed frame to the inertial coordinate frame
defined as (Dierks & Jagannathan, 2008)

(2)
where the abbreviations have been used for sin(•) and cos(•), re-
spectively. It is important to note that
. It is also necessary to define a rotational transformation matrix from the
fixed body to the inertial coordinate frame as (Dierks & Jagannathan, 2008)

(3)
where the abbreviation t (•) has been used for tan(•). The transformation
matrices R and T are nonsingular as long as
and −π ≤ψ ≤ π . These regions will be assumed throughout the development of
this work, and will be referred to as the stable operation regions of the UAV.
Under these flight conditions, it is observed that
for known constants Rmax andTmax (Neff et al., 2007). Finally, the kinematics
of the UAV can be written as

(4)

Neural Networks
In this work, two-layer NN’s are considered consisting of one layer of
randomly assigned constant weights in the first layer and one layer
of tunable weights in the second with a inputs, b outputs, and L
hidden neurons. A compromise is made here between tuning the number of
layered weights with computational complexity. The universal approximation
property for NN’s (Lewis et al., 1999) states that for any smooth function
, there exists a NN such that where εN
is the bounded NN functional approximation error such that ,for
 Neural Network Control and Wireless Sensor Network-based ... 183

a known constant εM and a is the activation function in

the hidden layers. It has been shown that by randomly selecting the input
layer weights VN , the activation function forms a stochastic
basis, and thus the approximation property holds for all inputs, a N x ℜ∈
, in the compact set S . The sigmoid activation function is considered
here. Furthermore, on any compact subset of ℜn , the target NN weights
are bounded by a known positive value, WM , such that . For
complete details of the NN and its properties, see (Lewis et al., 1999).

Three Dimensional Leader-Follower Formation Control

Throughout the development, the follower UAV’s will be denoted with a
subscript ‘j’ while the formation leader will be denoted by the subscript
‘i’. To begin the development, an alternate reference frame is defined by
rotating the inertial coordinate frame about the z-axis by the yaw angle of
follower j,ψ j , and denoted by a Ej . In order to relate a vector in a E to Eaj ,
the transformation matrix is given by

(5)
where .

Figure 1. UAV leader-follower formation control

The objective of the proposed leader-follower formation control
approach is for the follower UAV to maintain a desired separation, s jid , at a
desired angle of incidence, α jid ∈ Eaj , and bearing, β jid ∈ Eaj , with respect to
its leader. The incidence angle is measured from the plane of follower
j while the bearing angle is measured from the positive xaj - axis as shown
184 Unmanned Aerial Vehicles (UAV) and Drones

in Figure 1. It is important to observe that each quantity is defined relative

to the follower j instead of the leader i (Fierro et al., 2001), (Desai et al.,
1998). Additionally, in order to specify a unique configuration of follower j
with respect to its leader, the desired yaw of follower j is selected to be the
yaw angle of leader i, ψ i ∈ E a as in (Saffarian & Fahimi, 2008). Using this
approach, the measured separation between follower j and leader i is written
as

(6)
where

(7)
Thus, to solve the leader-follower formation control problem in the
proposed framework, a control velocity must be derived to ensure

(8)
Throughout the development, will be taken as constants,
while the constant total mass,mj , is assumed to be known. Additionally,
it will be assumed that reliable communication between the leader and its
followers is available, and the leader communicates its measured orientation,
Θi , and its desired states, . This is a far less stringent
assumption than assuming the leader communicates all of its measured
states to its followers (Gu et al., 2006). Additionally, future work will relax
this assumption. In the following section, contributions from single UAV
control will be considered and extended to the leader-follower formation
control of UAV’s.

LEADER-FOLLOWER FORMATION TRACKING

CONTROL
In single UAV control literature, the overall control objective UAV j is often
to track a desired trajectory, , and a desired yaw ψ
jd
while maintaining a stable flight configuration (Dierks & Jagannathan,
 Neural Network Control and Wireless Sensor Network-based ... 185

2008). The velocity vjzb is directly controllable with the thrust input.
However, in order to control the translational velocities vjxb and vjyb , the
pitch and roll must be controlled, respectively, thus redirecting the thrust.
With these objectives in mind, the frameworks for single UAV control are
extended to UAV formation control as follows.

Follower UAV Control Law

Given a leader i subject to the dynamics and kinematics (1) and (4),
respectively, define a reference trajectory at a desired separation s jid , at a
desired angle of incidence, α jid , and bearing, β jid for follower j given by

(9)
where Rajd is defined as in (5) and written in terms of ψ jd , and Ξ jid is
written in terms of the desired angle of incidence and bearing, α jid β jid ,
,respectively, similarly to (7). Next, using (6) and (9), define the position
tracking error as
(10)
which can be measured using local sensor information. To form
the position tracking error dynamics, it is convenient to rewrite (10) as
revealing

(11)
Next, select the desired translational velocity of follower j to stabilize (11)

(12)
where is a diagonal positive definite
design matrix of positive design constants and v id is the desired translational
velocity of leader i. Next, the translational velocity tracking error system is
defined as

(13)
Applying (12) to (11) while observing and similarly
, reveals the closed loop position error dynamics to be
rewritten as
186 Unmanned Aerial Vehicles (UAV) and Drones

(14)
Next, the translational velocity tracking error dynamics are developed.
Differentiating (13), observing

substituting the translational velocity dynamics in (1), and adding and

subtracting reveals

(15)
Next, we rewrite (2) in terms of the scaled desired orientation vector,
where , and
and are the maximum desired roll and
pitch, respectively, define , and add and subtract
with yield

(16)
where and

(17)
is an unknown function which can be rewritten as
. In the forthcoming development, the
approximation properties of NN will be utilized to estimate the unknown
function by bounded ideal weights , such that
for an unknown constant WMc1 , and written as where
is the bounded NN approximation error where ε Mc1 is a known
constant. The NN estimate of jc1 f is written as
where is the NN estimate
of T Wjc1 , , 3,2,1 ˆW 1 i = T ijc is the ith row of is the NN
input defined as
 Neural Network Control and Wireless Sensor Network-based ... 187

Note that is an estimate of x jc1 since the follower does not know ωi .
However, Θi is directly related toωi ; therefore, it is included instead.
Remark 1: In the development of (16), the scaled desired orientation
vector was utilized as a design tool to specify the desired pitch and roll
angles. If the un-scaled desired orientation vector was used instead, the
maximum desired pitch and roll would remain within the stable operating
regions. However, it is desirable to saturate the desired pitch and roll before
they reach the boundaries of the stable operating region.
Next, the virtual control inputs θ jd andφ jd are identified to control
the translational velocities v jxb and v jyb , respectively. The key step in the
development is identifying the desired closed loop velocity tracking error
dynamics. For convenience, the desired translational velocity closed loop
system is selected as

(18)
where is a diagonal positive
definite design matrix with each . In the
following development, it will be shown that
; therefore, it is clear that Kv > 0 . Then, equating (16) and (18) while
considering only the first two velocity error states reveals

(19)
where was utilized. Then, applying basic math
operations, the first line of (19) can be solved for the desired pitch θ jd while
the second line reveals the desired roll φ jd . Using the NN estimates, , The
desired pitch θ jd can be written as

(20)

where and
. Similarly, the desired roll angle, φ jd , is found to be

(21)
188 Unmanned Aerial Vehicles (UAV) and Drones

where and
.
Remark 2: The expressions for the desired pitch and roll in (20) and
(21) lend themselves very well to the control of a quadrotor UAV. The
expressions will always produce desired values in the stable operation
regions of the UAV. It is observed that a tan(•) approaches ± π / 2 as its
argument increases. Thus, introducing the scaling factors in
results in , and the aggressiveness of
the UAV’s maneuvers can be managed.
Now that the desired orientation has been found, next define the attitude
tracking error as

(22)
where the dynamics are found using (4) to be . In order
to drive the orientation errors (22) to zero, the desired angular velocity, ω jd
, is selected as

(23)

where is a diagonal positive

definite design matrix all with positive design constants. Define the angular
velocity tracking error as

(24)
and observing , the closed loop orientation tracking error
system can be written as

(25)
Examining (23), calculation of the desired angular velocity requires
knowledge of ; however, is not known in view of the fact
are not available. Further, development of uj2 in the following section will
reveal is required which in turn implies must be known. Since
these requirements are not practical, the universal approximation property
of NN is invoked to estimate (Dierks and Jagannathan, 2008).
 Neural Network Control and Wireless Sensor Network-based ... 189

To aid in the NN virtual control development, the desired orientation,

, is reconsidered in the fixed body frame, Eb , using the relation
. Rearranging (23), the dynamics of the proposed virtual
controller when the all dynamics are known are revealed to be

(26)
For convenience, we define a change of variable as ,
and the dynamics (26) become

(28)
Defining the estimates of , respectively,
and the estimation error , the dynamics of the proposed NN
virtual control inputs become

(28)
where K jΩ1
and K jΩ2
are positive constants. The estimate is then
written as

(29)
where K jΩ3 is a positive constant.
In (28), universal approximation property of NN has been utilized
to estimate the unknown function by bounded ideal weights
, such that for a known constant WMΩ , and written as
where ε jΩ is the bounded NN approximation
error such that for a known constant ε ΩM . The NN estimate of
f jΩ is written as where is the NN estimate of
is the NN input written in terms of the virtual control estimates,
desired trajectory, and the UAV velocity. The NN input is chosen to take the
form of .
190 Unmanned Aerial Vehicles (UAV) and Drones

Observing , subtracting (28) from (27)

and adding and subtracting , the virtual controller estimation error
dynamics are found to be

(30)

where , and
. Furthermore, with a computable
constant with NΩ the constant number of hidden layer neurons in the virtual
control NN. Similarly, the estimation error dynamics of (29) are found to be

(31)
where . Examination of (30) and (31)
reveals , and to be equilibrium points of the estimation error
dynamics when .
To this point, the desired translational velocity for follower j has been
identified to ensure the leader-follower objective (8) is achieved. Then, the
desired pitch and roll were derived to drive ,
respectively. Then, the desired angular velocity was found to ensure
. What remains is to identify the UAV thrust to guarantee and
rotational torque vector to ensure . First, the thrust is derived.
Consider again the translational velocity tracking error dynamics (16),
as well as the desired velocity tracking error dynamics (18). Equating (16)
and (18) and manipulating the third error state, the required thrust is found
to be

(32)

where is the NN estimate in (17) previously defined. Substituting

the desired pitch (20), roll (21), and the thrust (32) into the translational
velocity tracking error dynamics (16) yields
 Neural Network Control and Wireless Sensor Network-based ... 191

and adding and subtracting reveals

(33)

with
, for a known constant
, and for a computable constant
.
Next, the rotational torque vector, , will be addressed. First, multiply
the angular velocity tracking error (24) by the inertial matrix J j , take the first
derivative with respect to time, and substitute the UAV dynamics (1) to reveal

(34)

with . Examining
, it is clear that the function is nonlinear and contains unknown
terms; therefore, the universal approximation property of NN is utilized
to estimate the function by bounded ideal weights

, such that for a known constant WMc2 and written as

where is the bounded NN functional

reconstruction error such that for a known constant . The

NN estimate of is given by where ˆ is

the NN estimate of and is the input to the NN
written in terms of the virtual controller estimates. By the construction of
the virtual controller, is not directly available; therefore, observing (29),
the terms have been included instead.

Using the NN estimate and the estimated desired angular velocity

tracking error , the rotational torque control input is written
as
192 Unmanned Aerial Vehicles (UAV) and Drones

(35)
and substituting the control input (35) into the angular velocity dynamics
(34) as well as adding and subtracting , the closed loop dynamics
become

(36)

where , and

. Further, for a computable constant

where Nc2 is the number of hidden layer

neurons.

As a final step, we define and

so that a single NN can be utilized with Nc hidden

layer neurons to represent 6 In the following

theorem, the stability of the follower j is shown while considering eiv = 0 .
In other words, the position, orientation, and velocity tracking errors are
considered along with the estimation errors of the virtual controller and the
NN weight estimation errors of each NN for follower j while ignoring the
interconnection errors between the leader and its followers. This assumption
will be relaxed in the following section.
Theorem 3.1.1: (Follower UAV System Stability) Given the dynamic
system of follower j in the form of (1), let the desired translational velocity
for follower j to track be defined by (12) with the desired pitch and roll
defined by (20) and (21), respectively. Let the NN virtual controller be
defined by (28) and (29), respectively, with the NN update law given by

(37)

where are design parameters. Let

the dynamic NN controller for follower j be defined by (32) and (35),
respectively, with the NN update given by
 Neural Network Control and Wireless Sensor Network-based ... 193

(38)

where are
constant design parameters. Then there exists positive design
constants and positive definite design matrices
, such that the virtual controller estimation errors

and the virtual control NN weight estimation errors, , the

position, orientation, and translational and angular velocity tracking errors,
, respectively, and the dynamic controller NN weight
estimation errors, , are all SGUUB. Proof: Consider the following
positive definite Lyapunov candidate

(39)
where

whose first derivative with respect to time is given by

. Considering first , and substituting the closed loop virtual control
estimation error dynamics (30) and (31) as well as the NN tuning law (37)
, reveals

where and
provided and . Observing

for a known constant, , and

can then be rewritten

194 Unmanned Aerial Vehicles (UAV) and Drones

as

Now, completing the squares with respect to

, an upper bound for is found to be

(40)

where . Next, considering and

substituting the closed loop kinematics (14) and (25), dynamics (33) and
(36), and NN tuning law (38) while considering reveals

Then, observing and completing the squares with respect

to and ~ , and upper bound for is found to be

(41)

where ,and K jωmin

are the
minimum singular values of , and K jω
, respectively, and
. Now, combining (40) and (41),
an upper bound for is written as

(42)
Finally, (42) is less than zero provided

(43)
and the following inequalities hold:
 Neural Network Control and Wireless Sensor Network-based ... 195

(44)
Therefore, it can be concluded using standard extensions of Lyapunov
theory (Lewis et al., 1999) that is less than zero outside of a compact
set, revealing the virtual controller estimation errors, , and the NN
weight estimation errors, , the position, orientation, and translational and
angular velocity tracking errors, , respectively, and
the dynamic controller NN weight estimation errors, , are all SGUUB.

Formation Leader Control Law

The dynamics and kinematics for the formation leader are defined similarly
to (1) and (4), respectively. In our previous work (Dierks and Jagannathan,
2008), an output feedback control law for a single quadrotor UAV was
designed to ensure the robot tracks a desired path, , and
desired yaw angle,ψ id . Using a similarly approach to (10)- (14), the state
feedback control velocity for leader i is given by (Dierks and Jagannathan,
2008)

(45)
The closed loop position tracking error then takes the form of

(46)
Then, using steps similar to (15)-(21), the desired pitch and roll angles
are given by

(47)

where and
196 Unmanned Aerial Vehicles (UAV) and Drones

(48)

where

and and

is a NN estimate of the unknown function

(Dierks and Jagannathan, 2008). The desired angular velocity as
well as the NN virtual controller for the formation leader is defined similarly
to (23) and (28) and (29), respectively, and finally, the thrust and rotation
torque vector are found to be

(49)

(50)

where is a NN estimate of the unknown function

. The closed loop orientation, virtual control,
and velocity tracking error dynamics for the formation leader are found to
take a form similar to (25), (30) and (31), and (33) and (36), respectively
(Dierks and Jagannathan, 2008).
Next, the stability of the entire formation is considered in the following
theorem while considering the interconnection errors between the leader
and its followers.

Quadrotor UAV Formation Stability

Before proceeding, it is convenient to define the following augmented error
systems consisting of the position and translational velocity tracking errors
of leader i and N follower UAV’s as
 Neural Network Control and Wireless Sensor Network-based ... 197

Next, the transformation matrix (2) is augmented as

(51)
while the NN weights for the translational velocity error system are
augmented as

Now, using the augmented variables above, the augmented closed loop
position and translational velocity error dynamics for the entire formation
are written as

(52)

(53)

where defined similarly to Ajc in

terms of , , GF is a
constant matrix relating to the formation interconnection errors defined as

(54)

and is constant and dependent on the specific formation

topology. For instance, in a string formation where each follower follows
198 Unmanned Aerial Vehicles (UAV) and Drones

the UAV directly in front of it, follower 1 tracks leader i, follower 2 tracks
follower 1, etc., and FT becomes the identity matrix.
In a similar manner, we define augmented error systems for the virtual
controller, orientation, and angular velocity tracking systems as

(55)
respectively. It is straight forward to verify that the error dynamics of
the augmented variables (55) takes the form of (30), (31), (25), and (36),
respectively, but written in terms of the augmented variables (55).
Theorem 3.3.1: (UAV Formation Stability) Given the leader-follower
criterion of (8) with 1 leader and N followers, let the hypotheses of Theorem
3.1.1 hold. Let the virtual control system for the leader i be defined similarly
to (28) and (29) with the virtual control NN update law defined similarly to
(37). Let control velocity and desire pitch and roll for the leader be given
by (45), (47), and (48), respectively, along with the thrust and rotation
torque vector defined by (49) and (50), respectively, and let the control NN
update law be defined identically to (38). Then, the position, orientation, and
velocity tracking errors, the virtual control estimation errors, and the NN
weights for each NN for the entire formation are all SGUUB.
Proof: Consider the following positive definite Lyapunov candidate

(56)
where

and

. The first derivative of (56) with respect to

time is given by and performing similar steps as those used in
(40)-(41) reveals

(57)
 Neural Network Control and Wireless Sensor Network-based ... 199

(58)

where is the minimum singular value of

is the minimum singular value of

with I being the identity matrix, KΩ min3 is the minimum singular value

of , NΩ is the number of
hidden layer neurons in the augmented virtual control system, and η Ω
is a computable constant based on . Similarly,
are the minimum singular values
of the augmented gain matrices respectively, where
are a known computable constants
and ηc is a computable constant dependent on . Now,
using (57) and (58), an upper bound for is found to be

(59)
Finally, (59) is less than zero provided

(60)
and the following inequalities hold:

(61)
Therefore, it can be concluded using standard extensions of Lyapunov
theory (Lewis et al., 1999) that V$ is less than zero outside of a compact set,
revealing the position, orientation, and velocity tracking errors, the virtual
200 Unmanned Aerial Vehicles (UAV) and Drones

control estimation errors, and the NN weights for each NN for the entire
formation are all SGUUB.
Remark 3: The conclusions of Theorem 3.3.1 are independent of any
specific formation topology, and the Lyapunov candidate (56) represents
the most general form required show the stability of the entire formation.
Examining (60) and (61), the minimum controller gains and error bounds are
observed to increase with the number of follower UAV’s, N. These results
are not surprising since increasing number of UAV’s increases the sources
of errors which can be propagated throughout the formation.
Remark 4: Once a specific formation as been decided and the form of
FT is set, the results of Theorem 3.3.1 can be reformulated more precisely.
For this case, the stability of the formation is proven using the sum of the
individual Lyapunov candidates of each UAV as opposed to using the
augmented error systems (51)-(55).

OPTIMIZED ENERGY-DELAY SUB-NETWORK

ROUTING (OEDSR) PROTOCOL FOR UAV LOCALI-
ZATION AND DISCOVERY
In the previous section, a group of UAV’s was modeled as a nonlinear
interconnected system. We have shown that the basic formation is stable
and each follower achieves its separation, angle of incidence, and bearing
relative to its leader with a bounded error. The controller assignments for the
UAV’s can be represented as a graph where a directed edge from the leader to
the followers denotes a controller for the followers while the leader is trying
to track a desired trajectory. The shape vector consists of separations and
orientations which in turn determines the relative positions of the follower
UAV’s with respect to its leader.
Then, a group of N UAV’s is built on two networks: a physical network
that captures the constraints on the dynamics of the lead UAV and a sensing
and communication network, preferably wireless, that describes information
flow, sensing and computational aspects across the group. The design of the
graph is based on the task in hand. In this graph, nodes and edges represent
UAV’s and control policies, respectively. Any such graph can be described
by its adjacency matrix (Das et al., 2002).
In order to solve the leader-follower criterion (8), ad hoc networks
are formed between the leader(s) and the follower(s), and the position of
each UAV in the formation must be determined on-line. This network is
 Neural Network Control and Wireless Sensor Network-based ... 201

dependent upon the sensing and communication aspects. As a first step, a

leader is elected similar to the case of multi-robot formations (Das et al.,
2002) followed by the discovery process in which the sensory information
and physical networks are used to establish a wireless network. The outcome
of the leader election process must be communicated to the followers in
order to construct an appropriate shape. To complete the leader-follower
formation control task (8), the controllers developed in the previous section
require only a single-hop protocol; however, a multi-hop architecture is
useful to relay information throughout the entire formation like the outcome
of the leader election process, changing tasks, changing formations, as well
alerting the UAV’s of approaching moving obstacles that appear in a sudden
manner.
The optimal energy-delay sub-network routing (OEDSR) protocol
(Jagannathan, 2007) allows the UAV’s to communicate information
throughout the formation wirelessly using a multi-hop manner where each
UAV in the formation is treated as a hop. The energy-delay routing protocol
can guarantee information transfer while minimizing energy and delay for
real-time control purposes even for mobile ad hoc networks such as the case
of UAV formation flying.
We envision four steps to establish the wireless ad hoc network. As
mentioned earlier, leader election process is the first step. The discovery
process is used as the second step where sensory information and the
physical network are used to establish a spanning tree. Since this is a multi-
hop routing protocol, the communication network is created ondemand
unlike in the literature where a spanning tree is utilized. This on-demand
nature would allow the UAV’s to be silent when they are not being used
in communication and generate a communication path when required. The
silent aspect will reduce any inference to others. Once a formation becomes
stable, then a tree can be constructed until the shape changes. The third
step will be assignment of the controllers online to each UAV based on
the location of the UAV. Using the wireless network, localization is used
to combine local sensory information along with information obtained
via routing from other UAV’s in order to calculate relative positions and
orientations. Alternatively, range sensors provide relative separations,
angles of incidence, and bearings. Finally cooperative control allows the
graph obtained from the network to be refined. Using this graph theoretic
formulation, a group is modeled by a tuple , where ϑ is the
reference trajectory of the robot, Ρ represents the shape vectors describing
202 Unmanned Aerial Vehicles (UAV) and Drones

the relative positions of each vehicle with respect to the formation reference
frame (leader), and H is the control policy represented as a graph where
nodes represent UAV and edges represent the control assignments. Next, we
describe the OEDSR routing protocol where each UAV will be referred to
as a “node.”
In OEDSR, sub-networks are formed around a group of nodes due to an
activity, and nodes wake up in the sub-networks while the nodes elsewhere
in the network are in sleep mode. An appropriate percentage of nodes in the
sub-network are elected as cluster heads (CHs) based on a metric composed
of available energy and relative location to an event (Jagannathan, 2007) in
each sub-network. Once the CHs are identified and the nodes are clustered
relative to the distance from the CHs, the routing towards the formation leader
(FL) is initiated. First, the CH checks if the FL is within the communication
range. In such case, the data is sent directly to the FL. Otherwise, the data
from the CHs in the sub-network are sent over a multi-hop route to the
FL. The proposed routing algorithm is fully distributed since it requires
only local information for constructing routes, and is proactive adapting to
changes in the network. The FL is assumed to have sufficient power supply,
allowing a high power beacon from the FL to be sent such that all the nodes
in the network have knowledge of the distance to the FL. It is assumed that
all UAV’s in the network can calculate or measure the relative distance to
the FL at any time instant using the formation graph information or local
sensory data. Though the OEDSR protocol borrows the idea of an energy-
delay metric from OEDR (Jagannathan, 2007), selection of relay nodes
(RN) does not maximize the number of two hop neighbors. Here, any UAV
can be selected as a RN, and the selection of a relay node is set to maximize
the link cost factor which includes distance from the FL to the RN.

Selection of an Optimum Relay-Node-Based link cost factor

Knowing the distance information at each node will allow the UAV to
calculate the Link Cost Factor (LCF). The link cost factor from a given node
to the next hop node ‘k’ is given by (62) where Dk represent the delay that
will be incurred to reach the next hop node in range, the distance between
the next hop node to the FL is denoted by k Δxk , and the remaining energy,
Ek , at the next hop node are used in calculation of the link cost as

(62)
 Neural Network Control and Wireless Sensor Network-based ... 203

In equation (62), checking the remaining energy at the next hop node increases
network lifetime; the distance to the FL from the next hop node reduces the
number of hops and endto-end delay; and the delay incurred to reach the
next hop node minimizes any channel problems. When multiple RNs are
available for routing of the information, the optimal RN is selected based on
the highest LCF. These clearly show that the proposed OEDSR protocol is
an on demand routing protocol. For detailed discussion of OEDSR refer to
(Jagannathan, 2007). The route selection process is illustrated through the
following example. This represents sensor data collected by a follower UAV
for the task at hand that is to be transmitted to the FL.

Routing Algorithm through an Example

Figure 2. Relay node selection

Consider the formation topology shown in Figure 2. The link cost factors
are taken into consideration to route data to the FL. The following steps are
implemented to route data using the OEDSR protocol:
1. Start with an empty relay list for source UAV n: Relay(n)={ }.
Here UAV n4 and n7 are CHs.
2. First, CH n4 checks with which nodes it is in range with. In this
case, CH n4 is in range with nodes n1, n2, n3, n5, n8, n9, n12, and n10.
3. The nodes n1, n2, and n3 are eliminated as potential RNs because
the distance from them to the FL is greater than the distance from
CH n4 to the FL.
4. Now, all the nodes that are in range with CH n4 transmit
RESPONSE packets and CH n4 makes a list of possible RNs,
which in this case are n5, n8, n9, n12, and n10.
5. CH n4 sends this list to CH n7. CH n7 checks if it is range with any
of the nodes in the list.
6. Nodes n9, n10, and n12 are the nodes that are in range with both CH
n4 and n7. They are selected as the potential common RNs.
204 Unmanned Aerial Vehicles (UAV) and Drones

7. The link cost factors for n9, n10, and n12 are calculated.
8. The node with the maximum value of LCF is selected as the RN
and assigned to Relay(n). In this case, Relay(n)={n12}.
9. Now UAV n12 checks if it is in direct range with the FL, and if it
is, then it directly routes the information to the FL.
10. Otherwise, n12 is assigned as the RN, and all the nodes that are in
range with node n12 and whose distance to the FL is less than its
distance to the FL are taken into consideration. Therefore, UAV’s
n13, n16, n19, and n17 are taken into consideration.
11. The LCF is calculated for n13, n16, n19, n14, and n17. The node
with the maximum LCF is selected as the next RN. In this case
Relay(n) = {n19}.
12. Next the RN n19 checks if it is in range with the FL. If it is, then
it directly routes the information to the FL. In this case, n19 is in
direct range, so the information is sent to the FL directly.

Optimality Analysis for OEDSR

To prove that the proposed route created by OEDSR protocol is optimal in
all cases, it is essential to show it analytically.
Assumption 1: It is assumed that all UAV’s in the network can calculate
or measure the relative distance to the FL at any time instant using the
formation graph information or local sensory data.
Theorem 4.3.1: The link cost factor-based routing generates viable RNs
to the FL.
Proof: Consider the following two cases
Case I: When the CHs are one hop away from the FL, the CH selects
the FL directly. In this case, there is only one path from the CH to the FL.
Hence, OEDSR algorithm does not need to be used.
Case II: When the CHs have more than one node to relay information,
the OEDSR algorithm selection criteria are taken into account. In Figure 3,
there are two CHs, CH1 and CH2. Each CH sends signals to all the other
nodes in the network that are in range. Here, CH1 first sends out signals to
n1, n3, n4, and n5 and makes a list of information about the potential RN.
The list is then forwarded to CH2. CH2 checks if it is in range with any of
the nodes in the list. Here, n4 and n5are selected as potential common RNs.
A single node must be selected from both n4 and n5 based on the OEDSR
link cost factor. The cost to reach n from CH is given by (2). So based on the
 Neural Network Control and Wireless Sensor Network-based ... 205

OEDSR link cost factor, n4 is selected as the RN for the first hop. Next, n4
sends signals to all the nodes it has in range, and selects a node as RN using
the link cost factor. The same procedure is carried on till the data is sent
to the FL. Lemma 4.3.2: The intermediate UAV’s on the optimal path are
selected as RNs by the previous nodes on the path. Proof: A UAV is selected
as a RN only if it has the highest link cost factor and is in range with the
previous node on the path. Since OEDSR maximizes the link cost factor,
intermediate nodes that satisfy the metric on the optimal path are selected
as RNs.
Lemma 4.3.3: A UAV can correctly compute the optimal path (with lower
end to end delay and maximum available energy) for the entire network
topology.
Proof: When selecting the candidate RNs to the CHs, it is ensured that
the distance form the candidate RN to the FL is less than the distance from
the CH to the FL. When calculating the link cost factor, available energy is
divided by distance and average end-to-end delay to ensure that the selected
nodes are in range with the CHs and close to the FL. This helps minimize
the number of multi-point RNs in the network.
Theorem 4.3.4: OEDSR protocol results in an optimal route (the path
with the maximum energy, minimum average end-to-end delay and minimum
distance from the FL) between the CHs and any source destination.

Figure 3. Link cost calculation

SIMULATION RESULTS
A wedge formation of three identical quadrotor UAV’s is now considered in
MATLAB with the formation leader located at the apex of the wedge. In the
simulation, follower 1 should track the leader at a desired separation sjid = 2
m , desired angle of incidence (0 rad) α jid = , and desired bearing (3 rad) β jid
=π while follower 2 tracks the leader at a desired separation sjid = 2 m desired
206 Unmanned Aerial Vehicles (UAV) and Drones

angle of incidence, (10 rad) α jid −= π , and desired bearing (3 rad) β jid −= π .
The desired yaw angle for the leader and thus the formation is selected to be
sin( 3.0 t) ψ d = π . The inertial parameters of the UAV’s are taken to be as m
= 9.0 kg and 2 J = diag {63.0,42.0,32.0} kg m2 , and aerodynamic friction is
modeled as in (Dierks and Jagannathan, 2008). The parameters outlined in
Section 3.3 are communicated from the leader to its followers using single
hop communication whereas results for OEDSR in a mobile environment
can be found in (Jagannathan, 2007). Each NN employs 5 hidden layer
neurons, and for the leader and each follower, the control gains are selected
to be, KΩ1 = ,23 KΩ2 = ,80 KΩ3 =20, K = diag {30,10,10} ρ , kv1 = ,10 kv2 =
,10 kv3 = 30, K = diag {30,30,30} Θ , and K = diag {25,25,25} ω . The NN
parameters are selected as, FΩ = ,10 κ Ω =1, and Fc = ,10 κc = 1.0 , and the
maximum desired pitch and roll values are both selected as 2π 5/ .
Figure 4 displays the quadrotor UAV formation trajectories while
Figures 5-7 show the kinematic and dynamic tracking errors for the leader
and its followers. Examining the trajectories in Figure 4, it is important to
recall that the bearing angle, βji , is measured in the inertial reference frame
of the follower rotated about its yaw angle. Examining the tracking errors
for the leader and its followers in Figures 5-7, it is clear that all states track
their desired values with small bounded errors as the results of Theorem
3.3.1 suggest. Initially, errors are observed in each state for each UAV, but
these errors quickly vanish as the virtual control NN and the NN in the
actual control law learns the nonlinear UAV dynamics. Additionally, the
tracking performance of the underactuated states xv and yv implies that the
desired pitch and roll, respectively, as well as the desired angular velocities
generated by the virtual control system are satisfactory.

Figure 4. Quadrotor UAV formation trajectories

 Neural Network Control and Wireless Sensor Network-based ... 207

Figure 5. Tracking errors for the formation leader

Figure 6. Tracking errors for follower 1

Figure 7. Tracking errors for follower 2

CONCLUSIONS
A new framework for quadrotor UAV leader-follower formation control was
presented along with a novel NN formation control law which allows each
follower to track its leader without the knowledge of dynamics. All six DOF
are successfully tracked using only four control inputs while in the presence
of unmodeled dynamics and bounded disturbances. Additionally, a discovery
and localization scheme based on graph theory and ad hoc networks was
presented which guarantees optimal use of the UAV’s communication
links. Lyapunov analysis guarantees SGUUB of the entire formation, and
numerical results confirm the theoretical conjectures.
208 Unmanned Aerial Vehicles (UAV) and Drones

REFERENCES
1. Das, A.; Spletzer, J.; Kumar, V.; & Taylor, C. (2002). Ad hoc networks for
localization and control of mobile robots, Proceedings of IEEE Conference
on Decision and Control, pp. 2978-2983, Philadelphia, PA, December 2002
2. Desai, J.; Ostrowski, J. P.; & Kumar, V. (1998). Controlling formations of
multiple mobile robots, Proceedings of IEEE International Conference on
Robotics and Automation, pp. 2964-2869, Leuven, Belgium, May 1998
3. Dierks, T. & Jagannathan, S. (2008). Neural network output feedback
control of a quadrotor UAV, Proceedings of IEEE Conference on
Decision and Control, To Appear, Cancun Mexico, December 2008
4. Fierro, R.; Belta, C.; Desai, J.; & Kumar, V. (2001). On controlling
aircraft formations, Proceedings of IEEE Conference on Decision and
Control, pp. 1065-1079, Orlando, Florida, December 2001
5. Galzi, D. & Shtessel, Y. (2006). UAV formations control using
high order sliding modes, Proceedings of IEEE American Control
Conference, pp. 4249-4254, Minneapolis, Minnesota, June 2006
6. Gu, Y.; Seanor, B.; Campa, G.; Napolitano M.; Rowe, L.; Gururajan,
S.; & Wan, S. (2006). Design and flight testing evaluation of formation
control laws. IEEE Transactions on Control Systems Technology, Vol.
14, No. 6, (November 2006) page numbers (1105- 1112)
7. Jagannathan, S. (2007). Wireless Ad Hoc and Sensor Networks, Taylor
& Francis, ISBN 0-8247- 2675-8, Boca Raton, FL
8. Lewis, F.L.; Jagannathan, S.; & Yesilderek, A. (1999). Neural Network
Control of Robot Manipulators and Nonlinear Systems, Taylor &
Francis, ISBN 0-7484-0596-8, Philadelphia, PA
9. Neff, A.E.; DongBin, L.; Chitrakaran, V.K.; Dawson, D.M.; & Burg,
T.C. (2007). Velocity control for a quad-rotor uav fly-by-camera
interface, Proceedings of the IEEE Southeastern Conference, pp. 273-
278, Richmond, VA, March 2007
10. Saffarian, M. & Fahimi, F. (2008). Control of helicopters’ formation using non-
iterative nonlinear model predictive approach, Proceedings of IEEE American
Control Conference, pp. 3707-3712, Seattle, Washington, June 2008
11. Xie, F.; Zhang, X.; Fierro, R; & Motter, M. (2005). Autopilot based
nonlinear UAV formation controller with extremum-seeking,
Proceedings of IEEE Conference on Decision and Control, pp 4933-
4938, Seville, Spain, December 2005
 CHAPTER

10
Dubins Waypoint, Navigation of
Small-Class Unmanned Aerial Vehicles

Larry M. Silverberg, Dahan Xu

Mechanical and Aerospace Engineering, North Carolina State University,
Raleigh, USA

ABSTRACT
This paper considers a variation on the Dubins path problem and proposes
an improved waypoint navigation (WN) algorithm called Dubins waypoint
navigation (DWN). Based on the Dubins path problem, an algorithm is
developed that is updated in real-time with a horizon of three waypoints. The
purpose of DWN is to overcome a problem that we find in existing WN for

Citation: Silverberg, L. and Xu, D. (2019), “Dubins Waypoint, Navigation of

Small-Class Unmanned Aerial Vehicles”. Open Journal of Optimization, 8, 59-72.
doi: 10.4236/ojop.2019.82006.
Copyright: © 2019 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
210 Unmanned Aerial Vehicles (UAV) and Drones

small-class fixed-wing unmanned aerial vehicles (UAV) of not accurately

reaching waypoints. This problem results at times in high overshoot and, in
the presence of wind disturbances, it can cause a vehicle to miss the waypoint
and swirl around it. To prevent this, the DWN creates “new waypoints”
that are in the background, called turning points. Examples illustrate the
improvement of the performance of WN achieved using the DWN algorithm
in terms of the targeting of waypoints while reducing fuel and time.
Keywords: Dubins Path, Waypoint Navigation, Unmanned Aerial
Vehicles, Autonomy, Shortest Path, Fuel, Optimization

INTRODUCTION
The user community of unmanned aerial vehicle (UAV) systems has been
growing significantly; the commercial market net worth reached a reported
$8.3 billion in 2018 [1] , the largest growth in the commercial markets
being in small-class (under 55 pounds) vehicles. This paper develops an
algorithm for a subset of this class of UAV, in particular, for fixed-wing
vehicles. Users of small-class UAV are presently navigating autonomously
by open-source algorithms such as Mission Planner, Cape, and Pix4D. These
navigational systems employ waypoint navigation (WN), wherein the user
enters waypoints, whether a priori (static) or not (dynamic). The practice of
WN in this community is a constraint to which development work adheres.
The community is growing, and with that, is expected to demand higher
accuracy. In RC flight, the racing community already requires reaching
targets accurately and the growth in autonomous flight with its varied
missions will create more demand to reach waypoints accurately.
The shortest path between two waypoints with a limited turning radius
is called a Dubins path [2] . WN employing Dubins paths has been studied
considerably. Ariff and Go conducted a thorough examination of the
dynamics involved in WN with a focus on dynamic soaring algorithms [3] .
Goerzen, Kong, and Mettler made rigorous comparisons between different
path planning algorithms for unmanned vehicle navigation and identified
advantages and disadvantages [4] . Manyam, et al. expanded the Dubins
path problem from two waypoints to three waypoints where all three points
and their headings are constrained [5] . Wolek and Woolsey implemented
the Dubins path problem with estimates of unsteady velocity disturbances
[6] . McGee and Hedrick considered the situations in which the disturbance
is less significant and more predictable [7] . The vehicle was placed in a
moving frame as a means of finding a piecewise continuous optimal path.
 Dubins Waypoint, Navigation of Small-Class Unmanned Aerial Vehicles 211

Milutinovic et al. applied the characteristics of a Dubins vehicle to the

circumnavigation algorithm, together with a feedback controller to improve
the vehicle’s performance [8] . Ketema and Zhao formulated the optimization
problem of path planning under wind disturbances and identified situations
in which the target points may be not reachable [9] . Meyer, Isaiah and Shima
examined and proposed a solution for a vehicle to intercept a moving target
with a Dubins path [10] . Criteria were given whether a target can or cannot
be intercepted and an analytical solution was obtained by a path elongation
algorithm. Wang et al. proposed a solution using a Dubins path to solve
the formation rendezvous problem of multiple UAVs [11] . Medeiros and
Urrutia optimized both path length and direction changes for a Dubins path
problem, by dividing the path planning process into several stages, bringing
in the algorithm of the travelling salesman’s problem [12] .
This paper considers a variation on the Dubins path problem and proposes
an improved navigation algorithm called Dubins waypoint navigation
(DWN). The proposed DWN algorithm considers a current vehicle state
(position and velocity) along with the next two prescribed waypoints to
plan the vehicle’s path. Compared with traditional WN, one more waypoint
is involved in the determination of the path, enabling consideration of the
manner in which the vehicle should circle around its previous waypoint. The
objective is similar to the Dubins path problem in that a shortest path and
a minimum fuel path between waypoints consists of straight-line segments
and arcs. The major difference is that in DWN the heading at the waypoints
are not specified and we consider three points instead of two. Also, for
simplicity, the paper only considers the case in which the vehicle’s turning
radius is constant.
The interest here lies in an algorithm that is updated in real-time with
a horizon of three waypoints. The proposed DWN algorithm employs a
current vehicle position and heading for the first waypoint, uses a prescribed
vehicle position for the second waypoint, not specifying its heading, and
uses a third waypoint to determine how the vehicle should circle around the
second waypoint.
The purpose of DWN is to overcome a problem that we find in existing
WN for small-class fixed-wing vehicles of not accurately reaching waypoints.
This problem results from the present use of a waypoint radius as a threshold
that determines when to move on to a next waypoint. Presently, a vehicle
approaches a first waypoint and switches to navigating to a second waypoint
when the vehicle is within a waypoint radius around the first waypoint,
212 Unmanned Aerial Vehicles (UAV) and Drones

thereby not accurately reaching the first waypoint. This leads the operator
to desire a smaller waypoint radius. However, the smaller waypoint radius
causes several problems. First, it causes high overshoot. Secondly, in the
presence of wind disturbances, it can cause a vehicle to miss the waypoint
and swirl around it.
When correcting this problem, it became important to preserve the
simplicity of prescribing waypoints without adding operator complexity.
Therefore, the DWN does not have to change anything in the foreground,
including the operator’s specified waypoints. Instead, the DWN creates
“new waypoints” that can be in the background, called turning points, as
explained in the method section.
Section 2 describes the DWN algorithm. Section3 compares performances
of WN and DWN with and without unknown wind disturbances, and the
paper finishes in Section 4 with a summary and conclusions.

METHOD

Flow Chart
This section develops the DWN. The DWN eliminates waypoint circles
and, instead, employs the new concept of turning points along with the new
criterion for updating turning points based on whether or not a turning point
is reachable. See Figure 1 showing the DWN flow chart. As shown, the
DWN approaches navigation as an optimization problem with in a local
optimization space that consists of a current point and two future waypoints.
At any instant, the vehicle seeks to reach a turning point that is determined by
minimizing fuel along the path up to the second future waypoint constituting
the horizon. The trajectory, and hence the optimization problem, is updated
once the turning point is unreachable. The reachability condition is
determined from the vehicle’s position, heading, and turning radius. For the
purposes of real-time implementation, the trajectory may be updated more
frequently than when the turning point is updated (like under the conditions
of high winds). The stated optimization problem yields desired paths that are
determined in closed-form.

Finding Turning Points

Consider the illustrative WN problem shown in Figure 2 and, in particular,
the navigation to the second and third waypoints shown in Figure 3.
 Dubins Waypoint, Navigation of Small-Class Unmanned Aerial Vehicles 213

Figure 1. Flow chart of the DWN.

Figure 2. Illustrative WN problem.

Figure 3. Navigating to the 2nd and 3rd waypoints.

214 Unmanned Aerial Vehicles (UAV) and Drones

The vehicle starts at W0 and heads to W1. At this point, the DWN considers
waypoints W0, W1, and W2. The other points are beyond the horizon.
Figure 3 shows the paths from W0 to W2 (left drawing) and from W1 to
W3 (right drawing).
As shown, T1 denotes the turning point associated with W1. It lies on
the line tangent to a turning circle (not to be confused with the waypoint
circle in classical WN) that has a turning radius R. The turning radius,
specified by the user, is the desirable radius of the turns around waypoints.
An optimization problem minimizes the fuel from W0 around and touching
W1 to W2. The orientation of the turning circle is free to rotate about W1 so
the optimization problem is a minimization problem of fuel expressed in
terms of the turning circle’s orientation. The fuel consumed during a turn
is greater than the fuel consumed while flying in a straight line over the
same distance. Even though the minimization is over the distance extending
to W2, the DWN updates the vehicle’s path once it reaches point T1. Thus,
the vehicle does not follow the section of the path from T1 to W2 (In Figure
3, the paths followed are solid lines and the paths not followed are dashed
lines.) The DWN recalculates that not-followed section of the path in the
next iteration. Under ideal conditions, DWN updates the planned path when
the vehicle crosses T1, at which point T1 is determined to be unreachable.
Under real conditions, disturbances can result in the vehicle not reaching
point T1 accurately. Point T1 is determined to be unreachable at a point C
that is different but nearpoint T1. Referring to the right drawing, the new
horizon is set to W3 and the new turning point T2 is determined from the
vehicle’s current point C and current direction of flight and the locations of
points W2 and W3. As shown, the optimization problem now minimizes the
fuel from C around and touching W2 to W3. The optimization problem is
now a minimization problem of fuel from C to W3 expressed in terms of the
new turning circle’s orientation. The DWN updates the iterations until the
vehicle reaches its last waypoint.

Key Parameters
Figure 4 shows the nth iteration. The vehicle is heading to Wn. A coordinate
system was set up so its x-axis points from C to Wn and its y-axis is
perpendicular to x in the direction of Wn+1. The data is scaled such that the x
and y coordinates represent non-dimensional lengths of distance divided by
turning radius R; equivalently R = 1. We perform all of the calculations after
the geometric parameters are normalized with respect to R.
 Dubins Waypoint, Navigation of Small-Class Unmanned Aerial Vehicles 215

1
The x axis, because it is along the line through C and Wn, can cause β
to exceed 90˚.
As shown, α is the heading angle (between –180˚ and 180˚) and β is the
turn angle (β is between 0˚ and 90˚)1. We assume that there are no sharp turns
between specified waypoints (the angle between any two lines connecting
three consecutive waypoints is never greater than 90˚). At point C, the
vehicle turns counter-clockwise (CCW), or clockwise (CW). Likewise, it
turns around Wn CCW or CW. In total, there are four cases: CCW-CCW,
CCW-CW, CW-CCW, and CW-CW.

Figure 4. Set up of the coordinate system.

The interest lies in finding closed-form expressions for the paths as
functions of A, B, α and β by curve fitting. Toward this end, note that the
minimum fuel paths are continuous with respect to these parameters for
each case, individually. Discontinuities occur when transitioning from one
case to another. Therefore, the curve-fitting needs to be performed for each
case and the case that consumes the smallest amount of fuel yields the true
minimum (See Figure 5).

Optimization Problem
Figure 6 shows the geometry of the CW-CCW case. The fuel from C to
Wn+1 is a function of the orientation angle θ of the turning circle. (The other
cases, not shown, are similar.) The fuel is

(1)
where La is the total length of the two arcs (around the first and second
turning circles), Ls is the length of the two straight segments (after the first
and second turning circles), and wa and ws are corresponding weights (in
units of fuel per length). Letting wa = ws = 1 (in which the weights have units
of 1), yields the shortest path problem:
216 Unmanned Aerial Vehicles (UAV) and Drones

(2)
In the results section, we will show that the shortest path and minimum
fuel solutions are nearly indistinguishable under a broad range of conditions,
allowing the shortest path solution to approximate the minimum fuel solution.
This is important because the shortest path solution is independent of the
vehicle, increasing ease of implementation and the versatility of DWN.

Figure 5. Different directions of turning.

Figure 6. Geometry for the CW-CCW case.

Referring again to Figure 6, the first straight line segment starts at
tangential point (x1, y1) and ends at turning point Tn = (x2, y2). The second
line segment starts at tangential point (x3, x4). Note that we define the first
circle by the current position C and the vehicle’s heading; it is either CCW
or CW. The location of the second circle depends on the orientation angle θ.
The second line segment is tangent to the second circle and passes through
Wn+1 and therefore is uniquely expressed in terms of the angle δ shown,
which, in turn, depends on θ.
The coordinates x1, y1, x2, y2, x3, and y3 for each of the four cases are
given in Table 1 and Table 2.
By appropriately manipulating the geometric relationships, the path
length (the objective function) becomes a function of the orientation angle
 Dubins Waypoint, Navigation of Small-Class Unmanned Aerial Vehicles 217

θ of the turning circle, and the solution of this 3-point optimization problem
becomes a function of θ. This part of the iteration can be solved off-line
and curve-fit to determine Tn. There are two ways to fit Tn. One way is to
directly fit Tn to A, B, α, and β. The second way is to fit θ to A, B, α, and β
followed by determining Tn from θ. The second way of curve fitting the data
will inherently lead to a more accurate curve fit than the first way because
the relationship between Tn and θ is rather complicated and its analytical
relationship is available, too [13] .

Table 1. Location of the end points of the first line segment.

Table 2. Location of second line segment.

Updating the Turning Point

As described earlier, the DWN guides a vehicle to a waypoint until it becomes
218 Unmanned Aerial Vehicles (UAV) and Drones

unreachable, at which point it updates the turning point. Figure 7 shows the
reachability condition. As shown, a reachability zone consists of CCW and CW
circles “enclosed” on the rear by a tangent line. The reachability zone is fixed
to the vehicle. The vehicle is trying to reach waypoint Tn. At a given instant of
time, Tn could be either inside or outside the reachability zone. (In the figure,
Tn is outside of the reachability zone.). As soon as it is inside the reachability
area, Tn becomes unreachable, and the DWN updates the turning point to Tn+1.
Note that the purpose of “enclosing” the CCW and CW with the line
segment was to prevent the possibility of point C passing Tn undetected,
which would otherwise be possible because the detection is not truly
performed continuously but only at discrete instances of time.
Comparing the reachability zone used in DWN with the waypoint
circle used in WN, first notice that the reachability zone, in the absence
of disturbances, allows the vehicle to accurately reach a first waypoint, as
opposed to beginning its turn to a second waypoint before reaching the first
waypoint. Secondly, in WN, the operator can easily be fooled into the natural
desire to select a waypoint radius that is smaller than the vehicle’s turning
radius toward the goal of more closely reaching a waypoint. However, when
doing so in the presence of a wind disturbance, this can result in missing
the waypoint, finding it difficult to reach, and causing the vehicle to swirl
around the waypoint. In the DWN, no waypoint radius is considered.

Parameterization
We formulated the 3-point horizon optimization problem to keep the
computational effort minimal for real-time implementation. To further reduce
computational effort and for robustness, we parameterized the solution,
effectively reducing it to a look-up table. Toward this end, we determined
the orientation angle θ of a turning circle as a function of the four parameters
A, B, α, and β. We parameterized the orientation angle by smoothly fitting
the data to the four parameters separately for each case (CCW-CCW, CCW-
CW, CW-CCW, and CW-CW). Toward distinguishing between the different
cases, we needed to parameterize the boundaries of each of the cases in
terms of the four parameters. In particular, α and β transition from case to
case due to changes in A and B. For example, consider the graph of α versus
β shown in Figure 8 for particular parameters (A = B = 4, wa = ws = 1).
 Dubins Waypoint, Navigation of Small-Class Unmanned Aerial Vehicles 219

Figure 7. Examples of reachable and unreachable points.

As shown, there are three boundaries and three interior regions. (It turns
out that the CW-CW case never produces an optimal orientation angle.) We
express the boundary of the turn angle β as a function of the heading angle
α in which its coefficients are expressed as a function of the distances A and
Bas follows:

(3)

(4)
We determined the coefficients b0i through b6i (i = 1, 2, 3) separately for
each of the cases. The real-time procedure of updating the turning points is
as follows:
1) Calculate A, B, α, and β at an instant of time from a current state
of the vehicle.
2) Calculate the coefficients associated with the boundaries between
the interior regions (See [13]).
3) Determine the interior region in which the vehicle lies.
4) Calculate the orientation angle θ of the next turning circle (See
[13]).
5) Calculate the next turning point Tn (expressed in terms of θ
analytically).
220 Unmanned Aerial Vehicles (UAV) and Drones

RESULTS
Let us continue with the 6-waypoint example and compare WN and DWN.
In WN, navigation performance depends on the minimum turning radius
and the waypoint radius. In DWN, navigation performance depends on
minimum turning radius alone. The feedback control algorithms are the same
for WN and DWN. However, the resulting overshoots differ. An illustrative
feedback control algorithm described below accounts for turning toward an
endpoint limited by the vehicle’s minimum turning radius R and serves as a
basis for the comparison between WN and DWN.

Figure 8. Distributions of different turning conditions.

An Illustrative Feedback Control Algorithm

Consider Figure 9 showing a reference path with endpoints that are waypoints
in the case of WN and that are turning points in the case of DWN. A vehicle
at time ti heads in a direction ϕHi (not shown) a distance si from the reference
path. The vehicle travels a distance vΔt over a time step. At time ti+1, the
vehicle heads in a direction ϕi+1 = ϕi + ψi in which ψi is the turn angle over
one iteration, given by

(5)
Above, ϕTi is a target angle and ϕRi is a reference angle (all of the angles
are positive counter-clockwise), g and h are control gains, and ϕT/|ϕT|
determines whether ϕR is counter-clockwise or clockwise.
The parameters used in the results, both with and without wind disturbances,
 Dubins Waypoint, Navigation of Small-Class Unmanned Aerial Vehicles 221

are as follows: The minimum turning radius was R = 0.7, the time step was
0.02, the vehicle speed was v = 0.8, and the control gains were g = 0.1 and h
= 5.25. With these parameters the turning radius was 0.7.

The Shortest Path Approximation

Let us compare the minimum fuel and the shortest path solutions, the latter
determined when wa = ws = 1. In practice, the distance between waypoints is
at least four times larger than the turning radius R. Therefore, we will only
consider distances A and B that are more than or equal to 4 times greater
than R. When the distances A and B are exactly 4 times greater than R,
the vehicle is flying in a straight line the least amount of the time and, for
this case, we expect the difference between the minimum fuel solution and
the shortest path solution to be the greatest. Figure 10 shows minimum
fuel solutions in which wa varies from 1— the shortest path case—to wa =
1.5, for which the fuel cost during turning flight is 50% greater than the
fuel cost during straight flight (In small-class fixed-wing vehicles the fuel
penalty of turning is less than 50%). As shown, the orientation angle of
the turning circle does not change more than 4˚. The minimum-fuel path is
very close to the shortest path and therefore the shortest path approximates
the minimum fuel path. The use of the shortest path would therefore be
acceptable in the vast majority of small-class UAV problems. This finding is
significant because the shortest path solutions apply to any vehicle whereas
the minimum fuel solutions need to consider vehicle parameters. In the
results below, we employ the shortest path solutions.

Figure 9. Parameters for the controller.

222 Unmanned Aerial Vehicles (UAV) and Drones

Figure 10. Example of a CCW-CCW condition (A = B = 4, α = −45˚, β = 30˚,

CCW-CCW).

Navigation in the Absence of an Unknown Wind Disturbance

Figure 11 compares WN and DWN in the absence of an unknown wind
disturbance. Indeed, when the wind disturbance is completely accommodated
for (by subtracting it out), DWN and WN exhibit the performance shown
in Figure 11. The minimum turning radius R of the vehicle is the same in
each case. The three cases consider WN waypoint radii of 0.1R, 0.5R, and
R, respectively.
First consider DWN (solid lines). In each case, the vehicle passes through
waypoints exactly and follows the minimum fuel paths—by design. Next,
consider WN. In the first case, the vehicles passes close to the waypoints but
with significant overshoot. In the other two cases, we see as the waypoint
radius increases that the vehicle passes farther from the waypoints. In the
third case, the vehicle undershoots the desired path. Note that in the WN and
DWN cases given above, the feedback control had very little effect on the
response because there was no unknown wind disturbance.

Navigation in the Presence of an Unknown Wind Disturbance

Figure 12 shows WN and DWN responses in the presence of a wind
disturbance. The wind velocity is constant and unknown, set to 10% of
the drone’s speed directed from the bottom to the top of the figures (in the
horizontal plane) and not accommodated for.
In the left figure, the waypoint radius was set to be small, intending
to achieve higher accuracy. However, the wind exceeds the capability of
the feedback controller and the vehicle just misses the second waypoint
 Dubins Waypoint, Navigation of Small-Class Unmanned Aerial Vehicles 223

sending it into a swirl. DWN also misses the turning point paired with
the second waypoint but moves on to complete its journey. It can do this
because it recognizes the waypoint to be unreachable. In the middle and the
right figures, the waypoint radii are larger and WN manages to navigate the
vehicles. It is worth noticing that errors exist for both WN and DWN. The
DWN errors result from using the turning points for navigation instead of
waypoints. It is clear that the DWN errors are smaller than the WN errors
due to its features already discussed, hence one still recognizes that DWN
improves the performance of the WN navigation.

Figure 11. WN and DWN (RWP = 0.1R, 0.5R, and R).

Figure 12. WN with and without the DWN in the presence of a disturbance
(RWP = 0.1R, 0.5R, and R).

CONCLUSION
WN suffers from an existing trade-off between minimum turning radius
224 Unmanned Aerial Vehicles (UAV) and Drones

and waypoint radius that prevents vehicles from reaching waypoints and
closely following desired paths. This paper developed an algorithm, called
Dubins waypoint navigation (DWN) that remedies this problem by not using
waypoint circles. Instead, we introduced a reachability zone and turning
points. The DWN significantly reduces undershoot, overshoot, and swirl.
Furthermore, the algorithm remedies the problem in a way that can be hidden
from the operator to avoid confusion. The paper showed, by establishing a
horizon that includes two future waypoints, how to improve the performance
of WN in terms of the targeting of waypoints while reducing fuel and time.
Pertaining to real-time implementation, we also reduced the computational
effort, essentially eliminating it, by parameterizing the shortest path solution.
We also compared WN and DWN tracking performance (in the absence of
an unknown disturbance) and regulation performance (in the presence of an
unknown disturbance). The comparisons are illustrative of the improvements
in path following (tracking and regulation) obtained by DWN.

ACKNOWLEDGEMENTS
The authors gratefully recognize the Namibia Wildlife Aerial Observatory
(WAO) for its support of this work.

NOMENCLATURE
A = distance between current position and the following waypoint
a = coefficient for curve fitting β with respect to δ
B = distance between the following two successive waypoints
b = coefficient for curve fitting a with respect to A and B
C = current position
Fu = total fuel cost
g = proportional term for PID controller
h = derivative term for PID controller
La = total length of the arcs
Ls = total length of the straight segments
R = turning radius of the vehicle
T = target point
v = speed of the vehicle
 Dubins Waypoint, Navigation of Small-Class Unmanned Aerial Vehicles 225

W = waypoint
wa = fuel cost per unit length travelled on an arc
ws = fuel cost per unit length travelled on a straight line
α = angle of heading in local system
β = angle between the following two successive waypoints in local system
γ = angular position of the first tangential point in local system
δ = angular position of the last tangential point in local system
θ = angular position of the center of the turning circle
ϕi = angle that denotes the heading in global frame at the th time instance
ϕR = angle of turning resulted from the feedback controller
ϕT = angle between heading and the target
ψ = angle of turn in one time step

CONFLICTS OF INTEREST
The authors declare no conflicts of interest regarding the publication of this
paper.
226 Unmanned Aerial Vehicles (UAV) and Drones

REFERENCES
1. Drubin, C. (2013) UAV Market Worth $8.3 B by 2018. Microwave
Journal, 56, 37.
2. Tsourdos, A., White, B. and Shannugavel, M. (2011) Path Planning
in Two Dimensions. In: Tsourdos, A., White, B. and Shanmugavel,
M., Eds., Cooperative Planning of Unmanned Aerial Vehicles, Wiley,
Chichester, 30. https://doi.org/10.1002/9780470974636
3. Ariff, O. and Go, T. (2011) Waypoint Navigation of Small-Scale UAV
Incorporating Dynamic Soaring. The Journal of Navigation, 64, 29-44.
https://doi.org/10.1017/S0373463310000378
4. Goerzen, C., Kong, Z. and Mettler, B. (2010) A Survey of Motion
Planning Algorithms from the Perspective of Autonomous UAV
Guidance. Journal of Intelligent and Robotic Systems, 57, 65-100.
https://doi.org/10.1007/s10846-009-9383-1
5. Manyam, S., Rathinam, S., Casbeer, D. and Garcia, E. (2017) Tightly
Bounding the Shortest Dubins Paths through a Sequence of Points.
Journal of Intelligent & Robotic Systems, 88, 495-511. https://doi.
org/10.1007/s10846-016-0459-4
6. Wolek, A. and Woolsey, C. (2015) Feasible Dubins Paths in Presence
of Unknown, Unsteady 5 Velocity Disturbances. Journal of Guidance
Control and Dynamics, 38, 782-786. https://doi.org/10.2514/1.
G000629
7. McGee, T. and Hedrick, J. (2007) Optimal Path Planning with a
Kinematic Airplane Model. Journal of Guidance, Control, and
Dynamics, 30, 629-633. https://doi.org/10.2514/1.25042
8. Milutinovic, D., Casbeerm, D., Cao, Y. and Kingston, D. (2017)
Coordinate Frame Free Dubins Vehicle Circumnavigation Using Only
Range-Based Measurements. International Journal of Robust and
Nonlinear Control, 27, 2937-2960. https://doi.org/10.1002/rnc.3718
9. Ketema, Y. and Zhao, Y. (2010) Micro Air Vehicle Trajectory Planning
in Winds. Journal of Aircraft, 47, 1460-1463. https://doi.org/10.2514/1.
C000247
10. Meyer, Y., Isaiah, P. and Shima, T. (2015) On Dubins Paths to Intercept
a Moving Target. Automatica, 53, 256-263. https://doi.org/10.1016/j.
automatica.2014.12.039
11. Wang, Z., Lium L., Long, T. and Xu, G. (2018) Efficient Unmanned
Aerial Vehicle Formation Rendezvous Trajectory Planningf Using
 Dubins Waypoint, Navigation of Small-Class Unmanned Aerial Vehicles 227

Dubins Path and Sequential Convex Programming. Engineering

Optimization, 51, 1412-1429. https://doi.org/10.1080/030521
5X.2018.1524461
12. Medeiros, A. and Urrutia, S. (2010) Discrete Optimization Methods
to Determine Trajectories for Dubins’ Vehicles. Electronic Notes
in Discrete Mathematics, 36, 17-24. https://doi.org/10.1016/j.
endm.2010.05.003
13. Silverberg, L. and Xu, D. (2018) Minimum-Fuel Hidden Layer
Heuristic for SmallClass UAV. Master of Science Thesis, Mechanical
Engineering, North Carolina State University, Raleigh.
 CHAPTER

11
Modelling Oil-Spill Detection with
Swarm Drones

F. Aznar,1 M. Sempere,1 M. Pujol,1 R. Rizo,1 and M. J. Pujol2

Department of Computer Science and Artificial Intelligence, University of
1

Alicante, 03090 Alicante, Spain

Department of Applied Mathematics, University of Alicante, 03090
2

Alicante, Spain

ABSTRACT
Nowadays, swarm robotics research is having a great increase due to the
benefits derived from its use, such as robustness, parallelism, and flexibility.
Unlike distributed robotic systems, swarm robotics emphasizes a large number
of robots, and promotes scalability. Among the multiple applications of such

Citation: Chen-Charpentier, Benito, Aznar, F., Sempere, M., Pujol, M., Rizo, R.,
“Modelling Oil-Spill Detection with Swarm Drones”, Hindawi Journal on Analysis and
Models in Interdisciplinary Mathematics, Volume 2014, Article ID 949407, 14 pages,
https://doi.org/10.1155/2014/949407.
Copyright: © 2014 F. Aznar et al. This is an open access article distributed under the
Creative Commons Attribution License.
230 Unmanned Aerial Vehicles (UAV) and Drones

systems we could find are exploring unstructured environments, resource

monitoring, or distributed sensing. Two of these applications, monitoring,
and perimeter/area detection of a given resource, have several ecological
uses. One of them is the detection and monitoring of pollutants to delimit
their perimeter and area accurately. Maritime activity has been increasing
gradually in recent years. Many ships carry products such as oil that can
adversely affect the environment. Such products can produce high levels
of pollution in case of being spilled into sea. In this paper we will present a
distributed system which monitors, covers, and surrounds a resource by using
a swarm of homogeneous low cost drones. These drones only use their local
sensory information and do not require any direct communication between
them. Taking into account the properties of this kind of oil spills we will
present a microscopic model for a swarm of drones, capable of monitoring
these spills properly. Furthermore, we will analyse the proper macroscopic
operation of the swarm. The analytical and experimental results presented
here show the proper evolution of our system.

INTRODUCTION
Maritime activity has been increasing gradually in recent years. For example,
around 42000 vessels (excluding the fishing ones) pass throughout the
North Sea, carrying products that can adversely affect the environment, such
as oil, which can produce high levels of pollution in case of being spilled
into sea. Moreover, many pollutants are accidentally spilled from ships
during “normal” operations. These spills are probably small but become
significant due to the large number of ships.Dramatic incidences of marine
pollution, such as the Prestige oil spill off the Spanish north coast [1–3],
have highlighted the potential for human-caused environmental damage. In
attempting to mitigate or avoid future damage to valuable natural resources
caused by marine pollution, research has been undertaken by the scientific
community to study the processes affecting the fate and distribution of
marine pollution and especially to model and simulate these processes.
Furthermore, active systems, able to detect and track such kind of spills, are
an invaluable tool to help to locate and clean the affected resources [4–6].
Moreover, swarm robotics is an approach to solve problems inspired by
the collective behaviour of social animals and it is focused on the interaction
of multiple robots. It is a different approach to classical artificial intelligence,
where the main goal is usually to develop behaviours that mimic human
brain function. Swarm robotics is based on the metaphor of social insects and
 Modelling Oil-Spill Detection with Swarm Drones 231

emphasizes aspects like decentralized control, limited communication between

agents, local information, emergence of a global behaviour, and robustness [7–
10]. As it has been shown [10] it is different from other multirobotic systems,
because, for example, the swarm robots are relatively simple.
However, the distributed nature of these systems makes it difficult to
develop an architecture that correctly models a swarm and coordinates it to
perform complex tasks. In order to design a swarm behaviour, a mathematical
model must be provided for both the individual agents and the whole swarm
[11, 12]. These models will be tested to evaluate the performance of the swarm
before its deployment in real UAVs. This is mainly because this kind of systems
has emergent properties that make it difficult to predict the overall operation of
the swarm when only the local behaviour of the agents is being analysed.
In this paper, we will present a swarm behaviour to be used with robotic
drones in order to cover the perimeter or the area of an oil spill. We will use
GNOME [13], a very realistic environmental simulation system for marine
pollutants. With the data provided by this simulation, using real maritime
information and air currents predictions, we will create a virtual oil spill in
the Spanish coast. We will show that our behaviour is able to both detect
and track the spill for a given period of time. Moreover, we will also show,
through the use of the macroscopic model derived by the local rules of the
robots, that the swarm will track the oil pollutants. A Fokker-Plank equation
will be derived from the microscopic behaviour in order to predict the
probability that a drone stays in a given position at time t.

POLLUTANT DISPERSION MODEL

Whether the trajectory model is accurate, adequate, or correct is often
questioned. The accuracy and adequacy of a model are associated with
the data used for the calculations and modelled physical processes. The
modelling of an oil spill is not a simple task, mainly due to the number of
factors influencing the trajectories of pollutants: sea currents, winds, and
even the gravitational force or surface tension of water.There are several
applications that model and simulate pollutant discharges into the sea. These
applications are based on the modelling of the most important forces that
interact on an oil slick. In this paper, one of these applications has been used
to obtain realistic data of an oil spill off Spanish coast. As it will be seen in
subsequent sections, we use both ocean currents and wind data. These data
will be used for testing a realistic oil spill using a swarm of flying robots
(drones).
232 Unmanned Aerial Vehicles (UAV) and Drones

More specifically, we have used GNOME, an interactive environmental

simulation system designed for the rapid modelling of pollutant trajectories
in the marine environment. GNOME simulates oil movement due to winds,
currents, tides, and spreading. It was developed by the Hazardous Materials
Response Division (HAZMAT) of the National Oceanic and Atmospheric
Administration Office of Response and Restoration (NOAA OR&R).
HAZMAT uses GNOME during spill events to calculate the spill trajectory
and the associated uncertainty in that trajectory. As it has been mentioned
in [14], GNOME can be used for predicting how winds, currents, and
other processes might move and spread oil on the water, for studying how
predicted oil trajectories are affected by inexactness (uncertainty) in current
and wind observations and forecasts, and for predicting the chemical and
physical changes of the oil spill during the time that it remains on the water
surface.
One of the benefits of this tool is the possibility of using external files
containing information on tides, currents, and so forth. If this information
is added, then GNOME uses it to make the trajectory prediction of the spill
(even using uncertainty during simulation) in the specified region. The output
from the model consists of graphics, movies, and data files for postprocessing
in a geographic information system (GIS) or NOAA Emergency Response
Division’s (ERD1) in-house model GNOME Analyst.
The model used by this tool is general and applies to trajectory problems.
It is an Eulerian/Lagrangian model that is two-dimensional (2-D) in space,
but vertically isolated layered systems can be modelled. Shoreline maps are
inputs to the model, so any area can be modelled. The model automatically
handles either hemisphere or east or west longitude.
More specifically, random spreading (diffusion) is calculated by a simple
random walk with a square unit probability. The random walk is based on
the diffusion value, D, that represents the horizontal turbulent diffusion of
the spill in the water. During a spill, the value is calibrated as based on
overflight data. In this way diffusion and spreading are treated as stochastic
processes. Gravitational and surface tension effects are ignored, as these are
only important during the first moments of a spill. Complex representation
of subgrid diffusion and spreading effects are ignored.
The main diffusion equation used in GNOME is presented in [13], where
𝐷𝑥 and 𝐷𝑦 are the scalar diffusion coefficients in 𝑥 and 𝑦 directions and 𝐶
is the pollutant concentration:
 Modelling Oil-Spill Detection with Swarm Drones 233

(1)
GNOME simulates this diffusion with a random walk with any
distribution, with the resulting diffusion coefficient being one-half the
variance of the distribution of each step divided by the time step:

(2)
where 𝜎 is the variance of the distribution of diffused points and Δ𝑡 is
2

the time elapsed between time steps.

Evaporation in GNOME is modelled by a simplistic 3-phase evaporation
algorithm where the pollutant is treated as a three-component substance with
independent half-lives [15]:

(3)
where 𝑡𝑖 and 𝑡𝑖−1 are the time elapsed (age; in hours) at time step 𝑖 and
the previous time step 𝑡−1, respectively, since the LEs release. 𝐻1, 𝐻2, and
𝐻3 are the half-lives of each constituent (in hours) for the pollutant, and
𝑃1, 𝑃2, and 𝑃3 are the percentages of each constituent (as decimals) for the
pollutant.
Spilled substances are modelled as point masses (up to 10, 0004 ) called
LEs (Lagrangian elements) or “splots” (from “spill dots”). Spills can be
initialized as one-time or continuous releases and as point or line sources or
evenly spaced in a grid on the map for diagnostic purposes.
Once GNOME executes a simulation, the solution is produced in the
form of a trajectory. GNOME provides two solutions to an oil spill scenario,
the best estimate trajectory and the uncertainty trajectory. The best estimate
solution shows the model result with all of the input data assumed to be
correct. The uncertainty solution allows the model to predict other possible
trajectories that are less likely to occur, but which may have higher associated
risks. In this paper we will use the uncertainty solution of pollutant particles
(represented by its LEs) for generating a continuous pollutant map. More
details of this mathematical model could be obtained in [13].
234 Unmanned Aerial Vehicles (UAV) and Drones

SWARM DESIGN
In this section we will analyse the features that must have a robotic swarm to
detect and follow an oil spill in effective way. We will also propose a specific
microscopic behaviour for this task. As mentioned above, the modelling of
this kind of pollutants is a complex task because of the interaction of many
factors. However, it is easy to extract a set of characteristics that a system
needs to locate this kind of spills.
On the one hand, the swarm must be able to detect and follow pollutants
that can change and move in time mainly by the action of advection and
diffusion. Depending on the application, two situations can be found: the
origin of the spill is known or the swarm initially makes the detection
process (in this work it is assumed that detection is one of the tasks to be
performed). On the other hand, the appearance of several polluting slicks
is very probable due, among other factors, to the transport and deposit of
sediments on the coast while the oil slick disperses and evaporates.
The behaviour of the swarm must be highly robust and tolerant to failures
and should be totally distributed. Therefore, all agents must have the same
importance when performing the behaviour, without the existence of any
agent more prominent than another. Finally, the behaviour should be highly
scalable for two reasons: robust issues and performance of the behaviour,
since as a first step it may be beneficial to use a reduced number of agents
until they find any evidence of a particular discharge.
Although in this paper behaviours will be analysed in a simulated way,
the features of agents are directly based on flying robots (our intention is
to use these algorithms in drones). These drones will have a video camera
that will use a vision algorithm to determine the existence of any oil slick
in its visual field (as presented in [4, 16]). For security reasons, we assume
that drones will fly at different heights, so that collisions in a position (x,
y) are avoided. We also assume that due to flight altitude (about 500 m
above sea level) the differences in the visual field caused by different flying
altitudes are negligible. All tests, unless otherwise specified, will be carried
out by a swarm of 200 drones. It is a medium-sized swarm, appropriate for
simulation.
Our main goal, once the behaviour is designed, is to determine the
ability of the swarm to locate, converge, and follow an oil spill. Therefore,
a macroscopic model to predict the global behaviour of the swarm and to
verify its performance will be specified. More specifically, we propose a
homogeneous behaviour, executed by all agents, consisting of three states.
 Modelling Oil-Spill Detection with Swarm Drones 235

Initially, drones look for any trace of the spill in the environment. Once the
spill is detected, the drone will head to it. Finally, the drone will try to stay
inside it (to cover the slick) or in its perimeter (to mark it).
Figure 1 shows the finite state machine (FSM) that governs the behaviour
of each robot. The initial state is Wander, since we initially assume that
the position of the spill is unknown. The transition from Wander state is
performed when the agent’s visual sensor detects an oil slick. In this case,
the new state will be Resource. This state (Resource) has two possible
transitions: the spill is not detected (transition d) so the system returns to
Wander state or the amount of oil detected is >80% of the image (transition
b), so, the system switches to InResource state. The agent remains in this
state if the detected oil percentage is >80%; otherwise, the system returns to
Resource state.

Figure 1: Finite state machine (FSM) that governs the operation of each agent.
The initial state is Wander. The transition a is triggered when the agent’s visual
sensor detects an oil slick. Transition b occurs when the amount of oil detected
is >80% of the image. Transition cis triggered when the amount of oil is ≤80%
of the image. Finally, transition d is triggered if no oil is detected.
Next, each behaviour specified in the FSM will be described in more
detail.
At the beginning of the execution agents start at Wander state. At that
time they have no idea where to find an oil spill. Therefore, all directions
are equally valid and they will move randomly. The velocity of the drone at
time t is defined as
236 Unmanned Aerial Vehicles (UAV) and Drones

(4)
where rand is a random uniform vector defined within the interval of
the maximum and minimum velocity of the drone, and 𝜇1 is the coefficient
of variability on the current velocity. With values close to 1, the robot will
move in a totally random way.
If a drone detects the resource, it heads to the resource to position itself
at the perimeter or over it, depending on the desired behaviour. The velocity
of the agent is defined by three factors:

(5)
where 𝛼1 + 𝛼2 + 𝛼3 = 1 and these values define the intensity of each factor.
More specifically, k𝑐 specifies the direction of the robot. This direction
is determined by the area with the larger intensity resource average:

(6)
where 𝑆 is sensor readings that detect the resource at a given time, pos(𝑠)
is the position vector of a reading, and pos(rob) is the position of the robot.
We assume that the intensity of the resource is in the range [0, 1], where 0
is the complete lack of resource and 1 is the unambiguous detection of it.
𝛾 determines the direction of the velocity vector, depending on whether
the robot is outside or inside de resource. The aim of this variable is,
therefore, to keep the robot on the perimeter of the resource:

(7)
where 𝜂 is a threshold that determines, from the quantity of resource
detected (0 means no resource and 1 maximum quantity of resource), if the
agent is located on the perimeter. If the main objective of the system is that
drones cover the pollute slick, then (𝑆) will be defined as 1 for any set of
readings 𝑆 at a given time.
vo specifies an avoidance vector with respect to all robots detected at a
given moment:
 Modelling Oil-Spill Detection with Swarm Drones 237

(8)
where 𝑅 is the set of detected robots, pos(𝑟𝑖) is the position of the detected
robot 𝑖, and pos(rob) is the position of the current robot.
Moreover, we will take into account the accuracy of the transmitted
locations: there are several factors that could make these locations not to
be optimal. We will include, therefore, a random component to model this
uncertainty in the movement of the robot: kr(𝑡) = k𝑟(𝑡−1)+rand⋅ 𝜇2, where
𝜇2 is the coefficient of variability on the velocity
Finally, when drones are located inside the spill and, therefore, the
borders of the resource are not detected, we assume that drones will develop
a random wander behaviour until they find water again because there is no
other information about which direction is better to follow:

(9)
where 𝜇3 is the coefficient of variability on the velocity.
In this section we have defined a microscopic behaviour for the detection
and tracking of an oil spill. Each agent of the swarm executes this behaviour
locally. It is a simple behaviour that can be run in low processing capacity
systems. This behaviour does not require any communication between
agents or a global positioning system.

EXPERIMENTATION
We will now present a set of experiments in order to test the operation of
the proposed microscopic behaviour. As we will see, these experiments use
simulations of oil spills based on real data.
We have simulated an oil spill in the coastal area of “Sant Antoni de
Portmany,” on Ibiza island, using an area of approximately 350 km2 . This
area has been chosen because of the various currents that affect it and its
proximity to the peninsula. Real weather data have been used for both sea
and air currents, considering the start date of the spill on April 10, 2013, at
0:00 h. We use these real data to run a simulation for seven days from the
start of the spill, simulating a ship loaded with 100000 barrels of oil. Figure
7 shows the map of the main area affected and the initial place of impact.
In order to work with data generated by GNOME we use the average of the
points of discharge generated by the application (More specifically, (𝑡)’ =
238 Unmanned Aerial Vehicles (UAV) and Drones

𝐺(𝑀(𝑡)⊕𝑠shape) where 𝑀(𝑡) is the map obtained with GNOME at time

𝑡, 𝐺 is a Gaussian filter with 𝜎 = 10|𝑆|, and ⊕ is a morphological binary
operator) (splots in GNOME), making uniform the continuity between
nearby contaminated areas and simplifying the simulation. This allows us to
reduce the number of splots needed in the simulation:

(10)
The previous equation shows default parameters used in simulations for
the microscopic model. These parameters have been adjusted taking into
account the microscopic behaviour experiments.
The simulation of the microscopic model has been developed using the
software MASON [17]. We use a swarm of 200 agents randomly distributed
by the environment that moves uniformly at 60 km/h. The simulation uses
small size drones (< 3 m2 ) that are able to visually detect an area of 1 km2 .
Using these parameters, several tests will be performed to verify the
correct operation of our model at local level. Initially, first tests check the
convergence of the swarm for a single static slick. Using the same example,
will be modified to cover the slick instead of marking its perimeter. Next
tests will verify the convergence of the swarm choosing an instant with
several active slicks. Finally, we will check the tracking of the movement
of the spill.

Spill Detection
First tests have been developed to detect and monitor a static slick. On the
one hand, the swarm will mark the perimeter of the slick. Figure 2 shows
 Modelling Oil-Spill Detection with Swarm Drones 239

the initial distribution of agents with their geographical location (a) and the
position of agents at time 𝑡 = 15000 s and 𝑡 = 30000 s ((b) and (c)). At time
𝑡 = 30000 s the swarm completely surrounds the oil slick. Figure 3 presents
the percentage of agents over an oil slick for ten independent simulations.
As it can be appreciated, this percentage increases progressively.

Figure 2: Distribution of agents with respect to time (in seconds) for the task of
detecting the perimeter of an oil spill: (a) geographical location of the spill and
initial position of agents, (b) distribution of agents at time 𝑡 = 1500 s, and (c)
distribution of agents at time 𝑡 = 30000 s.

Figure 3: Percentage of agents on an oil slick with respect to time (in seconds).
Ten different experiments have been carried out, showing the average and vari-
ance.
240 Unmanned Aerial Vehicles (UAV) and Drones

Moreover, as it has been discussed in the definition of the microscopic

model of the swarm, when (𝑆) = 1 the swarm will cover the slick instead of
marking its perimeter. Figure 4 shows the results of the swarm for this case,
where 10 independent simulations have been carried out.

Figure 4: Distribution of agents for the task of covering the spill ((𝑆) = 1): (a)
geographical location of the slick and initial position of agents, (b) position of
agents at time 𝑡 = 15000 s, (c) position of agents at time 𝑡 = 30000 s.
The experiments in Figure 2 show that the operation of the swarm is
correct: 50% of agents are able to detect the spill in less than 4 hours since
the beginning of the execution, marking correctly the perimeter of the spill.
However, in a real environment the appearance of several slicks is common.
This case will be analysed in the next section.

Several Spills
When an oil spill occurs, the oil may spread across kilometres generating
several oil slicks. For this reason, it is very important to verify the correct
operation of the swarm with different slicks that can spread out over a
wide area. In the simulation, at early hours, several areas containing large
proportions of pollutant are produced. For example, underwater currents can
move oil slick close to the coast, while the ship continues spilling oil into
the sea.
In order to check the operation of the microscopic model in these cases,
we have chosen an instant of the simulation where the slick has a complex
structure. Figure 5 shows the evolution of the distribution of robots. As it
 Modelling Oil-Spill Detection with Swarm Drones 241

can be seen in this figure, even in this complex case and without previous
data that indicates the spill trend, our model is able to locate and mark the
perimeter of the slicks. More specifically, Figure 6 presents the percentage
of agents that are located on a polluted area. In this figure it can be seen that
the number of agents increases with time, therefore distributing uniformly
on the slick.

Figure 5: Distribution of agents for the task of detecting the perimeter of an

oil spill on a map with several slicks: (a) geographical location of the spill and
initial position of agents, (b) distribution of agents at time 𝑡 = 15000 s, and (c)
distribution of agents at time 𝑡 = 30000 s.

Figure 6: Percentage of agents on an oil slick with respect to time (in seconds)
for a complex slick. Ten different experiments have been carried out, showing
the average and variance.
242 Unmanned Aerial Vehicles (UAV) and Drones

Figure 7: Origin and temporal evolution of the spill. Several snapshots that
show the position of the spill at different instants, measured in hours from the
beginning of the spill, are shown.

Spill Movement
Figure 5 tests have shown that the swarm is able to detect and mark the
perimeter/area with complex slicks.
However, in a real spill, finding the location of the spill is as important
as effective monitoring is. In this section we will analyse the behaviour of
the swarm taking into account that the spill will spread as time progresses.
A simulation of 168 hours of the evolution of the spill has been performed,
starting April 10, 2013, at 0:00 h.
In order to simplify the simulation, we capture an image of the state of
the spill every 4.5 hours. Although the spill will experience changes in this
period, the swarm can deal with these changes.
Figure 7 shows the origin point of the spill and some captures of its
evolution.
As in previous experiments, the swarm is distributed randomly by the
environment. In this case, in the simulation process, the location of the
resource obtained by GNOME is updated every 4.5 hours. Figure 8 shows
the evolution of the swarm with respect to the state of the oil spill.
 Modelling Oil-Spill Detection with Swarm Drones 243

Figure 8: Evolution of the swarm with respect to the oil slick. Multiple snap-
shots at different instants in time (measured in hours) are shown.
Figure 9 shows the percentage of drones that are on top of a resource at a
given time. In order to produce this graph, 10 different simulations have been
carried out. The graph shows the average and variance of these simulations.
In this graph a series of steps produced artificially by the simulator (because
of the sudden change from one state to the next) can be seen.

Figure 9: Percentage of agents over an oil slick with respect to time (in sec-
onds). Ten different experiments have been carried out, showing the average
and variance.
244 Unmanned Aerial Vehicles (UAV) and Drones

Nonetheless, the evolution of the swarm is correct. The swarm is able to

track the spill and mark each of the slicks.

MACROSCOPIC MODEL
Once the microscopic behaviour has been described, it is interesting to see the
global behaviour of the swarm. There are several techniques used to analyse
this behaviour [18] such as the use of recurrence equations, generated from
a microscopic behaviour defined by a PFSM or the definition of differential
equations. However, most of these methods allow only the analysis of the
evolution of the state transitions globally.
In this work, we consider the framework proposed in [19] in order to
obtain the probability distribution of the swarm position for any time t. This
will enable us to predict, in great detail, the behaviour of the overall system.
As described by [19], once the microscopic behaviour has been defined, the
global behaviour of the system can be calculated using the Fokker-Planck
equation:

(11)
where 𝑄 is the displacement by a collision. 𝜌(r, 𝑡)𝑟𝑥𝑑𝑟𝑦 is the probability
of encountering a robot at position 𝑟 within the rectangle defined by 𝑑𝑟𝑥 and
𝑑𝑟𝑦 at time t
This equation provides a method to statistically model a swarm of robots
based on modelling techniques of multiparticle systems from the field of
quantum physics. From a Langevin equation that represents the behaviour
of a single particle, the Fokker-Planck equation is derived for all the system.
As we have already seen in [20], the Fokker-Planck equation implements
the necessary abstraction of microscopic details as described above and treats
rapidly changing parameters such as noise. The equation is still exact if this
noise is generated by a Gaussian process, that is, if it is fully determined
by the first two moments. It gives the temporal evolution of the probability
density describing the positions of the agents.
Initially, the swarm designer must specify the functions A and 𝐵 of
(11), in accordance with the desired microscopic behaviour. Function A is
a direction and describes the deterministic motion based on information
provided by the environment and the information indirectly provided by other
robots via the environment. Function 𝐵 describes the random component of
the motion. A and 𝐵 are characterized by the underlying control algorithm.
 Modelling Oil-Spill Detection with Swarm Drones 245

𝐵 typically incorporates influences by other robots that disturb the robot,

for example, by the need of collision avoidance. A might incorporate an
external influence such as a light gradient.
The most difficult task is the definition of these functions. This design is
not unique and it requires finding two functions that correctly describe the
behaviour of the swarm. These functions will be analytically presented and
their performance will be tested in following sections.
A function determines the displacement of the swarm. A depends
primarily on a vector representing the directional information. A potential
field 𝑃 is commonly used to define it. In our case, we need to establish a
function that takes into account the following things based on the proposed
microscopic model: the random motion states of the robot, the probability
that a movement of an agent fails in its execution (e.g., due to a collision),
and the potential field where the robots move. Although it is possible to
model a probability distribution for each state, as our microscopic model
has no interaction between agents (except purely physical, as collisions) and
the behaviour of states is relatively simple, the macroscopic behaviour of the
swarm can be comprised in a single distribution:

(12)
where 𝑃 is directly related to the sensor readings in point r at time 𝑡 and
𝜇5 is a normalization term:

(13)
In (12), a function A that takes into account previous aspects is proposed.
When the density of agents increases, the probability of collision also
increases and, therefore, this situation reduces the rate at which robots are
directed by vector specified by A. More specifically, consider the following.
br is the probability that comprises the influences of all states that develop
a random or pseudorandom state. 𝑃(r, 𝑡) as commented before, is defined as
the probability of encountering a robot at position r at time 𝑡. Γ is a function
applied to the potential field that produces, for each position of 𝑃, the direction
to be taken by drones, as a single agent would do using the function 𝛾 In our
case, we use a “slidingneighbourhood” filter as commented, for example,
in [21] to perform the same calculation on 𝑃 as on the microscopic model,
changing the sign of the displacement when a ratio greater than 80% of
pollutant is detected. 𝐾 is a convolution operator. Being𝐺a square gaussian
246 Unmanned Aerial Vehicles (UAV) and Drones

kernel of size 𝑑, generated with 𝜎 = 4 ⋅ |𝑆|, and assuming two-dimensional

coordinates so that r = (𝑖, 𝑗), the sum of the components of this kernel is ∑𝑑
𝑖=1
∑𝑑 𝑗=1 (𝑖, 𝑗). 𝐹 is defined as this sum if it is different from 0; otherwise, it
will be defined as 1. In this way,

(14)
Function 𝐵 describes the nondeterministic motion and, therefore, it
takes into account the random motion of agents. Two forces, which must
be considered, take part in the microscopic behaviour. On the one hand,
some influences derived from agents that are on Wander and InResource
states. These states have a random motion, depending on the intensity of
parameters 𝜇1, 𝜇3. On the other hand, the behaviour itself causes that the
environment has areas with a higher density of agents. In these areas the
probability of collision can be increased depending on the density of agents
at a given time:

(15)
Thus, in (15) two terms can be observed: 𝑏𝑟 comprises the influences
of all states that develop a random or pseudorandom state, as previously
commented, and 𝜌(r, 𝑡)𝜇4 is a term that defines the connection between the
density of robots and their probability of collision.

Spill Detection
The operation of the microscopic behaviour has been analysed in the
previous section. It has been shown that the swarm is able to locate and
mark the perimeter of an oil spill with relatively simple rules. Several tests
have been developed to verify the validity of the presented model in various
cases.
In addition to the tests above, it is possible to establish, for a given
discharge, the areas of the map with the highest probability of containing
a robot, independently of the number of agents (on condition that you use
enough of them), by using the macroscopic model. This provides a more
accurate visualization of the behaviour for large swarms without being
limited by the number of agents to be simulated.
This section, using the previously presented macroscopic definition,
presents how the swarm is able to locate and mark, in a given time, an oil
spill:
 Modelling Oil-Spill Detection with Swarm Drones 247

(16)
The previous equation shows default parameters used in simulations
for the macroscopic model. These parameters have been adjusted taking
into account the microscopic behaviour experiments. We will present the
analysis of the macroscopic behaviour of the swarm for two time instants 𝑡
= (140 h, 168 h).
The simulation process is simple, once the functions A and 𝐵 of Fokker-
Planck equation have been defined. Initially, an origin point for the swarm
must be established. Although in microscopic simulations it is possible to
establish different origin points (as setting randomly the position of agents),
Fokker-Planck equation requires a single point. Nonetheless, several
simulations for different origin points have been performed, observing that
the results for large values of 𝑡 are very similar: only slight variations can
be seen if the origin point is located just on an oil slick. In this case, the
probability of this slick in relation to the rest of the spill can be higher.
The same origin point has been used for all tests, discretizing the
simulation area in 100 × 100 units. The origin point is established in Fokker-
Planck equation as (𝑥, 𝑦) = (40, 25).
Once the Fokker-Planck equation has been defined, the probability
distribution that a certain agent is in a position of the environment at a given
time can be obtained in an iterative way. This distribution can be calculated
iterating for each instant of time 𝑡 all the positions of the map, updating for
each position the new probability described by the equation.
The macroscopic state of the swarm is presented in Figure 10 for 𝑡 =
140 h. A clear convergence in the perimeter of the spill can be observed.
A three-dimensional representation of the probability distribution that an
agent is in a certain position of the environment at 𝑡 = 168 h is presented in
Figure 11. As it can be observed, the macroscopic model correctly predicts
the behaviour presented at the microscopic level of the swarm.
248 Unmanned Aerial Vehicles (UAV) and Drones

Figure 10: (a) Map (𝑀) generated from GNOME data for 𝑡 = 140 h. (b) Prob-
ability that a robot is in a certain position of the space at 𝑡 = 140 h. Sampled
using the macroscopic model of the swarm with the Fokker-Planck equation.

Figure 11: Three-dimensional representation of the probability that a robot is in

a particular position at 𝑡 = 168 h. Sampled using the macroscopic model of the
swarm with the Fokker-Planck equation. A map for the probability 0.002 where
the environment and the state of the spill at 𝑡 = 168 h are represented is added.

Model Comparison
In the previous section the behaviour of the macroscopic model and how
this model predicts the overall behaviour of the swarm have been presented.
Now, we will compare predictions of microscopic model and macroscopic
model for a specific case.
 Modelling Oil-Spill Detection with Swarm Drones 249

Figure 12(a) shows a probability distribution obtained from the microscopic

model. In order to do this, we have discretized the simulation environment
and, with the same parameters used in previous sections, we have developed
a simulation with 200 agents at𝑡 = 140 h. Along the simulation, we save the
number of agents that pass on a discrete cell (in order to compare it with the
macroscopic model that we have also discretized in 100 × 100 units) and
then we calculate the probability that an agent is in this cell at a given time.

Figure 12: Comparison of the microscopic prediction and macroscopic model

for the map 𝑀 at 𝑡 = 140 h. (a) (log(𝑃micro)) Logarithm of the probability
distribution obtained simulating 200 agents for 30,000 seconds. Logarithmic
distribution is used to highlight the states with low probability. (b) (𝑃macro) Prob-
ability distribution obtained with Fokker-Planck equation. (c) Product of both
distributions, decreasing the importance of high values in the microscopic sim-
ulation (specifically 𝑃macro × 𝑃1/2 micro).
250 Unmanned Aerial Vehicles (UAV) and Drones

In the same way, Figure 12(b) shows the probability distribution that predicts
the macroscopic model. In this figure it can be seen that the area of interest
is covered in both models. There are minor differences within the models,
due to the deficiencies of microscopic simulation that, among other things,
depends on the number of agents used in the simulation.
Nevertheless, we can compare both approaches by multiplying both
distributions. In this way only high probabilities remain and, therefore, it
is easier to observe if the areas of the spill are correctly identified in both
models. Bearing this in mind, we have slightly rectified the microscopic
distribution. When we use a limited number of agents in the simulation, we
discover, in some cases, that high probabilities hide important information
in the distribution.
In order to avoid this loss of information, we have used the square root
of the distribution in order to compare both models.
The product of macroscopic distribution and the square root of
microscopic distribution are presented in Figure 12(c). This figure shows
how the most important parts of the spill are detected with both distributions,
predicting the macroscopic model the same perimeter areas detected by the
microscopic model.

DISCUSSION
This paper describes a microscopic model that is able to locate and mark
the perimeter of an oil spill. The microscopic behaviour presented does not
require direct communication between agents. This limitation can cause that
the convergence of the swarm on the spill takes more time, depending on the
number of drones and the size of the spill.
However, this behaviour is versatile and easy to implement and develop,
even in areas without GPS coverage. It is important to highlight that a swarm
system, which requires direct communication between agents, is a limited
system because of the maximum range of each agent and the saturation of
the radiofrequency space if the system needs a large number of agents.
Moreover, we have demonstrated that the process of locating and marking
the perimeter of the spill without communication is robust and efficient. We
have shown that the swarm system is able to completely delimit the spill
if the number of agents is sufficient. In order to achieve this task, an agent
must be able to detect drones that are nearby. There are several ways, as, for
example, using a camera or transmitting the GPS position.
 Modelling Oil-Spill Detection with Swarm Drones 251

We propose the use of signal intensity (at a given frequency) for obstacle
avoidance tasks. This strategy may show some problems (we have
implemented it by using a reactive behaviour); however, it has several
advantages. Many countries require that drones broadcast a continuous
signal indicating their position. Europe uses 433 MHz frequency for this
purpose. The intensity of the signal in a particular area can be detected by
using the same infrastructure. If the intensity of the signal grows with the
movement of the agent, this agent must change its direction. We emphasize
that, as a swarm approach, this is not a communication between agents but
simply a beacon that we can use, if necessary, to know the position of drones.
The proposed macroscopic model demonstrates that the tendency of the
swarm, for a sufficient number of drones, is the same that can be perceived in
the microscopic model. The connection of both models has been tested for a
complex spill, generated with GNOME. These experiments have shown that
the fundamental characteristics of the behaviour (detection and monitoring)
are reflected in both models. It is advisable not to forget the differences
between the two models.
The microscopic model defines the individual behaviour. Because of this
it is easy to understand at local level. However, this model does not define
the behaviour of all the swarm. In order to analyse the global behaviour, a
set of tests can be defined for a large number of agents. However, these tests
can be expensive and difficult and are not exempt from problems.
The macroscopic model defines the global behaviour of the swarm. It
allows us to verify the emergent behaviour from the interaction between all
agents that run the microscopic model. The macroscopic model demonstrates
the tendency of the swarm for a large number of agents. The analysis of
this model is complex, because of the use of differential equations that, for
example, force us to choose a single point to start the simulation. Even so, this
model has remarkable advantages [20], for example, continuous analysis for
any point of the environment, time of the probability that an agent is located
in a given location, and simulation time negligible compared to microscopic
model.
We are currently working on the implementation of this system in a real
swarm of drones. Our immediate future research focuses on this real swarm,
since it allows us to adjust the algorithms for a real system.
We are already in the testing phase for small swarms (5 drones), obtaining
satisfactory results in our preliminary tests. We are using low-cost, custom-
developed hexacopters to test this behaviour. The low computational needs
252 Unmanned Aerial Vehicles (UAV) and Drones

required by this behaviour make it possible to use cheap Arduino control

boards to control the UAVs. We have developed several tests for our swarm
behaviors using artificial color landmarks. While flying our agents can
localize (by wandering) a landmark and follow it (if moving) in real time.

CONFLICT OF INTERESTS
The authors declare that there is no conflict of interests regarding the
publication of this paper.

ACKNOWLEDGMENT
This work has been supported by the Spanish Ministerio de Ciencia e
Innovación, Project TIN2009-10581.
 Modelling Oil-Spill Detection with Swarm Drones 253

REFERENCES
1. J. I. Medina-Bellver, P. Marín, A. Delgado et al., “Evidence for in situ
crude oil biodegradation after the prestige oil spill,” Environmental
Microbiology, vol. 7, no. 6, pp. 773–779, 2005.
2. S. Díez, J. Sabaté, M. Viñas, J. M. Bayona, A. M. Solanas, and J.
Albaigés, “The prestige oil spill. I. Biodegradation of a heavy fuel oil
under simulated conditions,” Environmental Toxicology and Chemistry,
vol. 24, no. 9, pp. 2203–2217, 2005.
3. J. J. González, L. Viñas, M. A. Franco et al., “Spatial and temporal
distribution of dissolved/dispersed aromatic hydrocarbons in seawater
in the area affected by the prestige oil spill,” Marine Pollution Bulletin,
vol. 53, no. 5–7, pp. 250–259, 2006.
4. F. Nirchio, M. Sorgente, A. Giancaspro et al., “Automatic detection of
oil spills from sar images,” International Journal of Remote Sensing,
vol. 26, no. 6, pp. 1157–1174, 2005.
5. M. N. Jha, J. Levy, and Y. Gao, “Advances in remote sensing for oil
spill disaster management: state-of-the-art sensors technology for oil
spill surveillance,” Sensors, vol. 8, no. 1, pp. 236–255, 2008.
6. C. Brekke and A. H. S. Solberg, “Oil spill detection by satellite remote
sensing,” Remote Sensing of Environment, vol. 95, no. 1, pp. 1–13,
2005.
7. M. Dorigo, E. Tuci, R. Groß et al., “The swarm-bot project,” in
Kunstliche Intelligenz, pp. 31–44, Springer, Berlin, Germany, 2005.
8. M. Dorigo and E. Şahin, “Autonomous Robots: guest editorial,”
Autonomous Robots, vol. 17, no. 2-3, pp. 111–113, 2004.
9. G. Beni, “From swarm intelligence to swarm robotics,” in Swarm
Robotics, pp. 1–9, Springer, Berlin, Germany, 2005.
10. E. Şahin, “Swarm robotics: from sources of inspiration to domains
of application,” in Swarm Robotics, vol. 3342 of Lecture Notes in
Computer Science, pp. 10–20, 2005.
11. O. Soysal and E. Şahin, “A macroscopic model for self-organized
aggregation in swarm robotic systems,” in Swarm Robotics, pp. 27–42,
Springer, Berlin, Germany, 2007.
12. W. Liu, A. F. T. Winfield, and J. Sa, “Modelling swarm robotic systems:
a case study in collective foraging,” in Proceedings of the Towards
Autonomous Robotic Systems (TAROS ‘07), pp. 25–32, 2007.
254 Unmanned Aerial Vehicles (UAV) and Drones

13. United States. National Ocean Service. Office of Response and

Restoration, General Noaa Operational Modeling Environment
(Gnome) Technical Documentation, U.S. Dept. of Commerce, National
Oceanic and Atmospheric Administration, National Ocean Service,
Office of Response & Restoration, 2012.
14. K. Wojtaszek, Application of transport model for building contingency
maps of oil spills on the North Sea [M.S. thesis], Delft University
of Technology, 2003, A-1, http://risk2.ewi.tudelft.nl/research-and-
publications/doc_download/70-katarzynawojtaszekthesispdf.
15. P. D. Boehm, D. L. Flest, D. Mackay, and S. Paterson, “Physical-
chemical weathering of petroleum hydrocarbons from the ixtoc
I blowout: chemical measurements and a weathering model,”
Environmental Science and Technology, vol. 16, no. 8, pp. 498–505,
1982.
16. K. Karantzalos and D. Argialas, “Automatic detection and tracking of
oil spills in SAR imagery with level set segmentation,” International
Journal of Remote Sensing, vol. 29, no. 21, pp. 6281–6296, 2008.
17. S. Luke, C. Cioffi-Revilla, L. Panait, and K. Sullivan, “Mason: a new
multi-agent simulation toolkit,” in Proceedings of the Swarm Fest
Workshop, vol. 8, 2004.
18. L. Bayindir and E. Şahin, “A review of studies in swarm robotics,”
Turkish Journal of Electrical Engineering and Computer Sciences,
vol. 15, no. 2, pp. 115–147, 2007.
19. H. Hamann and H. Wörn, “A framework of space-time continuous
models for algorithm design in swarm robotics,” Swarm Intelligence,
vol. 2, no. 2–4, pp. 209–239, 2008.
20. H. Hamann, Space-Time Continuous Models of Swarm Robotic
Systems: Supporting Global-to-Local Programming, vol. 9, Springer,
Berlin, Germany, 2010.
21. J. L. Semmlow, Biosignal and Medical Image Processing, Signal
Processing and Communications, Taylor & Francis, 2004.
 CHAPTER

12
A Survey of Modelling and
Identification of Quadrotor Robot

Xiaodong Zhang,1 Xiaoli Li,2 Kang Wang,2 and Yanjun Lu1

School of Automation, Shenyang Aerospace University, Shenyang 110136,
1

China
2
School of Automation and Electrical Engineering, University of Science
and Technology Beijing, Beijing 100083, China

ABSTRACT
A quadrotor is a rotorcraft capable of hover, forward flight, and VTOL and
is emerging as a fundamental research and application platform at present
with flexibility, adaptability, and ease of construction. Since a quadrotor is
basically considered an unstable system with the characteristics of dynamics

Citation: Xiaodong Zhang, Xiaoli Li, Kang Wang, and Yanjun Lu, “A Survey of
Modelling and Identification of Quadrotor Robot”, Hindawi Journal on Abstract
and Applied Analysis, Volume 2014 |Article ID 320526 | 16 pages | https://doi.
org/10.1155/2014/320526.
Copyright: © 2014 Xiaodong Zhang et al. This is an open access article distributed
under the Creative Commons Attribution License.
256 Unmanned Aerial Vehicles (UAV) and Drones

such as being intensively nonlinear, multivariable, strongly coupled, and

underactuated, a precise and practical model is critical to control the vehicle
which seems to be simple to operate. As a rotorcraft, the dynamics of a
quadrotor is mainly dominated by the complicated aerodynamic effects of the
rotors. This paper gives a tutorial of the platform configuration, methodology
of modeling, comprehensive nonlinear model, the aerodynamic effects, and
model identification for a quadrotor.

INTRODUCTION
A quadrotor is agile to attain the full range of motion propelled by four
rotors symmetrically across its center with smaller dimension and simple
fabrication, unlike a conventional helicopter with complicated mechanism.
Generally, it should be classified as a rotary-wing aircraft according to
its capability of hover, horizontal flight, and vertical take-off and landing
(VTOL) [1]. In 1920s, the prototypes of manned quadrotors were introduced
for the first time [1, 2]; however, the development of this new type of air
vehicle is interrupted for several decades due to various reasons such as
mechanical complexity, large size and weight, and difficulties in control
especially. Only in recent years a great deal of interests and efforts have
been attracted on it; a quadrotor has even become a more optional vehicle
for practical application, such as search-and-rescue and emergency response
amazingly. As a small, unmanned aerial vehicle (UAV), it has versatile
forms from 0.3 to 4 kg. Up to now, some large quadrotors already have
sufficient payload and flight endurance to undertake a number of indoor
and outdoor applications, like Bell Boeing Quad TiltRotor and so forth
[3]. With the improvements of high energy lithium battery, MEMS sensor
and other technologies, especially, the scope for commercial opportunities
is rapidly increasing [4]. As a quadrotor is inexpensive and easy to be
designed and assembled, as well as the complex dynamics, such a rotorcraft
is emerging as a fundamental research platform for aerial robotics research
for the problems related to three-dimensional mobility and perception [5].
Furthermore, a quadrotor’s design and implementation have even become a
Multidisciplinary Engineering Undergraduate Course nowadays for the aim
to teach students to cope with the challenges, for instance, fast and unstable
dynamics and tight integration of control electronics and sensors [6].
For the specific purposes including academic research, commercial
usage, and even military aim, many research groups or institutions have
fabricated various quadrotors, such as the X4-flyer [7], OS4 [8], STARMAC
 A Survey of Modelling and Identification of Quadrotor Robot 257

[9], and Pixhawk [10] which have become the shining stars mentioned on
the network, magazines, and all kinds of academic journals. It is worthy to
note that the Draganflyer X4, Asctec Hummingbird, Gaui Quad flyer, and
DJI Wookong have been introduced and developed in the comprehensive
commercial market. For the powerful operation, some new types of quadrotors
with tilting propellers or a new configuration, have been constructed in [10–
15] in order to address the issues such as underactuated system. In addition,
a number of OSPs (open- source projects) for quadrotors have emerged with
contributions from RC hobbyists, universities, and corporations [16].
A quadrotor helicopter is a highly nonlinear, multivariable, strongly
coupled, underactuated, and basically an unstable system (6 DOF with only
4 actuators), which acts as preliminary foundation for design of control
strategy. Many controllers have been presented to overcome the complexity
of the control resulting from the variable nature of the aerodynamic forces
in different conditions of flight [17]. Many works have been published on
control issues about quadrotors, such as PID controllers [18–21], linear
quadratic LQR algorithm [22, 23], 𝐻∞ loop forming method [24], sliding
mode variable structure control, feedback linearization [25, 26], backstepping
[27, 28], and even intelligent control [29, 30]. In those works above, the
linearization of the nonlinear model around hover fight regime is conducted
and used to construct controller to stabilize the quadrotor’s attitude under
small roll and pitch angles. The treatments to the vehicle dynamics, based
on some simplistic assumptions, have often ignored known aerodynamic
effects of rotorcraft vehicles. In the case of hovering and forward flight with
slow velocity, those assumptions are approximately reasonable.
As the quadrotor research shifts to new research areas (i.e., mobile
manipulation, aerobatic moves, etc.) [31, 32], the need for an elaborate
mathematical model arises, and the simplistic assumption is no more suitable.
When aggressive maneuvers such as fast forward and heave flight actions,
VTOL, and the ground effect appear, the dynamics of quadrotors could
be influenced significantly under these aerodynamic force and moment.
It is shown in [33] that existing techniques of modeling and control are
inadequate for accurate trajectory tracking at higher speed and in uncertain
environments if aerodynamic influence is ignored. The model incorporated
with a full spectrum of aerodynamic effects that impact on the quadrotor in
faster climb, heave, and forward flight has become an area of active research
with considerable effort focusing on strategies for generating sequences of
controllers to stabilize the robot to a desired state.
258 Unmanned Aerial Vehicles (UAV) and Drones

Traditionally, first principle assumptions and measurements of vehicle

parameters can be used to derive nonlinear models, from which linear models
can be obtained. As an alternative, system identification is also powerful
to derive dynamic models directly from flight test data to overcome the
challenges such as hardly obtained parameters of the underlying physics of
the vehicle. Even though there exists a large volume of multirotor research,
there is very little research into system ID of multirotors [34]. The reason
for this absence is partially due to the unstable system dynamics of the
quadrotor, which makes open-loop identification nonpractical.
This paper provides a tutorial introduction to configuration, modeling,
aerodynamics effects analysis, and model identification for quadrotor.
The paper’s basic structure is as follows: above all, the characteristics
and configuration of quadrotor are introduced, then two formulations for
the model of quadrotor are compared, and a comprehensive nonlinear
equation characterizing the dynamics of quadrotor is derived. Thereupon,
aerodynamic effects that impact on the quadrotor in aggressive maneuvers
are revealed. At last, several methods about identification are reviewed.

CHARACTERISTICS OF QUADROTOR
Typically, the structure of a quadrotor is simple enough, which comprises
four rotors attached at the ends of arms under a symmetric frame. The
dominating forces and moments acting on the quadrotor are given by rotors
driven with motors, especially BLDC motors. According to the orientation
of the blades, relative to the body coordinate system, there are two basic
types of quadrotor configurations: plus and cross-configurations shown in
Figure 1.

Figure 1: Plus and X quadrotor configurations.

 A Survey of Modelling and Identification of Quadrotor Robot 259

In the plus configuration selected by most of the quadrotors, a pair of

blades, spinning in the same clockwise or counter-clockwise direction, are
fabricated on x and y coordinates of the body frame coordinate system,
such as the assemble of the Draganflyer XPro. On the contrary, a different
cross-configuration is adopted by some other quadrotors, such as the
Convertawings model A, the Piasecki PA-39, or the Curtiss-Wright VZ-
7AP, in which there is no rotor at the front or the rear but instead two rotors
are on the right side and two on the left.
In contrast with the plus configuration, for the same desired motion,
the cross-style provides higher momentum which can increase the
maneuverability performance as each move requires all four blades to vary
their rotation speed [35]. However, the attitude control is basically analogous
[36].
It is the configuration of a quadrotor that shows the inherent
characteristics. Basic control sequences of cross-configuration are shown
in Figure 2. The quadrotor’s translational motion depends on the tilting
of rotorcraft platform towards the desired orientation. Hence, it should be
noted that the translational and rotational motion are tightly coupled because
the change of rotating speed of one rotor causes a motion in three degrees
of freedom. This is the reason that allows the quadrotor with six degrees of
freedom (DOF) to be controlled by four inputs; therefore the quadrotor is an
underactuated system.

Figure 2: Quadrotor dynamics: (a) and (b) difference in torque to manipulate

the yaw angle (Ψ); (c) and (d) hovering motion and vertical propulsion due to
balanced torques; (e) and (f) difference in thrust to manipulate the pitch angle
(𝜃) and the roll angle (𝜙).
260 Unmanned Aerial Vehicles (UAV) and Drones

In principle, a quadrotor is dynamically unstable and therefore proper control

is necessary to make it stable. Despite of unstable dynamics, it is good for
agility. The instability comes from the changing rotorcraft parameters and
the environmental disturbances such as wind [37]. In addition, the lack of
damping and the cross-coupling between degrees of freedom make it very
sensitive to disturbances.

6-DOF AIRFRAME DYNAMICS

Dominating methods as Euler-Lagrange formalism and Newton-Euler
formalism are applied to model the dynamics for an aircraft [38–44]. It
has been noted that the Newton-Euler method is easy to be understood and
accepted physically despite of the compact formulation and generalization
shown by Euler-Lagrange formalism. Nevertheless, two methods are
consistent for the description of dynamics. That is to say, it is indicated that
after a speed transform matrix the Lagrange equation is an expression form
of the second Newton Law [45].

Euler-Lagrange Formalism
The generalized coordinates of the rotorcraft are given in [46]:

(1)
where (𝑥, 𝑦, 𝑧) = 𝜉 ∈ 𝑅 denotes the position of the mass center of the
3

quadrotor relative to the inertial frame and (𝜓, 𝜃, 𝜙) = 𝜂 ∈ 𝑅3 are the three
Euler angles (resp., yaw, pitch, and roll), under the conditions (−𝜋 ≤ 𝜓 ≤
𝜋) for yaw, (−𝜋/2 ≤ 𝜃 ≤ 𝜋/2) for pitch, and (−𝜋/2 ≤ 𝜙 ≤ 𝜋/2) for roll, which
represent the orientation of the rotorcraft (see Figure 3).

Figure 3: Quadrotor’s body-fixed and inertial coordinate systems.

 A Survey of Modelling and Identification of Quadrotor Robot 261

Naturally, translational and rotational coordinates are obtained from the

model

(2)
The translational and the rotational kinetic energy of the rotorcraft are

(3)
where 𝑚 denotes the mass of the quadrotor. 𝐽 = W 𝐼W is the moment of
𝑇

inertia matrix in the inertial coordinate system after being transformed from
the body frame, by matrix W:

(4)
The only potential energy to be considered is the gravitational potential
given by 𝑈 = 𝑚𝑔z𝐸.
The Lagrangian of the rotorcraft is

(5)
The full rotorcraft dynamics model is derived from the Euler-Lagrange
equations under external generalized forces:

(6)
where 𝐹𝜉 = 𝑅𝐹̂ is the translational force applied to the quadrotor due to
the throttle control input, 𝜏∈𝑅3 represents the pitch, roll, and yaw moments
and 𝑅 denotes the rotational matrix 𝑅(𝜓, 𝜃, 𝜙) ∈ SO(3), which represents
the orientation of the rotorcraft relative to a fixed inertial frame.
Since the Lagrangian contains no cross-terms in the kinetic energy
combining ̇ 𝜉 and 𝜂̇, the Euler-Lagrange equation partitions into two parts.
One obtains
262 Unmanned Aerial Vehicles (UAV) and Drones

(7)

(8)
Rewrite (8) as

(9)
where 𝐶(𝜂, 𝜂)̇ is referred to as the Coriolis terms and contains the
gyroscopic and centrifugal terms.

Newton-Euler Formalism
Typically, it is necessary to define two frames of reference, each with its
defined righthanded coordinate system, as shown in Figure 3. 𝑋, 𝑌, and 𝑍
are orthogonal axes of the body-fixed frame with its correspondent body
linear velocity vector 𝑉⃗ = [𝑢V 𝜔]𝑇 and angular rate vector Ω⃗ = [𝑝 𝑞 𝑟] 𝑇.
Another one is an Earth-fixed inertial (also known as navigation) coordinate
system 𝐸 = (𝑋𝐸, 𝑌𝐸, 𝑍𝐸) with which initially the bodyfixed coincides. The
attitude of the quadrotor, expressed in terms of the Euler angles 𝜙 (roll),
𝜃 (pitch), and 𝜓 (yaw), is evaluated via sequent rotations around each one
of the inertial axes. Herein, a reference frame by 𝑂NED (North-EastDown)
denotes an inertial reference frame and 𝑂𝐵 a bodyfixed reference frame.
Generally, a quadrotor is considered as a rigid body in a three-dimensional
space. The motion equations of a quadrotor subject to external force 𝐹∈𝑅3
and torque 𝜏 ∈ 𝑅3 are given by the following Newton-Euler equations with
respect to the body coordinate frame 𝐵 = (𝑋𝐵, 𝑌𝐵, 𝑍𝐵):

(10)
The rotorcraft orientation in space is presented by a rotation 𝑅 from 𝐵
to 𝐸, where 𝑅 ∈ SO3 is the rotation matrix. Here 𝑐𝜃 is for cos(𝜃) and 𝑠𝜃 is
for sin(𝜃):

(11)
With the transformation R, the first equation assessing the translational
 A Survey of Modelling and Identification of Quadrotor Robot 263

dynamics in (11) can be written in E:

(12)
Recall the kinematic relationship between the generalized velocities
and the angular velocity Ω=𝑊𝜂,̇ 𝑊∈𝑅3×3.
Defining a pseudoinertia matrix (𝜂) = 𝐽𝑊and a Coriolis vector 𝐶(𝜂, 𝜂) =
𝐼 ̇ 𝜂+𝑊̇ 𝜂×𝐼 ̇ 𝜂, one can obtain

(13)
This model has the same structure as the one obtained by the Euler-
Lagrange approach, in which the main difference is the expressions of 𝐼 and
𝐶, which are more complex and more difficult to implement and to compute
in the case of the Euler-Lagrange method. It is important to note that the
model (13) is common for all aerial robots with six degrees of freedom.

BASIC DYNAMIC MODEL OF A QUADROTOR

This section introduces the basic quadrotor dynamic modeling with rigid
body dynamics and kinematics. This model, based on the first order
approximation, has been successfully utilized in various quadrotor control
designs so far.
In the first place, some assumptions are reasonable and essential shown
as follows [44].
i The structure is supposedly rigid.
ii The structure is supposedly symmetrical.
iii The CoG (center of gravity) and the body fixed frame origin are
assumed to coincide.

Dynamic Model of a Quadrotor

As we know, Newton second law is applied to the translational motion in
inertial frames [47]. From the equation of Coriolis, one can obtain

(14)
where 𝑚 is the mass of the quadrotor, f ≜ (𝑓𝑥 𝑓𝑦 𝑓𝑧) 𝑇 is the total
𝑏

force applied to the quadrotor, and k is the translational velocity. 𝜔𝑏/𝑖 is

264 Unmanned Aerial Vehicles (UAV) and Drones

the angular velocity of the airframe with respect to the inertial frame. Since
the control force is computed and applied in the body coordinate system,
and since 𝜔 is measured in body coordinates, (14) is expressed in body
coordinates, where k𝑏 ≜ (𝑢, V, 𝜔) and w𝑏 𝑏/𝑖 ≜ (𝑝, 𝑞, 𝑟)𝑇.
For rotational motion, Newton’s second law state is

(15)
where h is the angular momentum and m is the applied torque. h𝑏 = Jw𝑏
𝑏/𝑖
; J is the constant inertia matrix. The quadrotor is essentially symmetric
about all three axes, which implies that J = diag(𝐽𝑥, 𝐽𝑦, 𝐽𝑧).
Given m𝑏 ≜ (𝜏𝜙, 𝜏𝜃, 𝜏Ψ) 𝑇, which denote the rolling torque, the pitching
torque, and the total yawing torque, are induced by the rotor thrust and rotor
drag acting on the airframe.
The six-freedom-degree model for the quadrotor kinematics and
dynamics can be summarized as follows:

(16)
Equation (16) is a full nonlinear model for a quadrotor, in which the
complex dynamics is shown obviously, such as strong nonlinearity like the
multiplication between system states, intensive coupling among the variables,
and the multivariable features intuitively, that imposes the difficulties on the
 A Survey of Modelling and Identification of Quadrotor Robot 265

controller design and, on the other hand, attracts great interest of research.

Forces and Moments

The forces and torques that act on the quadrotor are primarily due to
gravity and the four propellers shown in Figure 3. The steady-state thrust 𝐹𝑖
generated by a hovering rotor (i.e., a rotor that is not translating horizontally
or vertically) in free air coincides with −𝑍𝐵 axis. The total force acting on
the quadrotor is given by

(17)
The rolling torque, the pitching torque, and the total yawing torque are
given by

(18)
The gravity force acting on the center of mass is given by

(19)
Equation (16) shows strong coupled dynamics [48]: the speed change of
one rotor gives rise to motion in at least 3 degrees of freedom.
For instance, the speed decrease of the right rotor will roll the craft to
the right under the imbalance between left and right lift forces, coupled with
the rotorcraft’s yaw to the right due to the imbalance in torque between
clockwise and counter-clockwise, so the translation changes direction
toward the front.
Nevertheless, in some cases that the rotating movement is slight,
the Coriolis terms 𝑞𝑟, 𝑝𝑟, and 𝑝𝑞 are small and can be neglected. So the
dynamics of the quadrotor is simplified and given as [47]
266 Unmanned Aerial Vehicles (UAV) and Drones

(20)

This model is shown in Figure 4 to which two diagrams in [49, 50] are
similar. Note that the attitude of quadrotor is changed, subject to the input 𝜏
(moment) produced by each rotor. However, the position/altitude dynamics
block is affected by 𝑇𝑗 and angle variables.

Figure 4: Simplified block diagram of the quadrotor’s dynamics.

Intuitively, Figure 4 gives the insight of the dynamic of the quadrotor
that the angles and their time derivatives do not depend on translation
components, whereas the translations depend on angle (and not on angular
velocities) [50]. Based on the characteristics of the dynamics, a quadrotor
control problem can be split into two distinct control problems, the inner
attitude/altitude loop designed for stability and tracking of desired Euler
angles and the outer 𝑋, 𝑌, and 𝑍 position loops for regulating the vehicle
position [27].
 A Survey of Modelling and Identification of Quadrotor Robot 267

State space equations are applied in the control design and system
identification generally. Hence, the nonlinear system of a quadrotor is
illustrated as the formulation, which is described in different manner in [3,
27, 51]:
(21)
where

(22)
Herein the output 𝑦 is composed of 𝑥, 𝑦, 𝑧 and 𝜓 is for the trajectory
track, but if for the hovering control, 𝑦 = [𝜙, 𝜃, 𝜓, 𝑍]𝑇 should be selected
because in translation movement shown in (22), the three state variables, 𝑥,
𝑦, and 𝑧, are subordinated to the same control parameter 𝐹; hence only one
state is controllable and the others are subjected to the controlled translation
and angular motions.

Gyroscopic Torques
At the normal attitude, namely, Euler angles are zero and the axes of the
rotors with higher speeds spinning are coincident with the 𝑧𝐵 axis of the
robot frame. However, while the quadrotor rolls or pitches, the direction of
the angular momentum vectors of the four motors is forced to be changed.
A gyroscopic torque will be imposed on the airframe that attempts to turn
the spinning axis so that it aligns with the precession axis. It is noted that no
gyroscopic torque occurs with rotation around the 𝑧𝐵 axis (yaw) because the
spin and precession axes are already parallel [52].
The gyroscopic (inertial) moment is modeled in [53] as

(23)
where 𝐽⃗ 𝑟 is the gyroscopic inertia, namely, that of the rotating part of
the rotor and Ω𝑖 is the angular rate of the rotor 𝑖 (𝑖 = 1, 2, 3, 4).
268 Unmanned Aerial Vehicles (UAV) and Drones

Quaternion Differential Equations

A problem, so called gimbal lock, will appear with the Euler angle 𝜃 close
to 𝜃 = 90∘ , and then the Roll angle 𝜙 loses its meaning. To overcome this
problem, the quaternion method, which offers a mathematical notation
that allows the representation of three-dimensional rotations of objects in
4D space, is selected to be the alternative remedy. Reference [54] gives a
quaternion dynamics description and [55] proposes a new quaternion-based
feedback control scheme for the attitude stabilization of a quadrotor aircraft.
In fact, every parameterization fails to fully represent rigid body pose in
every case. That is to say, Euler angles cannot globally represent rigid body
pose owing to the gimbal lock, whereas quaternions cannot define it uniquely
[56]. Although researchers proved the effectiveness of using quaternions to
describe aircraft dynamics, Euler angles are still the most common way of
representing rigid body pose.

Linearized Model
The full nonlinear model is very useful, as it provides insight into the behavior
of the vehicle. However, a linear model is used widely, which attributes to
the abundance of well-studied tools available for the control system design.
As we can see, most of the controllers are based on the nonlinear model with
hover conditions and are stable and effective only around reasonably small
angles.
Typically, the linearization of a nonlinear state space model 𝑥 = (𝑥) +
𝑔(𝑥)𝑈 ̇ is executed at an equilibrium point of the model

(24)
Then, the linear model is derived by

(25)
As the hovering is one of the most important regimes for a quadrotor, at
this point, the condition of equilibrium of the quadrotor in terms of (24)-(25)
is given as in [54]:

(26)
 A Survey of Modelling and Identification of Quadrotor Robot 269

While hovering, some assumptions are reasonable, such as the negligible

influence of external forces and moments on the aircraft and small
velocities and rotational velocities. By performing a Taylor series expansion
and eliminating the higher order terms on (20), and using small angle
approximations, a linear model is given [38, 57–59]:

(27)

Then the state vector is 𝑥 = [𝑢, ⃗ V, 𝑤, 𝑝, 𝑞, 𝑟, 𝜙, 𝜃, 𝜓]. Here, the

deviations from the trim value act as the states to be considered, and all
further references to aircraft states are understood to refer to the perturbation
states.
Distinctly different to the linearization at hovering regime, [60] presents
a new scheme, which has never before been considered in quadrotor control,
in which the linearizations at four points of equilibria are conducted. The
four linearizations represent different operating modes in a quadrotor flight
mission. These operating modes are
i hover;
ii vertical motion with a constant velocity;
iii horizontal translation with a constant pitch angle tilt;
iv horizontal translation with a constant roll angle tilt.
All four linearizations produce four linear time-invariant systems and
four controllers accordingly that are simple, low-order, and decentralized
270 Unmanned Aerial Vehicles (UAV) and Drones

and have integral-action designed for the system stabilization despite of the
issues of controller switch between two linear systems.

AERODYNAMIC EFFECTS
In most of research projects, quadrotor dynamics has often ignored known
aerodynamic effects of rotorcraft vehicles because only the stability while
hovering is the aim, as stated before. At slow velocities, such as while
hovering, this is indeed a reasonable assumption [61]. However, in case
of demanding flight trajectories, such as fast forward and descent flight
manoeuvres, as well as in the presence of the In Ground Effect, these
aerodynamic phenomena could significantly influence quadrotor’s dynamics
[33], and the performance of control will be diminished if aerodynamic
effects are not considered [62], especially in situations where the aircraft is
operating at its limits (i.e., carrying heavy load, single engine breakdown,
etc.).
Acting as a propulsion system, the aerodynamics of rotors plays the most
important role on the movement of the quadrotor excepted with gravity and
air drag with respect to the airframe. The kinematics and dynamics of the
rotors are fairly complex, resulting from the combination of several types of
motion, such as rotation, flapping, feathering, and lagging; normally the last
two items are neglectable [3, 4]. The theoretical models based on the blade
element theory (BET) combined with momentum theory (MT) show many
advantages such as more flexible, more simple, and convenient in contrast
with the empirical models based on empirical data typically obtained in the
wind tunnel.
Note that the application of helicopter theory to a quadrotor is not
straightforward for the reason of many important differences between
conventional helicopter and quadrotor [1]. In order to address the issues,
the specific research, with the aim at a quadrotor vehicle, is necessary to
establish full model with complex dynamics subject to aerodynamic forces
and moments. Many works [33, 40, 62–66] on rotor model have been done
based on the results obtained for conventional helicopters [67].
Blade flapping is of significant importance in understanding the natural
stability of quadrotors [4]. Since it induces forces in the x-y rotor plane
of the quadrotor, the underactuated directions in the dynamics, high gain
control cannot easily be implemented against the induced forces. On the
other hand, the total thrust variation owing to the vertical maneuver also
imposes nonignorable influence on the quadrotor behavior.
 A Survey of Modelling and Identification of Quadrotor Robot 271

Total Thrust
In case of simplifications in aerodynamic effects, the assumption that a
rotor’s thrust is proportional to the square of its angular velocity is the most
common consideration. However, it is proved that this assumption about
rotor’s thrust is especially far from reality in the cases of nonhovering
regime.
The helicopter literatures [67–69] give analysis about many effects on
the total thrust in more detail, in which translation lift and change of angle of
attack act as the two related effects. As a rotorcraft flighs across translation,
the momentum of the airstream induces an increase in lift force, which is
known as translational lift. The angle of attack (AOA) of the rotor with
respect to the free-stream also influences the lift, with an increase in AOA
increasing thrust, just like in aircraft wings.
Applying blade element theory to quadrotor construction, the expression
for rotor thrust 𝑇 is given in [56]:

(28)
and a thrust coefficient 𝐶𝑇 is given in

(29)
where 𝜇, 𝜆𝑖, and 𝜆𝑐 are speed coefficients 𝑉𝑥𝑦/𝑅Ω, 𝑉𝑧/𝑅Ω, and V𝑖/𝑅Ω,
respectively. 𝑉𝑥𝑦 and 𝑉𝑧 are the horizontal and vertical components of the
total air stream velocity, respectively, and V𝑖 is the induced velocity, and Ω
is rotor angular speed. Herein, the other parameters and coefficients in the
formulation above will not be described and refer to the literature [56].
Especially at hovering regime, 𝜇=0 and 𝜆𝑐 = 0 (i.e., static conditions)
yield

(30)
So one can obtain 𝑇⊂Ω just like the relationship between 𝑇 and Ω used
2

in the upwards context.

For the calculation of the aerodynamic coefficient 𝐶𝑇 it is crucial to
know three airspeed coefficients 𝜇, 𝜆𝑐, and 𝜆𝑖. Two of them, 𝜇, 𝜆𝑐, can easily
be obtained from the available motion data 𝑉𝑥𝑦, 𝑉𝑧, and Ω𝑅. 𝜆𝑖 however is
very hard to know, because it is impossible to measure the induced velocity
V𝑖.
272 Unmanned Aerial Vehicles (UAV) and Drones

One can solve this problem by means of calculating the induced velocity
coefficient 𝜆𝑖 involved in the two aerodynamic principals, momentum, and
blade element theories. In view of the fact that the macroscopic momentum
equation and the microscopic blade element equation give the same rotor
thrust formulation:

(31)

Figure 5: 𝑇/𝑇0 ratio during 𝑥-𝑦 plane movement [61].

The results of solving this equation can be shown in Figure 5.
The induced velocity decreases with an increase of airflow produced
by quadrotor movement, which can be seen in Figure 5. Although both
direction’s movements in 𝑥-𝑦 plane tend to increase induced velocity, only
the vertical movement decreases the thrust coefficient. As a result, during
takeoff the quadrotor loses rotor thrust, but during horizontal movement that
same thrust is increased and enables more aggressive maneuvers [62, 70].

Blade Flapping
When a rotor translates horizontally through the air, the advancing blade of
the rotor has a higher velocity relative to the free-stream and will generate
more lift than the retreating blade which sees a lower effective airspeed.
This causes an imbalance in lift, inducing an up and down oscillation of the
rotor blades, which is known as blade flapping [63, 64]. When a quadrotor
is in steady state suffering the blade flapping, its rotor plane will tilt at some
angle off of vertical, causing a deflection of the thrust vector illustrated in
Figure 6.
 A Survey of Modelling and Identification of Quadrotor Robot 273

Figure 6: Effect of blade flapping [33].

In fact, the rotor thrust is perpendicular to the rotor plane and not to the hub
of the rotor. Thus, in the case of blade flapping the rotor disk tilts, the rotor
thrust is also inclined with respect to the airframe and imposes a component
in the x and y directions of the body-fixed frame. As a result, the flapping of
the blades results in a variety of effects on the dynamics of the vehicle, and
in particular affecting attitude control performance [65], owing to the reason
that a moment is produced for the rotor plane aligned with the vehicle’s
center of gravity, and the effect increases with speed.
Note that the lateral flapping was neglected in above considerations, as
due to quadrotor’s symmetry and in-pair counter rotation of rotors; its net
influence is negligibly small in all instances of forward flight. Hence, there
are two moments that need to be considered. First of all, the moment 𝑀𝑏,lon
is caused by the longitudinal thrust due to a deflection angle 𝛼1𝑆 between the
rotor plane and rotorcraft platform:

(32)
where ℎ is the vertical distance from the rotor plane to the center of
gravity of the vehicle and 𝑇 is the thrust.
In addition, in the case of stiff rotors without hinges at the hub, there is
also a moment 𝑀𝑏, generated directly at the rotor hub from the flapping of
the blades:

(33)
where 𝑘𝛽 is the stiffness of the rotor blade in Nm/rad. The total
longitudinal moment created by blade flapping 𝑀𝑏𝑓 is the sum of these two
moments:
274 Unmanned Aerial Vehicles (UAV) and Drones

(34)
Although a controller designed exactly is possibly successful to
counteract small disturbances, it is difficult to reject the large systematic
disturbances that result from the aerodynamic effects such as blade flapping.
For the improvement of control performance, it is necessary to design a
feed forward compensation block in order to cancel out moments and forces
resulting from blade flapping and variations in total thrust [33].

Ground Effect and Ceiling Effect

When a rotor operates near the ground (about at half rotor diameter), a
phenomenon always appears that thrust augmentation pushes the vehicle
away from the ground, which is related to a reduction of the induced airflow
velocity.
This is called ground effect [71, 72]. Different from other approaches,
an adaptive technique [73] is an option to deal with this effect. However,
for the aim of improvement of the autonomous VTOL controller, a principal
model of this effect is needed.
One proposed mathematical model of ground effect [74] is

(35)
Here 𝑅 is the radius of the rotor, 𝑧 is the vertical distance from the
ground, 𝑇 is the thrust produced by the propeller in ground effect, and 𝑇∞ is
the thrust produced at the same power outside of ground effect.
Note that for 𝑧/𝑅 = 2 the predicted ratio between 𝑇 and 𝑇∞ is just 1.016.
Therefore, this formula (35) predicts that ground effect is negligible when
the rotor is more than one diameter off the ground, that is, 𝑧/𝑅 > 2.
Except for the ground effects, a “ceiling effect” is another issue needs
to be researched, the reason is a quadrotor can flight indoor different from
conventional helicopter.
In fact, so called “ceiling effect” means when the vehicle is close to an
overhead plane, the ceiling effect pulls the vehicle towards the ceiling which
can cause a crash in the worst case.
The effects have been proved by a set of experiments. Unfortunately, no
formal formulation of a ceiling effect is presented so far.
 A Survey of Modelling and Identification of Quadrotor Robot 275

IDENTIFICATION OF A QUADROTOR MODEL

Model is the foundation and the first step of control and simulation. In
general, system models are derived from first principles, physical insight
and sometimes input and output data, and the last two items are classified
as system identification. As for a Quadrotor, the first principle scheme is
intuitive to obtain the dynamical model, perhaps from the inheritance of
traditional flight mechanics, by which the relationship between the inputs
and outputs, that is, the underlying dynamics is revealed distinctly. However,
on the other hand, the mathematical formulation proposed is characterized
by the unwelcome complexity and strong nonlinearity that is regarded as a
nightmare for controller design.
As a alternative solution, system identification is effective to derive a
model. System identification, as the art and science of building mathematical
models of dynamic systems from observed input-output data, has developed
for few decades, starting from the year 1965, and enormous methods
are presented. However, there are many open problems [75, 76]: such as
nonlinearity and closed-loop identification, which are just the characteristics
shown in the quadrotor dynamics.
Significantly, the attention on the model and identification aspect is paid
on the fixed-wing and helicopter [34], instead of quadrotor, or multirotor,
and the reason may be the fact of less applications of quadrotor aircraft
by now, as well as relative complicated dynamics which exhibit some
distinctive features on the modeling and identification schemes, presented
as follows [77].
i Continuous-time model is preferable. In aerospace applications, a
continuous-time model looks more popular than the discrete-time
one for the reason of intuition.
ii To support closed-loop identification, a quadrotor system is
basically open-loop unstable, so that identification experiments
have to be implemented in closed loop under automatic control,
or with a human operator.
iii To uncouple the cross-interaction among the operation axes,
the cross-coupling in dynamic behavior imposes difficulty on
the estimation of a model, so that isolation of different dynamic
modes is necessary to alleviate this effect.
To the best of my knowledge, although not many, some schemes
dedicated to the quadrotor model identification are proposed in the following.
276 Unmanned Aerial Vehicles (UAV) and Drones

Herein, by means of the classification in the literature [75], so called a whole

palette of grey shades from white to black, these methods are introduced
successively.

Off-White Models: Parameters Estimate Based on First Prin-

ciple Model
A set of parameter estimates in the nonlinear quadrotor model could be
taken directly from measurements, a CAD model using model software
like SOLIDWORKS to model all the parts of the quadrotor, or derive from
experiments, along with the associated error [66, 70, 78]. The parameters
listed in the following context can be obtained by a regular identification
method [79–81].
i Aerodynamic parameters: rotor, blade, and aerodynamic
parameters are obtained through measurement, computation,
simulation, or from references.
ii Masses and displacements: component masses and distances are
measured with respect to the rotor plane.
iii Rotational inertia is obtained through measurement and
computation [2].
iv Motor constants: resorting to some experiments conducted, the
mo tor model can be simplified to be a first order system and the
constants could be extracted from the experimental data [82, 83].
The nonlinear identification problem is to estimate such parameters
from the measured data. In general, this is a difficult problem that has not yet
been treated in full generality. In [84], a Levenberg-Marquardt optimization
method and a quadratic optimization method are applied, respectively, to
obtain the 𝑧 inertia and the rotor parameters.

Steel-Grey Models: Local Linear Models Identification

Nonlinear systems are often handled by linearization around a working point.
The idea behind composite local models is to deal with the nonlinearities
by developing local models, which are good approximations in different
neighborhoods, and then compose a global model from these. Often, the
local models are linear, so a common name for composite models is also
local linear models [75], as described in Section 4.5, where the hover regime
is acted as a working point. After the linearization at working point, the
identification issue is simplified and easy to tackle with the help of linear
 A Survey of Modelling and Identification of Quadrotor Robot 277

identification methods as follows. However, the linearization is a realistic

simplification after all, which results in the bias, even model mismatch, and
sometimes is unreasonable in the case of aggressive maneuvers. In addition,
a set of models need to be derived in the situations that many working points
exist, so the switch between two models in the model set should be paid
enough attention to weaken the disturbance.

Parameter Identification
Using the linearized system dynamics after some treatments such as
neglecting the nonlinear coupling terms, a parameter identification [85] is
performed to identify separately each quadrotor axis performed in closed
loop. The generic scheme of the identification process is depicted in Figure
7.

Figure 7: Indirect closed-loop identification setup.

The controller (𝑠) used during identification is a simple stabilizing,
hand-tuned PD-controller with known parameters. Using linearized system
dynamics equation, the identification signal, that is, a pseudorandom binary
sequence (PRBS) of full length, and the controller 𝐶 and output data (𝑦, 𝑦)̇
𝑇
, one can use nonlinear optimization to estimate that our parameter vectors
Δ𝜙,𝜃 = (𝑝, 𝐽, 𝑇𝑑) 𝑇 and Δ 𝜓 = (𝑝, 𝑧, 𝐽, 𝑇𝑑) 𝑇 are conducted, respectively.
The experimental results show a very good correlation with real data, which
confirms the proposed approach in which an iterative parameter identification
scheme is applied, the results of which can easily be reproduced and
offers great accuracy. In general, a regular procedure is implemented for
the parameter identification, in which a linear model is derived from the
nonlinear one with some simplification and neglecting, and then a parameter
identification problem is shown, at last a iterative algorithm is applied to
obtain the estimated value of the parameters, just like what is described in
this literature. Next, an adaptive controller could be designed based on the
parameter identification.
278 Unmanned Aerial Vehicles (UAV) and Drones

Time Domain Identification

Reference [86] presents the estimation of a linear mathematical model for
the dynamics of a quadrotor by time domain system identification.
At first, the model structure has to be determined by the begin with a
large pool of potential candidate regressors, meaning states or control inputs,
or some combination thereof, then calculate the potential correlation of each
regressor with a state derivative using a linear least-squares method. The
final step in the process is to retain those regressors which have a significant
correlation with the state derivative in question. The level of correlation is
determined by the user so as to capture as much of the measured behavior as
possible with a minimal number of regressors. An example of such a pool of
regressors is given as follows:

(36)
where the state derivative 𝑢̇ is linearly related to the regressors, 𝑢, 𝑞,
𝜃, and 𝛿lon, by their corresponding parameters, 𝑋𝑢, 𝑋𝑞, 𝑋𝜃, and 𝑋lon. This
process of regressor pool correlation is repeated for each state derivative.
After the appropriate model structure has been determined, the next
step is to determine the value and error of each parameter by a linear least
squares method. These values form the dynamics and control matrices, 𝐴
and 𝐵. At the same time, the error values are also adjusted to account for any
remaining uncharacterized behavior, known as colored residuals.
The method herein is considered as the basic algorithm in the realm of
system identification, which is used to address the model identification issue
of linear system. It should be noted that the method is desirable for the SISO
(single input and single output) system; therefore, such characteristics as
cross-coupling must be mitigated in advance.

Frequency-Domain Identification
A frequency-domain system identification method is used to obtain a linear
representation of the quadrotor dynamics [87]. Contrast to time domain
analysis, frequency-domain identification can obtain a relative robust model
with the treatment of cutting down the errors associated with bias effects and
processing noise, resulting in a robust model.
In the algorithm, the frequency response data acquired is validated by
evaluating its coherence, which is an indication of how well the output and
input data are correlated. The definition of coherence is given as
 A Survey of Modelling and Identification of Quadrotor Robot 279

(37)
where represent the autospectral densities
of the input, output, and cross-spectral density of the input and output,
respectively, and are the frequency point. A perfect correlation between input
and output would result in a coherence value of unity, while poor coherence
typically falls below a value of 0.6.
It might also be noted that the data must be decoupled such that the
inputs provided by off-axis commands are rejected from the output on the
axis of interest, after the coherence of the data is validated. The multiple
single output system estimation can be expressed in (38), where 𝐻̂ is the
system estimation:

(38)
In the system identification process, the transfer functions of each axis
will be acquired first, followed by state space representations and complete
system analysis. The single input-single output (SISO) transfer function
identification cost function can be defined as

(39)
where the parameters such as 𝑛𝜔, 𝜔1 refer to [87].
As shown in [87], based on the rational experimental setup, a frequency-
domain system identification method obtains a linear representation of the
quadrotor dynamics. It might also be noted that the choice of the periodic
excitation signal is to minimize leakage in the computation of frequency
spectra, which is still an open problem in the area.

Subspace Model Identification

A subspace model identification (SMI) method [77], which has been proved
extremely successful in dealing with the estimation of statespace models for
multiple-input multiple-output (MIMO) systems is used to the identification
of a quadrotor flight dynamics. More precisely, the continuous-time predictor
based subspace identification approach proposed is applied to flight data
collected during dedicated identification experiments, and at hovering flight
condition, a linear statespace model is derived. As an advantage over the
280 Unmanned Aerial Vehicles (UAV) and Drones

most identification techniques, this approach is feasible for the application

in a closed-loop system as the correlation between 𝑢 and 𝑤, V is not required.
The key ideas of the algorithm are provided in the following.
Consider the linear time-invariant continuous-time system:

(40)
where 𝑥∈𝑅 , 𝑢∈𝑅 , and 𝑦∈𝑅 are, respectively, the state, input, and
𝑛 𝑚 𝑝

output vectors and 𝑤∈𝑅𝑛 and V ∈ 𝑅𝑝 are the process and the measurement
noise, respectively, modeled as Wiener processes with incremental
covariance given by

(41)
The system matrices 𝐴, 𝐵, 𝐶, and 𝐷, of appropriate dimensions, are such
that (𝐴, 𝐶) is observable and ([𝐵, 𝑄1/2], ) is controllable. Assume that a data
set (𝑡𝑖), 𝑦(𝑡𝑖), 𝑖[1, 𝑁] of sampled input/output data obtained from (41) is
available. Then, the problem is to provide a consistent estimate of the state
space matrices 𝐴, 𝐵, 𝐶, and 𝐷 on the basis of the available data.
Note that both model order and the tuning parameters of the identification
algorithm (i.e., the position of the Laguerre pole a and the parameters of
the PBSIDopt algorithm) need to be achieved at the head of the procedure;
herein, a crossvalidation approach, explained in detail in the literature [77],
is used to address the issue.
As can be observed from the experiments, in which the input signal
adopted for identification experiments is the so-called 3211 piecewise
constant sequence, the identified models capture the essential features of
the response of the quadrotor along all the axes. As it is known that the
SMI method is rapid and easy to use, however, the model deserved from
the algorithm is not based on some kind of optimal criteria, so the model
obtained is also not believed to be optimum.

UKF Method
As we know, if the systems have severe nonlinearities, EKF can be hard
 A Survey of Modelling and Identification of Quadrotor Robot 281

to tune and often gives unreliable estimation due to the linearization relied
by the EKF in order to propagate the mean and covariance of the states.
Therefore, UKF is applied for the identification of a quadrotor model [88].
For the quadrotor system with continuous-time system dynamics,

(42)
Since the state vector of the quad-rotor

(43)
the system state equation can be derived according to the full nonlinear
equations. Finally, based on all of the system equations, the parameters to be
estimated and identified are formulated as follows:

(44)
Based the experiments, the error of the estimation for velocity at 𝑥-axes
is less than 0.001, while the errors at both 𝑦- axes and 𝑧-axes are less than
0.0015. For estimation of angular velocities at 𝑥, 𝑦, and 𝑧-axes, the errors
are less than 0.0015. From the errors computed, it can be concluded that
the UKF output matches with the measured output and the measured noise
is well filtered by the UKF. It has good convergence time and relatively
reliable for the estimations.

Slate-Grey Models: RBF-ARX Model

The RBF-ARX model is a nonlinear time-varying model whose structure
resembles the ARX model. Its independent variables are groups of signals
indicating the nonlinear status of the system, and its model parameters can
be promptly adjusted to the best by taking the advantages of the RBF neural
network. Owing to the self-adjusting parameters, the RBF-ARX model
not only has an outstanding approximation in local linear space but also
has superior global performance. RBF-ARX basic model structure of a
quadrotor is shown as follows [89, 90]:
282 Unmanned Aerial Vehicles (UAV) and Drones

(45)
where 𝑛𝑦, 𝑛𝑢, 𝑚, and ℎ are the order of the system, 𝑑 is delay factor of
system, and 𝑒(𝑡) is white noise. A quadrotor system is a nonlinear system
with four inputs and 3 outputs, and the inputs are the input electric voltage
of four rotors, that is, 𝑈(𝑡) = [𝑉𝑓(𝑡) 𝑉𝑟(𝑡) 𝑉𝑙(𝑡) 𝑉𝑏(𝑡)]𝑇, and the three outputs
are pitch, roll, and yaw angles, respectively, that is, 𝑌(𝑡) = [𝑝(𝑡) 𝑟(𝑡) 𝑦(𝑡)]𝑇
and 𝑤(𝑡) = [𝑝(𝑡) 𝑟(𝑡)]𝑇. A set of simulation tests show that the error of RBF-
ARX model is most close to a normal distribution, which indicates that the
good model is obtained. In addition, no matter how the state variable (𝑡)
slides, the distribution of system pole does not go beyond the stable scope.
So the RBF-ARX model is suitable for quadrotor.

Black Models

Neural Network Model

A black box model that uses a neural network to learn the dynamics of the
quadrotor is attempted [79]. The nonlinear autoregressive network with
exogenous inputs (NARX) architecture is setup which has 6 different nets,
one each for the 𝑋, 𝑌, and 𝑍 velocities and roll, pitch, and yaw rates. The other
variables are then integrated (position/angles) and derive (accelerations)
from the calculated velocities/rates. The procedure, described in Figure 8, is
designed to address the issue of the current prediction (𝑦𝑡) dependent on the
predicted output for the previous time point (𝑦𝑡−1), in which a series-parallel
architecture, (𝑡 − 1) is the correct output for time point 𝑡−1, 𝑢(𝑡) is the input
vector for time point 𝑡, and 𝑦(𝑡) is the prediction for time point 𝑡, is used for
training before converting the net to the parallel architecture.
 A Survey of Modelling and Identification of Quadrotor Robot 283

Figure 8: The flow for training the neural net.

Some tests result shows that the black box neural network model can predict
both the roll and pitch with very good accuracy and however perform not
better for the yaw rate, which will be improved by creating a larger net
or by adding more variables to the state vector for the nets. Surprisingly,
the model was tested on a manoeuvre for which it had not been trained; a
successful result is obtained. This result shows that the black box neural
network model learned not only the dynamics of the quadrotor but also the
dynamics of the trends and the noise. Neural networks have proved to be a
remarkable modeling scheme carried out in various applications; however,
on the other hand, substantial amount of tests need to be taken. Obviously,
the work is costly and time-consuming.

Data-Based Model
The main purpose of data-based techniques is to take full advantage of the
information acquired from huge amounts of process measurements. Without
recourse to physical models obtained from first principles, a relatively
overall perspective of system performance could be revealed via available
measurements. Through deep insights of process measurements, information
like system characteristics and regularity can be dug out for optimal
modeling and decision making. The description [91] aforementioned reveals
the insight and application of data-based model identification schemes.
It is noted that several typical data-based approaches, which only depend
on process measurements, principal component analysis (PCA), partial least
squares (PLS), and their variants, are successfully utilized in many areas
[92–96], In the realms of model and control, iterative learning control (ILC)
scheme, and model free adaptive control (MFAC)—in essence, model free
methods—show great advantages without a priori knowledge about the
underlying processes, such as time delay and system order, despite their
potential limit for processes with high complexity.
284 Unmanned Aerial Vehicles (UAV) and Drones

In addition, the recent developments on model-data integrated approaches,

which also rely on available process measurements and a prior known
limited knowledge about the processes for monitoring and control purposes,
the iterative feedback tuning (IFT), and virtual reference feedback tuning
(VRFT), have become the promising research topics. Although there are
no reports of the application of the data-based model identification on the
quadrotor so far, these approaches will soon find their utilizations.
Remark. Hover condition is the main status of the quadrotor, as quite a
few tasks, such as surveillance, search, and rescue, are implemented in the
condition.
Therefore, linear model is simplified on the complex nonlinear one
derived from first principle model, in which the feasibility is proved by
application result. However, aggressive maneuver shows the obvious
nonlinear characteristics, so that the nonlinear model is needed, in which
neural networks are an optional scheme despite the fact that a large amount
of tests are indispensable.
The data-based approaches have shown the distinctive advantages in
other application areas; the utilization on a quadrotor will be a commendable
attempt. It is noted that these methods aforementioned not only are applicable
in the quadrotor aspect but also is helpful to the field with the similar
characteristics, such as active suspension systems of vehicle investigated in
[97, 98].

CONCLUSION AND PERSPECTIVE

By now, a quadrotor has been a preferable platform of aircraft design and
autonomous control due to the distinct characteristics involved in the craft
configuration and flight dynamics, and the expansion to the commercial and
military application is underway.
The research results at present show that the static and dynamic
characteristics of a quadrotor have been developed, and the aerodynamic
effects on the craft flight performance have been revealed with the help of
the conventional rotor theory.
As a base and preliminary for the next control and simulation work,
the paper surveys the state of art of the modeling and identification of a
quadrotor; afterwards some opinions are given as follows.
1) The full nonlinear equations have been obtained by classical
Euler-Newtonian mechanics as well as Lagrange equation,
 A Survey of Modelling and Identification of Quadrotor Robot 285

with the knowledge of different force imposed on, especially

aerodynamic function. However, there is no specific research
of the model of a quadrotor, in which the dynamics is described
comprehensively and systematically.
2) The dynamics of quadrotor shows distinct characteristics for the
VTOL, horizontal flight, aggressive maneuver, and hovering mode
involved in the execution of flight task. Whereas at the present,
the hovering mode attracts over much attention, probably owing
to the fact that the hovering is the one of the important working
modes. Obviously, more efforts of research should be paid on the
other modes such as aggressive maneuver.
3) The method of multistatic status linearization model has been
presented at the aim of control, which may be an effective way
to address the issue of complex dynamics. However, a problem
of undisturbed switching between the multimodels should be
resolved.
4) In contrast with the other crafts, there is no more attention for
the model identification of a quadrotor for the reason of the
requirements mentioned in the paper. The effective transplantation
of the identification methods used in the other respects is a
practical choice.
5) A new configuration of quadrotor with tilt rotor can eliminate
the deficiency such as underactuate feature. The relative research
work has been in progress, which will be a focus hereafter.

CONFLICT OF INTERESTS
The authors declare that there is no conflict of interests regarding the
publication of this paper.

ACKNOWLEDGMENTS
This work was supported by the Fundamental Research Funds for the Central
Universities (Grant no. FRF-TP-12-005B), Program for New Century
Excellent Talents in Universities (Grant no. NCET-11-0578), Aeronautical
Science Foundation of China (Grant no. 2012ZD54013), and Specialized
Research Fund for the Doctoral Program of Higher Education (Grant no.
20130006110008).
286 Unmanned Aerial Vehicles (UAV) and Drones

REFERENCES
1. V. Martinez, Modeling of the flight dynamics of a quadrotor Helicopter
[M.S. thesis], Cranfield University, Cranfield, UK, 2007.
2. M. Y. Amir and V. Abbass, “Modeling of quadrotor helicopter
dynamics,” in Proceedings of the International Conference on Smart
Manufacturing Application (ICSMA ‘08), pp. 100–105, Kintex, April
2008.
3. G. Warwick, “US Army looking at three configuration concepts for
large cargo rotorcraft,” September 2013, http://www.flightglobal.com/
news/articles/us-army-looking-at-three-configuration-concepts-for-
large-cargo-rotorcraft-217956/.
4. R. Mahony, V. Kumar, and P. Corke, “Multirotor aerial vehicles:
modeling, estimation, and control of quadrotor,” IEEE Robotics and
Automation Magazine, vol. 19, no. 3, pp. 20–32, 2012.
5. R. Mahony and V. Kumar, “Aerial robotics and the quadrotor,” IEEE
Robotics and Automation Magazine, vol. 19, no. 3, p. 19, 2012.
6. I. Gaponov and A. Razinkova, “Quadcopter design and implementation
as a multidisciplinary engineering course,” in Proceedings of the 1st
IEEE International Conference on Teaching, Assessment, and Learning
for Engineering (TALE ‘12), pp. B16–B19, IEEE, Hong Kong, China,
August 2012.
7. N. Guenard, T. Hamel, and R. Mahony, “A practical visual servo control
for an unmanned aerial vehicle,” IEEE Transactions on Robotics, vol.
24, no. 2, pp. 331–340, 2008.
8. S. Bouabdallah and R. Siegwart, “Towards intelligent miniature flying
robots,” in Tractsaction in Advanced Robotics, vol. 25, pp. 429–440,
Springer, Berlin, Germany, 2006.
9. G. Hoffmann, D. G. Rajnarayan, S. L. Waslander, D. Dostal, J. S. Jang,
and C. J. Tomlin, “The stanford testbed of autonomous rotorcraft for
multi agent control (STARMAC),” in Proceedings of the 23rd Digital
Avionics Systems Conference (DASC ‘04), vol. 2, pp. 12.E.4–121.10,
Salt Lake City, Utah, USA, October 2004.
10. L. Meier, P. Tanskanen, F. Fraundorfer, and M. Pollefeys, “PIXHAWK:
a system for autonomous flight using onboard computer vision,” in
Proceedings of the IEEE International Conference on Robotics and
Automation (ICRA ‘11), pp. 2992–2997, IEEE, Shanghai, China, May
2011.
 A Survey of Modelling and Identification of Quadrotor Robot 287

11. S. K. Phang, C. Cai, B. M. Chen, and T. H. Lee, “Design and

mathematical modeling of a 4-standard-propeller (4SP) quadrotor,”
in Proceeding of the 10th World Congress on Intelligent Control and
Automation (WCICA’12), pp. 3270–3275, Beijing, China, July 2012.
12. M. S. Hossain, A. M. Kabir, and P. Mazumder, “Design and
development of an Y4 Copter control system,” in Proceedings of the
14th International Conference on Computer Modeling and Simulation
(UKSim’12), pp. 251–256, Cambridge, UK, March 2012.
13. F. Senkul and E. Altug, “Modeling and control of a novel tilt: roll rotor
quadrotor UAV,” in Proceeding of the International Conference on
Unmanned Aircraft Systems (ICUAS’13), pp. 1071–1076, Atlanta, Ga,
USA, May 2013.
14. M. Ryll, H. H. Bulthoff, and P. R. Giordano, “First flight tests for a
quadrotor UAV with tilting propellers,” in Proceedings of the IEEE
International Conference on Robotics and Automation (ICRA ‘13), pp.
295–302, IEEE, Karlsruhe, Germany, May 2013.
15. S. Driessens and P. E. I. Pounds, “Towards a more efficient quadrotor
configuration,” in Proceedings of the IEEE/RSJ International
Conference on Intelligent Robots and Systems (IROS ‘13), pp. 1386–
1392, Tokyo, Japan, 2013.
16. H. Lim, J. Park, D. Lee, and H. J. Kim, “Build your own quadrotor:
Open-source projects on unmanned aerial vehicles,” IEEE Robotics
and Automation Magazine, vol. 19, no. 3, pp. 33–45, 2012.
17. D. Mellinger and V. Kumar, “Minimum snap trajectory generation
and control for quadrotors,” in Proceedings of the IEEE International
Conference on Robotics and Automation (ICRA ‘11), pp. 2520–2525,
IEEE, Shanghai, China, May 2011.
18. Q. Zhan, J. Q. Wang, and X. Xi, “Control system design and experiments
of a quadrotor,” in Proceedings of the IEEE International Conference
on Robotics and Biomimetics (ROBIO ‘12), pp. 1152–1157, IEEE,
Guangzhou, China, December 2012.
19. L. Jun and Y. T. Li, “Dynamic analysis and PID control for a quadrotor,”
in Proceedings of the International Conference on Mechatronics and
Automation, pp. 573–578, IEEE, Beijing, China, August 2011.
20. A. L. Salih, M. Moghavvemi, H. A. F. Mohamed, and K. S. Gaeid,
“Flight PID controller design for a UAV quadrotor,” Scientific Research
and Essays, vol. 5, no. 23, pp. 3660–3667, 2010.
288 Unmanned Aerial Vehicles (UAV) and Drones

21. A. B. Milhim, Y. Zhang, and C. A. Rabbath, “Quadrotor UAV-high

fidelity modeling and nonlinear PID control,” in Proceedings of AIAA
Modeling and Simulation Technologies Conference, pp. 2010–8362,
Toronto, Canada, 2010.
22. R. Zawiski and M. Błachuta, “Model development and optimal control
of quadrotor aerial robot,” in Proceedings of the 17th International
Conference on Methods and Models in Automation and Robotics
(MMAR ‘12), pp. 475–480, Miedzyzdrojie, Poland, August 2012.
23. E. Reyes-Valeria, R. Enriquez-Caldera, S. Camacho-Lara, and J.
Guichard, “LQR control for a quadrotor using unit quaternions:
modeling and simulation,” in Proceedings of the 23rd International
Conference on Electronics, Communications and Computing
(CONIELECOMP ‘13), pp. 172–178, IEEE, Cholula, Mexico, March
2013.
24. M. Rich, N. Elia, and P. Jones, “Design and implementation of an
H∞ controller for a quadrotor helicopter,” in Proceeding of the 21st
Mediterranean Conference on Control and Automation (MED’13), pp.
1189–1198, Chania, Greece, June 2013.
25. P. Mukherjee and S. L. Waslander, “Direct adaptive feedback
linearization for quadrotor control,” in Proceedings of the AIAA
Guidance, Navigation, and Control Conference, pp. 2012–4917,
Minneapolis, Minn, USA, August 2012.
26. A. Khalifa, M. Fanni, A. Ramadan, and A. Abo-Ismail, “Modeling and
control of a new quadrotor manipulation system,” in Proceedings of
the 1st International Conference on Innovative Engineering Systems
(ICIES ‘12), pp. 109–114, Alexandria, Egypt, December 2012.
27. L. Pollini and A. Metrangolo, “Simulation and robust backstepping
control of a Quadrotor aircraft,” in Proceedings to AIAA Modeling
and Simulation Technologies Conference and Exhibit, pp. 2008–6363,
Honolulu, Hawaii, USA, 2008.
28. A. A. Mian and D. B. Wang, “Modeling and backstepping-based
nonlinear control strategy for a 6 DOF quadrotor helicopter,” Chinese
Journal of Aeronautics, vol. 21, no. 3, pp. 261–268, 2008.
29. F. Casolo, Motion Control, InTech, Croatia, Rijeka, 2010.
30. M. Santos, V. López, and F. Morata, “Intelligent fuzzy controller of a
quadrotor,” in Proceedings of the IEEE International Conference on
Intelligent Systems and Knowledge Engineering (ISKE ‘10), pp. 141–
 A Survey of Modelling and Identification of Quadrotor Robot 289

146, IEEE, Hangzhou, China, November 2010.

31. C. M. Korpela, T. W. Danko, and P. Y. Oh, “MM-UAV: mobile
manipulating unmanned aerial vehicle,” Journal of Intelligent and
Robotic Systems, vol. 65, no. 1–4, pp. 93–101, 2012.
32. D. Mellinger, N. Michael, and V. Kumar, “Trajectory generation and
control for precise aggressive maneuvers,” Springer Tractsaction in
Advanced Robotics, vol. 79, pp. 361–373, 2014.
33. H. M. Huang, G. M. Hoffmann, S. L. Waslander, and C. J. Tomlin,
“Aerodynamics and control of autonomous quadrotor helicopters in
aggressive maneuvering,” in Proceedings of the IEEE International
Conference on Robotics and Automation (ICRA ‘09), pp. 3277–3282,
IEEE, Kobe, Japan, May 2009.
34. N. V. Hoffer, C. Coopmans, A. M. Jensen, and Y. Q. Chen, “A survey
and categorization of small low-cost unmanned aerial vehicle system
identification,” Journal of Intelligent and Robotic Systems, vol. 74, no.
1-2, pp. 129–145, 2014.
35. S. Gupte, P. I. T. Mohandas, and J. M. Conrad, “A survey of quadrotor
unmanned aerial vehicles,” in Proceeding of the IEEE Southeastcon,
pp. 1–6, Orlando, Fla, USA, March 2012.
36. A. R. Partovi, A. Z. Y. Kevin, H. Lin, B. M. Chen, and G. Cai,
“Development of a cross style quadrotor,” in Proceedings of the AIAA
Guidance, Navigation, and Control Conference, St. Paul, Minn, USA,
August 2012.
37. E. Altu , J. P. Ostrowski, and R. Mahony, “Control of a quadrotor
helicopter using visual feedback,” in Proceedings of the IEEE
International Conference on Robotics and Automation, pp. 72–77,
IEEE, Washington, DC, USA, May 2002.
38. D. S. Miller, Open loop system identification of a micro quadrotor
helicopter from closed loop data [M.S. dissertation], University of
Maryland, College Park, Md, USA, 2011.
39. J. Dvorak, Micro quadrotor-design, modeling, identification and
control [M.S. thesis], Czech Technical University, Prague, Czech
Republic, 2011.
40. P. Pounds, R. Mahony, and P. Corke, “Modelling and control of a large
quadrotor robot,” Control Engineering Practice, vol. 18, no. 7, pp.
691–699, 2010.
41. J. H. Kim, M.-S. Kang, and S. Park, “Accurate modeling and robust
290 Unmanned Aerial Vehicles (UAV) and Drones

hovering control for a quad-rotor VTOL aircraft,” Journal of Intelligent

and Robotic Systems, vol. 57, no. 1–4, pp. 9–26, 2010.
42. M. Elsamanty, A. Khalifa, M. Fanni, A. Ramadan, and A. Abo-Ismail,
“Methodology for identifying quadrotor parameters, attitude estimation
and control,” in Proceedings of the IEEE/ASME International
Conference on Advanced Intelligent Mechatronics (AIM ‘13), pp.
1343–1348, Wollongong, Australia, July 2013.
43. J. Kim, M.-S. Kang, and S. Park, “Accurate modeling and robust
hovering control for a quad-rotor VTOL aircraft,” Journal of Intelligent
and Robotic Systems, vol. 57, no. 1–4, pp. 9–26, 2010.
44. T. Bresciani, Modelling, identification and control of a quadrotor
Helicopter [M.S. thesis], Lund University, Lund, Sweden, 2008.
45. H. J. Wang, “The essence of dynamic mechanics for lagrange equation,”
Journal of Hebei University of Science and Technology, vol. 24, no. 2,
pp. 56–60, 2003 (Chinese).
46. P. Castillo, R. Lozano, and A. Dzul, “Stabilization of a mini-rotorcraft
having four rotors,” in Proceedings of the IEEE/RSJ International
Conference on Intelligent Robots and Systems (IROS ‘04), pp. 2693–
2698, Sendal, Japan, October 2004.
47. R. W. Beard, Quadrotor Dynamics and Control, ‘Brigham Young
University, 2008.
48. P. McKerrow, “Modelling the draganflyer four-rotor helicopter,” in
Proceedings of the IEEE International Conference on Robotics and
Automation, pp. 3596–3601, New Orleans, La, USA, May 2004.
49. M. D. L. C. de Oliveira, Modeling, Identification and Control of a
Quadrotor Aircraft [M.S. thesis], Czech Technical University, Prague,
Czech Republic, 2011.
50. S. Bouabdallah, P. Murrieri, and R. Siegwart, “Design and control of an
indoor micro quadrotor,” in Proceeding of the 2004 IEEE International
Conference on Robotics and Automation (ICRA’04), vol. 5, pp. 4393–
4398, May 2004.
51. A. Benallegue, A. Mokhtari, and L. Fridman, “Feedback linearization
and high order sliding mode observer for a quadrotor UAV,” in
Proceedings of the International Workshop on Variable Structure
Systems (VSS ‘06), pp. 365–372, Alghero, Italy, June 2006.
52. S. Bouabdallah and R. Siegwart, “Full control of a quadrotor,” in
Proceedings of the IEEE/RSJ International Conference on Intelligent
 A Survey of Modelling and Identification of Quadrotor Robot 291

Robots and Systems (IROS ‘07), pp. 153–158, San Diego, Calif, USA,
November 2007.
53. C. A. Herda, Implementation of a quadrotor unmanned arial vehicle
[M.S. dissertation], California State University, Los Angeles, Calif,
USA, 2012.
54. J. M. B. Domingues, Quadrotor prototype [M. S. dissertation], Instituto
Superior Tecnico, Göttingen, Germany, 2009.
55. A. Tayebi and S. McGilvray, “Attitude stabilization of a VTOL
quadrotor aircraft,” IEEE Transactions on Control Systems Technology,
vol. 14, no. 3, pp. 562–571, 2006.
56. N. A. Chaturvedi, A. K. Sanyal, and N. H. McClamroch, “Rigid-body
attitude control,” IEEE Control Systems Magazine, vol. 31, no. 3, pp.
30–51, 2011. | MathSciNet
57. V. Mistler, A. Benallegue, and N. K. M’Sirdi, “Exact linearization and
noninteracting control of a 4 rotors helicopter via dynamic feedback,”
in Proceedings of the 10th IEEE International Workshop on Robot and
Human Communication, pp. 586–593, Paris, France, September 2001.
58. J. P. How, B. Bethke, A. Frank, D. Dale, and J. Vian, “Real-time
indoor autonomous vehicle test environment,” IEEE Control Systems
Magazine, vol. 28, no. 2, pp. 51–64, 2008. | MathSciNet
59. M. J. Stepaniak, A quadrotor sensor platform [Ph.D. dissertation],
Ohio University, Athens, Ohio, USA, 2008.
60. A. Nanjangud, “Simultaneous low-order control of a nonlinear
quadrotor model at four equilibria,” in Proceedings of the IEEE
Conference on Decision and Control, pp. 2914–2919, Florence, Italy,
December 2013.
61. R. K. Agarwal, Recent Advances in Aircraft Technology, InTech,
Rijeka, Croatia, 2012.
62. G. M. Hoffmann, H. M. Huang, S. L. Waslander, and C. J. Tomlin,
“Quadrotor helicoper flight dynamics and control- theory and
experiment,” in Proceedings of the AIAA Guidance, Navigation and
Control Conference and Exhibit, p. 6461, Hilton Head, SC, USA, 2007.
63. P. Pounds, R. Mahony, J. Gresham, P. Corke, and J. Roberts, “Towards
dynamically favourable quad-rotor aerial robots,” in Proceedings of
the Australasian Conference on Robotics and Automation, Canberra,
Australia, December 2004.
64. G. M. Hoffmann, S. L. Waslander, and C. J. Tomlin, “Quadrotor
292 Unmanned Aerial Vehicles (UAV) and Drones

helicopter trajectory tracking control,” in Proceedings of the AIAA

Guidance, Navigation and Control Conference and Exhibit, Honolulu,
Hawaii, USA, August 2008.
65. G. M. Hoffmann, H. Huang, S. L. Waslander, and C. J. Tomlin,
“Precision flight control for a multi-vehicle quadrotor helicopter
testbed,” Control Engineering Practice, vol. 19, no. 9, pp. 1023–1036,
2011.
66. P. Pounds, R. Mahony, and P. Corke, “Modelling and control of a
quadRotor robot,” in Proceedings to International Conference on
Robotics and Automation, Auckland, New Zealand, 2006.
67. R. W. Prouty, Helicopter Performance, Stability and Control, Krieger,
Melbourne, Fla, USA, 1995.
68. J. Seddon, Basic Helicopter Aerodynamics, BSP, London, UK, 1990.
69. J. G. Leishman, Principles of Helicopter Aerodynamics, Cambridge
University Press, London, UK, 2nd edition, 2006.
70. R. Zawiski and M. Błachuta, “Dynamics and optimal control of
quadrotor platform,” in Proceedings of the AIAA Guidance, Navigation,
and Control Conference, AIAA 2012-4915, Minneapolis, Minn, USA,
August 2012.
71. S. Bouabdallah, Design and control of quadrotors with application
to autonomous flying [M.S. thesis], Université Abou Bekr Belkaid
Tlemcen, Tlemcen, Algeria, 2007.
72. C. Powers, D. Mellinger, and A. Kushleyev, “Influence of aerodynamics
and proximity effects in quadrotor flight,” in Proceedings of the
International Symposium on Experimental Robotics, pp. 17–21,
Quebec, Canada, 2012.
73. N. Guenard, T. Hamel, and L. Eck, “Control laws for the tele operation
of an unmanned aerial vehicle known as an X4-flyer,” in Proceeding
of the IEEE/RSJ International Conference on Intelligent Robots and
Systems (IROS’06), pp. 3249–3254, Beijing, China, October 2006.
74. W. Johnson, Helicopter Theory, Dover, New York, NY, USA, 1980.
75. L. Ljung, “Perspectives on system identification,” Journal of Intelligent
and Robotic Systems, no. 74, pp. 129–145, 2014.
76. M. Gevers, “A personal view of the development of system
identification,” IEEE Control Systems Magazine, vol. 26, no. 6, pp.
93–105, 2006.
77. M. Bergamasco, “Identification of linear models for the dynamics
 A Survey of Modelling and Identification of Quadrotor Robot 293

of a hovering quadrotor,” IEEE Transactions on Control Systems

Technology, vol. 22, no. 5, pp. 1696–1707, 2014.
78. L. K. Burkamshaw, Towards a low-cost quadrotor research platform
[M.S. thesis], Naval Postgraduate School, Monterey, Calif, USA, 2010.
79. D. Sonntag, A study of quadrotor modeling [M. S. dissertation],
Linkopings Universitet, Linköping, Sweden, 2011.
80. L. Derafa, T. Madani, and A. Benallegue, “Dynamic modelling and
experimental identification of four rotors helicopter parameters,”
in Proceedings of the IEEE International Conference on Industrial
Technology (ICIT ‘06), pp. 1834–1839, Mumbai, India, December
2006.
81. P. Pounds, R. Mahony, and P. Corke, “System identification and
control of an aerobot drive system,” in Proceedings of the Information,
Decision and Control Conference (IDC ‘07), pp. 154–159, Adelaide,
Australia, February 2007.
82. S. Bouabdallah, P. Murrieri, and R. Siegwart, “Towards autonomous
indoor micro VTOL,” Autonomous Robots, vol. 18, no. 2, pp. 171–183,
2005.
83. Y. Naidoo, R. Stopforth, and G. Bright, “Quad-rotor unmanned aerial
vehicle helicopter modeling and control,” International Journal of
Advanced Robotic Systems, vol. 8, no. 4, pp. 139–149, 2011.
84. L. Derafa, T. Madani, and A. Benallegue, “Dynamic modelling and
experimental identification of four rotors helicopter parameters,”
in Proceedings of the IEEE International Conference on Industrial
Technology (ICIT ‘06), pp. 1834–1839, December 2006.
85. O. Falkenberg, J. Witt, U. Pilz, U. Weltin, and H. Werner, “Model
identification and H∞ control for MAV’s,” in Proceedings of the
International Conference on Intelligent Robotics and Applications, pp.
460–471, Montreal, Canada, 2012.
86. G. Gremillion, “System identification of a quadrotor micro air vehicle,”
in Proceedings of the AIAA Atmospheric Flight Mechanics Conference,
p. 7644, Toronto, Canada, 2010.
87. W. Wei, “Frequency-domain system identification and simulation of
a quadrotor controller,” in Proceedings of the AIAA Modeling and
Simulation Technologies Conference, pp. 1834–1839, National Harbor,
Maryland, Md, USA, 2014.
88. N. Abas, A. Legowo, and R. Akmeliawati, “Parameter identification
294 Unmanned Aerial Vehicles (UAV) and Drones

of an autonomous quadrotor,” in Proceeding of the 4th International

Conference on Mechatronics (ICOM’11), pp. 1–8, Kuala Lumpur,
Malaysia, May 2011.
89. L. L. Liu, Research on the modeling and control to a quadrotor
helicopter simulator [M.S. thesis], Central South University, Changsha,
China, 2009.
90. Q. L. Luo, Application of RBF-ARX model-based predictive control
on quad-rotor helicopter simulator [M. S. dissertation], Central South
University, Changsha, China, 2012, (Chinese).
91. S. Yin, X. W. Li, H. J. Gao, and O. Kaynak, “Data-based techniques
focused on modern industry: an overview,” IEEE Transactions on
Industrial Electronics, no. 99, pp. 1–11, 2014.
92. S. Yin, S. X. Ding, X. C. Xie, and H. Luo, “A review on basic data-driven
approaches for industrial process monitoring,” IEEE Transactions on
Industrial Electronics, vol. 61, no. 11, pp. 6418–6428, 2014.
93. S. Yin, S. X. Ding, A. Haghani, H. Hao, and P. Zhang, “A comparison
study of basic data-driven fault diagnosis and process monitoring
methods on the benchmark Tennessee Eastman process,” Journal of
Process Control, vol. 22, no. 9, pp. 1567–1581, 2012.
94. S. Yin, G. Wang, and H. R. Karimi, “Data-driven design of robust fault
detection system for wind turbines,” Journal of Mechatronics, vol. 24,
no. 4, pp. 298–306, 2014.
95. S. Yin, S. X. Ding, A. H. A. Sari, and H. Hao, “Data-driven monitoring
for stochastic systems and its application on batch process,”
International Journal of Systems Science, vol. 44, no. 7, pp. 1366–
1376, 2013. | MathSciNet
96. S. Yin, H. Luo, and S. X. Ding, “Real-time implementation of fault-
tolerant control systems with performance optimization,” IEEE
Transactions on Industrial Electronics, vol. 61, no. 5, pp. 2402–2411,
2014.
97. W. C. Sun, H. J. Gao, and O. Kaynak, “Adaptive backstepping
control for active suspension systems with hard constraints,” IEEE
Transactions on Mechatronics, vol. 18, no. 3, pp. 1072–1079, 2013.
98. W. C. Sun, Z. L. Zhao, and H. J. Gao, “Saturated adaptive robust control
for active suspension systems,” IEEE Transactions on Industrial
Electronics, vol. 60, no. 9, pp. 3889–3896, 2013.
 CHAPTER

13
Visual Flight Control of a Quadrotor
Using Bioinspired Motion Detector

Lei Zhang,1 Tianguang Zhang,1 Haiyan Wu,1 Alexander Borst,2 and

Kolja Kühnlenz1,3
Institute of Automatic Control Engineering (LSR), Technische Universität
1

München, 80290 München, Germany

2
Department of Systems and Computational Neurobiology, Max Planck
Institute of Neurobiology, Am Klopferspitz 18, D-82152 Martinsried,
Germany
3
Institute for Advanced Study (IAS), Technische Universität München,
80290 München, Germany

Citation: Lei Zhang, Tianguang Zhang, Haiyan Wu, Alexander Borst, and Kolja Küh-
nlenz, “Visual Flight Control of a Quadrotor Using Bioinspired Motion Detector”, In-
ternational Journal on Navigation and Observation, Volume 2012 |Article ID 627079, 9
pages, https://doi.org/10.1155/2012/627079.
Copyright: © 2012 Lei Zhang et al. This is an open access article distributed under the
Creative Commons Attribution License
296 Unmanned Aerial Vehicles (UAV) and Drones

ABSTRACT
Motion detection in the fly is extremely fast with low computational
requirements. Inspired from the fly’s vision system, we focus on a real-time
flight control on a miniquadrotor with fast visual feedback. In this work,
an elaborated elementary motion detector (EMD) is utilized to detect local
optical flow. Combined with novel receptive field templates, the yaw rate
of the quadrotor is estimated through a lookup table established with this
bioinspired visual sensor. A closed-loop control system with the feedback
of yaw rate estimated by EMD is designed. With the motion of the other
degrees of freedom stabilized by a camera tracking system, the yaw-rate of
the quadrotor during hovering is controlled based on EMD feedback under
real-world scenario. The control performance of the proposed approach
is compared with that of conventional approach. The experimental results
demonstrate the effectiveness of utilizing EMD for quadrotor control.

INTRODUCTION
Flying insects have tiny brains and mostly possess compound eyes which can
get panoramic scene to provide an excellent flying performance. Comparing
with state-of-the-art artificial visual sensors, the optics of compound eye
provide very low spatial resolution. Nevertheless, the behavior of flying
insects is mainly dominated by visual control. They use visual feedback to
stabilize flight [1], control flight speed, [2] and measure self-motion [3].
On the other hand, highly accurate real-time stabilization and navigation
of unmanned aerial vehicles (UAVs) or microaerial vehicles (MAVs) is
becoming a major research interest, as these flying systems have significant
value in surveillance, security, search, and rescue missions. Thus, the
implementation of a bio-plausible computation for visual systems could be
an accessible method to replace the traditional image processing algorithms
in controlling flying robots such as a quadrotor.
Most of early applications using insect-inspired motion detector focus
on motion detection tasks rather than velocity estimation. In robotics
and automation applications, EMDs are mainly used for a qualitative
interpretation of video image sequence, to provide general motion
information such as orientation and infront obstacles. In [4], a microflyer
with an onboard lightweight camera is developed, which is able to fly
indoor while avoiding obstacles by detecting certain changes in optic flow.
The recent approach for the navigation in a corridor environment on an
autonomous quadrotor by using optical flow integration is shown in [5].
 Visual Flight Control of a Quadrotor Using Bioinspired Motion Detector 297

Another example is a tethered optic flow-based helicopter that mimics

insect behaviors such as taking off, cruise, and landing [6, 7], and in [8]
the EMDs visual sensors were tested and characterized in field experiments
under various lighting conditions. Numerous authors have pointed out that
the Reichardt model, while sensitive to motion, does not measure velocity
[9–11]. However, some efforts have been made, examining the possibility
of velocity estimation tasks by introducing elaborated models [12, 13]. In
[14], yaw rate estimates on a coaxial helicopter testbed are obtained using
a matched filter approach which yet incorporates a virtual 3D environment
in the control loop. Although a lot of work has been done on a simulation
level or involving simulation tools, further robotic applications with EMDs
considering closed-loop velocity control are still to be investigated under
real-world scenarios.
In this paper, a quadrotor system with bioinspired visual sensor is
described. The novel image processing methods and the control laws are
implemented in real-time experiments. The yaw rate control is totally based
on the visual feedback of the on-board camera. The reference velocity value
is provided by the on-board inertial measurement unit (IMU). For a 6-DOF
(degrees of freedom) flying robot control, a tracking system of multicamera
configuration is also utilized to achieve the altitude and attitude stabilization
near hover. Due to the noise in real-world, the velocity estimation tasks
would be more challenging. In this work, the approach of building an
empirical lookup table from open-loop test results is introduced for this
task. The Reichardt motion detector is modified which describes, at an
algorithmic level, the process of local motion detection in flying behaviors.
Certain patterns of receptive fields, which respond to particular optic flow,
are utilized to estimate the global ego-motion through the environment.
Another main issue which needs to be tackled carefully in this work is that
the flying robot should be well stabilized during hovering. By multicamera
tracking, the absolute position as well as the pose is determined from the
positions of four on-board markers.
The remainder of this paper is organized as follows: in Section 2, we
firstly introduce the bioinspired visual image processing methods used in this
work. In Section 3, the 3D pose estimation using visual tracking system is
described. The control strategy of the system as well as the software structure
of algorithms is presented in Section 4. Then in Section 5, an overview of
the whole experimental platform is illustrated. The control performance is
also evaluated based on the experimental results in this section. Conclusions
are given in Section 6, with directions on future works.
298 Unmanned Aerial Vehicles (UAV) and Drones

BIOINSPIRED IMAGE PROCESSING

In this section, we introduce the essential part of this work: using biological
models for yaw rate estimation of a quadrotor. The EMDs are utilized for
this task. The whole methodology is introduced in detail. To achieve the
yaw rate control, the system also requires accurate visual tracking for pose
stabilization (Section 3) and efficient controllers (Section 4).
In an insect’s perspective, motion information has to be computed from
the changing retinal images by the nervous system [15]. For engineering
applications, some properties of the biological visual system are converted
into computational algorithms.
The elaborated EMD model used in this work is a modified model of
the famous Reichardt motion detector [16]. The original Reichardt motion
detector (Figure 1(a)) has only low-pass filters and two correlations. In this
work, a temporal high-pass filter is added before the low-pass filter to obtain
a simple response to step edges [17] (Figure 1(b)). The high-pass filters and
low-pass filters in this model are all designed to be of first order.

Figure 1: The simple Reichardt motion detector (a) and the elaborated EMD
model (b). A and B are two visual signal inputs. The moving luminance signal
is observed by this pair of visual sensors (such as the ommatidia of a fruit fly).
The detector compares visual sensor inputs and generates a direction sensitive
response R corresponding to the visual motion. D: delay block. Here it refers to
low-pass filter; HP: High-pass filter; M: multiplication.
In [12], a mathematical analysis of the original Reichardt motion detector
is given regarding the response to different images (sinusoidal gratings as
well as natural images). Without loss of generality, we firstly consider the
response of this modified model to a moving natural image (which possesses
energy at all spatial frequencies). Similar to the response of the simplified
model [12], for this modified model, the output is
 Visual Flight Control of a Quadrotor Using Bioinspired Motion Detector 299

where Δφ is the angular displacement between the two vision sensors,

fs is the spatial frequency of the image input to the detector, v stands for the
velocity of the moving image, τL and τH are the time constants of the low-
and high-pass filters, respectively, and P(fs) represents the power spectral
density. So according to (1), the local motion information is calculated.
To obtain a global ego-motion estimation, certain receptive fields of the
motion-sensitive widefield neurons in the fly brain are applied. Considering
the specified experimental scenario in this work, two novel templates of
receptive fields for rotation detection are utilized (Figure 2), which are
proposed in [17].

Figure 2: Rotation templates: horizontal (a) and vertical (b). The response of
horizontal rotation is calculated by subtracting the upper part and lower part of
the template of receptive fields, while the vertical rotation is detected by the dif-
ference between left part and right part of the global optical flow.
The algorithms for calculating the rotation global response can be
described as (image size: length × width; RH: response of local horizontal
motion; RV : response of local vertical motion):

(2)
Now we examine the feasibility of using this model for velocity
estimation tasks. In [12], two criteria are quantified for an accurate velocity
estimation system: (1) image motion at a fixed velocity should always have
300 Unmanned Aerial Vehicles (UAV) and Drones

approximately response; (2) at a given velocity, the response to motion

should be unambiguous over certain range. In simulation, we find that by
introducing this modified model, the response to a specific velocity of image
motion can meet the two basic requirements at low velocities (above which
the response output is ambiguous). Thus, for velocity estimation tasks,
the motion velocity should be limited in a certain range due to essential
property (bell-shaped response) of the Reichardt model. In order to reduce
the brightness sensitivity, logarithmic transformation could be also applied
(as the modified model in [17]). However, by doing this the discrimination
of response is also highly reduced. That means, the response at a given
velocity cannot differ significantly from the response at other velocities,
which is not appropriate for quantifying velocity. Moreover, regarding the
brightness sensitivity problem, a stable lighting condition is demanded in
real-time experiments.
We firstly examine the open-loop characteristics of the system only for
yaw rate estimation. The quadrotor is tethered in the air, with an on-board
camera looking directly to the ground texture (the complete system is further
introduced in Section 5). This scenario in the indoor environment involves
a black-white chessboard ground texture. It is considered to be the most
suitable scenario for detecting rotation motion of a flying robot. Compared
to other forms of textures, the high image contrast can also help to improve
the discrimination for quantifying velocity (due to the characteristics of the
biological model itself). The quadrotor is rotated on horizontal level without
control, and we can get the relationship between the response output and the
rotation velocity (yaw rate). A lookup table is then built. Due to the system
noise and discrimination limitations of the experiments, the curve has some
nonmonotonic regions. The polynomial minimum quadric method is used
to fit the curve, (where di is a residual which is defined as the difference
between the predicted value yi and the actual value f (xi)):

(3)

MULTICAMERA 3D POSE ESTIMATION

Quadrotor is an underactuated vehicle. The 6 DOFs of the quadrotor are
controlled by four motor inputs (pitch, roll, thrust, and yaw) by varying the
lift forces and the torque balance through changing the rotating speed of the
rotors (see Figure 3(a)). Yaw control is realized by tuning the differential
 Visual Flight Control of a Quadrotor Using Bioinspired Motion Detector 301

speed between the two counterrotating rotor pairs. Increasing the rotating
speed of all the four motors at the same amount will cause an upward
movement of the quadrotor. When tuning the differential speed between the
two motors of either single rotor pair, the quadrotor will fly sideways.

Figure 3: Quadrotor dynamics. (a) the overhead view of a quadrotor; (b) six
degrees of freedom of the flying robot.
The work in this section is based on the former related work in [18]. In
this work, we set up an indoor GPS system by using multicamera tracking
instead of the former two-camera tracking. By tracking the four markers
installed on the axis of the quadrotor, the 3D position as well as the pose
of the flying robot can be estimated. The experimental setup of 3D tracking
is further introduced in Section 5. The frame of quadrotor dynamics is the
same as the in Figure 3(b). We have the following definition:
Marker i position vector: Si = (xi, yi, zi) T (i = 1, 2, 3, 4).
Central point vector between two nonadjacent markers: Mj = (xMj , yMj ,
zMj) T (j = 1, 2);
Estimated central point vector of the quadrotor: Mq = (xM, yM, zM) T ;
Orientation of marker i: Vi = (xVi , yVi , zVi) T (i = 1, 2, 3, 4);
The counting of the markers is clockwise, while the first marker is on
the main axis. For 3D pose control, the central point should be used as
the reference position of the quadrotor. The central points of the distance
between marker 1 and marker 3 as well as between marker 2 and marker
4 are M1 = (1/2)(S1 + S3) and M2 = (1/2)(S2 + S4). In consequence of the
marker’s noise through the tracking system, the two central points in the two
equations above are not identical. Thus, the central point of the quadrotor
is Mq = (1/2)(M1 +M2). The vectors between the central point and marker 1
for pitch as well as between the central point and marker 2 for roll are Vi =
302 Unmanned Aerial Vehicles (UAV) and Drones

Si − Mq (i = 1, 2). The values of pitch θ, roll φ, and yaw ψ angles can be then
calculated, and thus the 3D pose of the quadrotor can be estimated:

(4)

(5)

(6)

CONTROLLER
At first the quadrotor should be regulated to hover in the air on horizontal
plane with little shaking. That means, the stable state commands (us p for
pitch, us r for roll, us y for yaw, and us t for thrust) should be adjusted firstly.
Basing on these parameters, the control commands can be calculated next.
For each controller, we have an output value (u q p for pitch, u q r for roll,
u q y for yaw, and u q t for thrust) between −1 and 1, which is then added
with the corresponding stable state command. Since we only consider the
rotation movement in this experiment, which means, the quadrotor is not
always heading with the main axis towards X direction, the pitch and roll
commands should be adjusted in the ψ direction with a rotation matrix:

(7)
We choose proportional-integral (PI) controller for the yaw rate control.
The pitch, roll, and thrust commands are controlled by proportional-integral-
derivative (PID) controllers:
 Visual Flight Control of a Quadrotor Using Bioinspired Motion Detector 303

(8)
In this experiment, the reference values θ0 and φ0 are set to zero. The
desired altitude z0 is 0.35 m near hover. The measured values θ and φ are
calculated from the received data of the visual multicamera tracking system
using (4) and (5), whereas ψ˙ is searched out from a certain empirical lookup
table using the response value, which is calculated by insect-inspired motion
detectors. The yaw velocity can also be obtained from (6) by time derivative,
which is, for the heading stabilization, used as a reference (ground truth).
The closed-loop results will be shown and further discussed in Section 5.
The main loop in the whole software architecture (Figure 4) consists of
two simultaneous processes: yaw rate control using on-board camera visual
feedback and X, Y, and Z position/poses control using visual tracking system.
A graphical user interface (GUI) is developed basing on Qt crossplatform
application, which provides data visualization (e.g., 3D trajectory of the
quadrotor, battery voltage and sensory data information), commands input
and real-time online configuration of control parameters.

Figure 4: Software architecture.

304 Unmanned Aerial Vehicles (UAV) and Drones

EXPERIMENTS AND RESULTS

Experimental Setup
The whole experimental platform is shown in Figure 5. It mainly consists
of a quadrotor testbed, off-board workstation, and video camera tracking
system.

Figure 5: Experimental setup.

1) Quadrotor
The miniquadrotor used in this work is a “Hummingbird” with an “AutoPilot”
central control board from ascending technologies. It offers a 1 kHz control
frequency and motor update rate, which guarantees fast response to changes
in the environment. The size of the whole quadrotor testbed is 36.5 cm in
diameter, and the four rotors (each with a propeller size of 19.8 cm) are
directly driven by four high-torque DC brushless motors respectively.
Powered by a state-of-the-art 3-cell 2100mAh lithium-polymer battery, the
vehicle is able to hover up to 15 minutes (with about 120 g of payload in this
work).
2) On-Board Camera
Considering the limited payloads of the quadrotor, the PointGrey Firefly
MV CMOS camera which has a light weight (14 g) and a tiny size (25 mm ×
40 mm) is selected as the on-board camera. We choose a standard resolution
of 640 × 480 (pixels), and the frame rate is 60 Hz. The camera is equipped
 Visual Flight Control of a Quadrotor Using Bioinspired Motion Detector 305

with a 6 mm microlens, providing a viewing angle of 56 deg and 38 deg in

the length and width directions, respectively. It uses a 5-pin USB 2.0 digital
interface with a 480 Mb/s transfer rate and 8-bit raw Bayer data format
(connected through IEEE 1394 to workstation). The camera is mounted
under the base board of the quadrotor, looking directly down to the ground
texture.
3) Workstation and Communication Module
An off-board Linux PC (AMD Athlon 5200+; 2 GB RAM) is used for image
data processing, 3D pose estimation and control law execution in this case.
The quadrotor is equipped with XBeePro wireless communication module
from MaxStream/Digi, which enables the data transmission from the on-
board inertial measurement unit (IMU) and the control command reception
(with R/C transmitter enabled) from workstation at a rate of 100 Hz.
4) Visual Tracking System and Marker Placement
The tracking system VisualeyezII VZ4000 from Phoenix Technologies
Incorporated is used to get the absolute position of the quadrotor. It has three
cameras inside which can capture the certain markers installed on the four
axes of the quadrotor in an accuracy of millimeter level. In this work, the
tracking system is installed on the ceiling of the lab (Figure 5). The software
VZSoft is installed in another Windows PC (AMD Athlon XP 3000+;
2.1 GHz). It gets the data from the tracking system through a COM interface.
The data will be then sent to the workstation with the interface Babelfish
which is developed by the Institute of Autonomic Control Engineering,
Technical University Munich, using Internet Communications Engine (ICE).

Velocity Estimation
To validate the designed templates of receptive fields for rotation detection,
the bioinspired image processing algorithm is implemented with C++
language using Open CV.
Under low velocities and within certain altitude range, the response can
be regarded as monotonic and near linear from the test results. In this work,
the yaw rate is under 100 deg/s and the altitude value is set to 0.35 m. The
lookup table is shown in Figure 6(a), with a polynomial curve fitting. From
the comparison in Figure 6(b), this approach provides a fairly accurate yaw
rate estimation (the mean error is 1.85 deg/s and the standard deviation of
error is 10.22 deg/s). This lookup table could be then used in the closed-loop
control under the same light condition.
306 Unmanned Aerial Vehicles (UAV) and Drones

Figure 6: (a) lookup Table (LUT) in this work. This empirical LUT shows the
relationship between yaw rate (in low speed) and visual response under the ex-
perimental environment; (b) open-loop characteristics for rotation motion near
hover (without control).

Heading Stabilization
In this experiment, we compare the heading control performance using EMD
with those using IMU or tracking system respectively. At first the stable
commands (us p, us r, us y , and us t) should be determined experimentally,
so that without any controllers off board, the quadrotor can be hovering in
the air nearly on a horizontal level and rotating as little as possible, with all
the payloads mounted (in this experiment, with on-board bread board for
tracking system using TCM8 mode, and with cable power supply instead
of battery). The X, Y, and Z positions should be further controlled using the
feedback from the tracking system, while the yaw position has no controllers
 Visual Flight Control of a Quadrotor Using Bioinspired Motion Detector 307

except the on-board IMU at the first attempt. The IMU controller is already
integrated on the base board so that no other off-board controller is needed
for IMU controlling. A major disadvantage of using IMUs for navigation is
that they typically suffer from accumulated error (see Figure 7 blue curve).
In this case, in about 20 seconds the yaw position will deviate by 30 degrees
if only IMU is used for the heading stabilization. The second reference is the
yaw position with the control using tracking system, but without using EMD
(the red curve in Figure 7). In Figure 8, the 3D position when using EMD
for heading stabilization is shown. By using tracking system and EMDs (the
green curve in Figure 7) a satisfying performance could be both achieved.
Despite some deviation (for tracking system maximal ±5 degrees and for
EMD maximal ±7 degrees), the quadrotor can hover very well with straight
heading direction.

Figure 7: Heading stabilization.

308 Unmanned Aerial Vehicles (UAV) and Drones

Figure 8: 3D position measurements when the quadrotor is in stabilized hover

((a), (b), and (c) for X, Y, and Z positions, resp.). The 3D position is stabilized
by multicamera tracking system exclusively. The X and Y positions are set to
zero, which means the on-board camera is heading directly to the central point
of the ground texture as well as the origin of world reference coordinate. The
altitude (Z position) is set to 0.35 m.

Yaw Rate Control

The next step is to achieve the velocity control using EMDs. The yaw rate
should be set in a low-speed area considering a monotonic relationship
between response and velocity. In this case, the desired velocity is 30
degrees/s. The results are shown in Figure 9(a). The settling time is about
 Visual Flight Control of a Quadrotor Using Bioinspired Motion Detector 309

4 seconds and the maximal error is about ±10 degrees/s. The inflight
performance of 3D pose is shown in Figures 9(b) and 9(c). Since the IMU
provides only the angular velocity values, the angle positions are integrated
by the base control board on the quadrotor in order to get the angle positions,
which are sent to the central workstation.

Figure 9: Attitude measurements of the flying robot near hover with yaw-rota-
tion (a): yaw rate control. After takeoff, the quadrotor is switched to a desired
yaw rate of 30 deg/s; (b) and (c): pitch and roll angles of the quadrotor.
For velocity estimation, although the EMD is not a pure velocity
detector, a closed-loop control of yaw rate is achieved with restrictions of
the structured environment and the limitation of velocity in low-speed area.
Including image translation delay through IEEE1394/USB cable, image
processing by CPU costs 10–20 ms (the program takes nearly 10 ms to wait
for images from the camera, while the actual computing time is only several
milliseconds). So the EMD computing is extremely fast, which provides an
evidence of the efficiency when using biological models.
310 Unmanned Aerial Vehicles (UAV) and Drones

CONCLUSIONS AND FUTURE WORKS

In this work, the closed-loop control of a flying robot is achieved by using
bioinspired image processing method, which proves to be an effective
approach with low computational cost. For real-time implementation, the
experimental results of heading stabilization show that, by using EMD
response as a feedback, the accumulating drift from the on-board IMU
is compensated. Another trial regarding the EMDs as a velocity sensor‌
has realized a low-speed control of the yaw rate on the quadrotor in real-
world scenario. In the future, all the 6 DOFs should be controlled by using
bioinspired image processing exclusively, without relying on any off-board
visual sensors or GPS. The absolute position should be determined by certain
advanced algorithms. For the future works, some efforts should be put in the
development of novel approaches for highly robust flying performance.

ACKNOWLEDGMENTS
This work is supported in part by the DFG excellence initiative research
cluster Cognition for Technical Systems-CoTeSys, see also www.cotesys.
org, the Bernstein Center for Computational Neuroscience Munich, see also
http://www.bccn-munich.de/, and the Institute for Advanced Study (IAS),
Technische Universität München, see also http://www.tum-ias.de/.
 Visual Flight Control of a Quadrotor Using Bioinspired Motion Detector 311

REFERENCES
1. M. Egelhaaf and A. Borst, “Motion computation and visual orientation
in flies,” Comparative Biochemistry and Physiology A, vol. 104, no. 4,
pp. 659–673, 1993.
2. M. V. Srinivasan and S. W. Zhang, “Visual navigation in flying insects,”
International Review of Neurobiology, vol. 44, pp. 67–92, 2000.
3. H. G. Krapp and R. Hengstenberg, “Estimation of self-motion by optic
flow processing in single visual interneurons,” Nature, vol. 384, no.
6608, pp. 463–466, 1996.
4. A. Beyeler, J.-C. Zufferey, and D. Floreano, “3D vision-based
navigation for indoor microflyers,” in Proceedings of the IEEE
International Conference on Robotics and Automation (ICRA ‘07), pp.
1336–1341, April 2007.
5. J. Conroy, G. Gremillion, B. Ranganathan, and J. S. Humbert,
“Implementation of wide-field integration of optic flow forautonomous
quadrotor navigation,” Autonomous Robots, vol. 27, no. 3, pp. 199–
200, 2009.
6. F. Ruffier and N. Franceschini, “Optic flow regulation: the key to
aircraft automatic guidance,” Robotics and Autonomous Systems, vol.
50, no. 4, pp. 177–194, 2005.
7. N. Franceschini, F. Ruffier, and J. Serres, “A bio-inspired flying robot
sheds light on insect piloting abilities,” Current Biology, vol. 17, no. 4,
pp. 329–335, 2007.
8. F. Expert, S. Viollet, and F. Ruffier, “Outdoor field performances of
insect-based visual motion sensors,” Journal of Field Robotics, vol.
28, no. 4, pp. 529–541, 2011.
9. E. Buchner, “Behavioural analysis of spatial vision in insects,” in
Photoreception and Vision in Invertebrates, M. A. Ali, Ed., pp. 561–
621, Plenum Press, New York, NY, USA, 1984.
10. M. Egelhaaf and W. Reichardt, “Dynamic response properties of
movement detectors: theoretical analysis and electrophysiological
investigation in the visual system of the fly,” Biological Cybernetics,
vol. 56, no. 2-3, pp. 69–87, 1987.
11. M. V. Srinivasan, S. W. Zhang, M. Lehrer, and T. S. Collett, “Honeybee
navigation en route to the goal: visual flight control and odometry,”
Journal of Experimental Biology, vol. 199, no. 1, pp. 237–244, 1996.
12. R. O. Dror, D. C. O’Carroll, and S. B. Laughlin, “Accuracy of velocity
312 Unmanned Aerial Vehicles (UAV) and Drones

estimation by Reichardt correlators,” Journal of the Optical Society of

America A, vol. 18, no. 2, pp. 241–252, 2001.
13. S. Rajesh, D. O’Carroll, and D. Abbott, “Elaborated Reichardt correlator
for velocity estimation tasks,” in Proceedings of the Biomedical
Applications of Micro- and Nanoengineering, pp. 241–253, December
2002.
14. S. Han, A. Straw, M. Dickinson, and R. Murray, “A real-time helicopter
testbed for insect-inspired visual flight control,” in Proceedings of the
IEEE International Conference on Robotics and Automation, Kobe,
Japan, 2009.
15. A. Borst, C. Reisenman, and J. Haag, “Adaptation of response transients
in fly motion vision. II: model studies,” Vision Research, vol. 43, no.
11, pp. 1311–1324, 2003.
16. B. Hassenstein and W. Reichardt, “Systemtheoretische Analyse der Zeit-
Reihenfolgen, und Vorzeihenauswertung bei der Bewegungsperzeption
des Ruesselkaefers Chlorophanus,” Naturforsch, vol. 11, pp. 513–524,
1956.
17. T. Zhang, H. Wu, A. Borst, K. Kuehnlenz, and M. Buss, “An FPGA
implementation of insect-inspired motion detector for high-speed
vision systems,” in Proceedings of the IEEE International Conference
on Robotics and Automation, Pasadena, Calif, USA, May 2008.
18. M. Achtelik, T. Zhang, K. Kuhnlenz, and M. Buss, “Visual tracking
and control of a quadcopter using a stereo camera system and inertial
sensors,” in Proceedings of the IEEE International Conference on
Mechatronics and Automation, Changchun, China, 2009.
 SECTION 4
APPLICATIONS OF UAV-S
AND DRONES
 CHAPTER

14
Development of Rescue Material
Transport UAV (Unmanned Aerial
Vehicle)

Daeil Jo, Yongjin Kwon

Department of Industrial Engineering, College of Engineering, Ajou
University, Suwon, South Korea

ABSTRACT
Recently, the market for drones is growing rapidly. Commercial UAVs
(Unmanned Aerial Vehicles, or drones) are increasingly being used for
various purposes, such as geographic survey, rescue missions, inspection
of industrial facilities, traffic monitoring and delivery of cargos and goods.
In particular, the drones have great potential for life-saving operations. A

Citation: Jo, D. and Kwon, Y. (2017) “Development of Rescue Material Transport

UAV (Unmanned Aerial Vehicle)”. World Journal of Engineering and Technology, 5,
720-729. doi: 10.4236/wjet.2017.54060.
Copyright: © 2017 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
316 Unmanned Aerial Vehicles (UAV) and Drones

missing person, for example, can be rapidly and effectively searched using
a drone in comparison with the conventional search operations. However,
there is no commercially available rescue UAV until now. The motivation for
this study is to design an unmanned aerial vehicle capable of vertical takeoff
and landing, while containing a high power propellant apparatus in order to
lift a heavy cargo that contains rescue materials (such as water, food, and
medicine). We used the EDF (Electric Ducted Fan) technology as opposed
to the conventional motor and prop combination. The EDF can produce
the power about three times higher than the motor-prop combination. This
became suitable for transportation of rescue goods, and can be widely used
in rescue operations in natural environments. Based on these results, the
UAV for rescue material transport capable of heavy vertical takeoff and
landing is developed, including airframe, flight control computer and GCS
(ground control station).
Keywords: High-Powered Propellant, Vertical Take-Off and
Landing, Rescue Drone, GCS (Ground Control Station), FCC (Flight
Control Computer)

INTRODUCTION
In recent years, the use of drones or UAVs (unmanned aerial vehicles) is
increasingly getting popular around the world. Drones are utilized in aerial
photography for both personal hobbies and commercial uses. In industrial
sectors, drones are deployed in facility inspection, power line inspections,
and monitoring of fire and flood purposes. In almost every imaginable part,
the drones are being used [1] [2] . The rescue operation is in no exception.
Due to its fast flying characteristics, which can cover a wide area in less
time, drones are now being used by law enforcement agencies for search
missing persons. The current rescue drones are typically equipped with
thermal image cameras; hence it can detect the missing person from the
surroundings, even though the missing person is not visually identifiable from
the surrounding environment. However, there is no rescue drones that can
carry the emergency materials, such as water, food, blanket, and medicine,
and being capable of conducting rescue operations [3] [4] [5] . In order to
do this, the drones must have a high power propellant technology that is
different from the conventional motor and prop combination. Additionally,
the rescue drone should be able to vertically takeoff and land, while carrying
the rescue materials. Therefore, the motivation of this study is to develop a
rescue drone that is capable of conducting a search and rescue operation with
 Development of Rescue Material Transport UAV (Unmanned ... 317

onboard carbo capacity. Through this development, we design and produce

the actual airframe and cargo handling device, set up the firmware SW of
flight control computer, and run a flight stability test to validate the safe
flying characteristics. We will actually build a structural drone that is still in
the concept phase and prepare a technology base so that a large number of
IPs can be created in the future.
Currently, various hobby and surveillance drones are being developed
and commercially available in Korea, but no high power drones are developed
yet, with the high load carrying capacity. Most hobby drones cannot carry
enough load, and even the large size drone can only carry less than 1 or
2 Kilograms. The biggest agricultural drones can carry about 10 to 15
Kilograms of payload, but they are too large for police rescue operations [6]
[7] . In this context, the drone should be small enough to fit in the police car
and easy to handle. At the same time, the drone should be generating enough
lifting power, so that it can carry the rescue materials for the stranded or
missing person. This will complement the shortcomings of the conventional
UAV that is lacking the power capacity. As the demand for UAV increases in
the future, it will lay the foundations for developing various types of UAVs
(courier, transportation, delivery, lifeguard, etc.) that can meet the customer
demands.

DEVELOPMENT OF SOFTWARE AND HARDWARE

Motors and Propellers

The drones that we want to make in this study basically have to deliver relief
goods to people located in the middle of the building in case of fire and
emergency situation in high-rise building. At this point, the most important
point is that the drones must be close to the outer wall of the building to
deliver relief supplies. In case of EDF (Electric Ducted Fan), it is possible
to approach close to the building because the rotating prop is contained
inside the turbine housing. However, when the conventional propeller
motor is used, the drone is difficult to approach the building because the
exposed propeller may hit the structure and the drone will fall from the
sky. Ultimately, this difference is the biggest reason to use the EDF motors
in developing the rescue drone [8] [9] [10] . There are also considerations
when comparing EDF and propeller motors. It is whether or not you can
protect the safety of your finger when you fly it. Due to the nature of the
relief donor drone, the person must directly reach out and remove the relief
318 Unmanned Aerial Vehicles (UAV) and Drones

item from the box in the lower part of the drone. At this time, according
to different rotation method of EDF motor and propeller motor, the safety
protection of the finger is different. In other words, EDF drone do not have
a risk of injuring your fingers unless you intentionally put your fingers into
the turbine housing. However, in propeller motors, the propeller itself is
located outside the drones, so there is a risk that the fingers can be cut any
time. This means that if a propeller motor is used, the drones used for the
rescue may cause secondary damage. Because of this, we have chosen a high
thrust EDF, instead of conventional moto and prop combination. Electronic
Ducted-Fan type airplane itself consists of engine, propeller, fixed wing and
flap. Theoretically, moving the fluid flow from the rotating body to the duct
increases the thrust efficiency by 41%. However, the weight of ducts to be
increased in comparison with the propeller should also be considered. The
aim of this study is to lift a total weight of about 10 kg, which has four
EDFs, each of which can maintain the duration of 50% of the maximum
thrust at a weight of up to 18 kg, as shown in Figure 1.

Battery Selection
The battery is a major factor in controlling the drones’ flight time and
output. The larger the capacity of the battery, the longer the flight time, and
the higher the number of cells in a series connected battery. The selected
EDF requires a higher voltage than the propeller and motor combination,
so we select a battery with 6 - 8 cells. The battery has a high stability and
uses a lithium ion polymer material. This is because the electrolyte of the
lithium ion battery is a liquid and basically has a problem in stability. We
also considered trade-offs in battery capacity and weight. Electric vehicles’
biggest dilemma is the proportional relationship between the capacity and
weight. When a heavy battery is used in a

Figure 1. Selected EDF: Highly safe, theoretically efficient EDF around the
wing with Duct. Maximum thrust per EDF 4.5 kg × 4 = 18 kg Expectations,
excluding frame and self weight-Estimated thrust at least 8 kg/900 g × 4 = 3.6
kg (left), EDF Specification (right).
 Development of Rescue Material Transport UAV (Unmanned ... 319

EDF with a duct, the weight of an automatic cardioverter defibrillator

(AED), a fire mask, a lantern, and other rescue materials need to be reduced.
Also, the weight of the drone itself can be increased significantly. The
selected battery is PT-B10000-NSR35 model and has specifications of 22.2
V, 6S1P, 35C+, as shown in Figure 2. To meet the selected EDF output and
minimum duration, the total capacitance that can be supplied by connecting
two batteries in parallel was increased.

Frame Design
Four EDFs are designed in four directions to balance the drones. Fixtures for
installing ESC (electric speed controller), battery holder, GPS transceiver
and other fixtures are installed on the center top plate. In addition, a fixture
for installing a structural goods transport box is installed in the lower part,
and two supports for stably loading and unloading the drones are installed.
The material of the frame was made of carbon material to minimize the
weight of the drone, and some aluminum parts were also used, as shown
in Figure 3. A cylindrical EDF fixture using a 3D printer was designed as an
EDF dedicated frame for fixing each EDF. In the end, thanks to the frame
made of carbon material, it was able to create a frame that can withstand a
high-output EDF, while lightening the weight as much as possible.

Figure 2. Selected batteries: Selecting a battery that can supply the appropriate
voltage for the EDF output (400 g × 2 = 800 g).

Figure 3. Frame drawing and specification: Carbon frame selection/2.2 kg

(left), EDF fixture design drawing (center), Drone assembly and fabrication
process (right).
320 Unmanned Aerial Vehicles (UAV) and Drones

Design and Manufacture of Rescue Material Box

In order to minimize the imbalance in the flight of the drones, a rectangular
box-shaped model penetrating the bottom of the drones was adopted. The
vertical length is the minimum length, which is larger than the length of the
frame including the EDF to avoid interference with the EDF. The height is
defined as the maximum length that can be installed between the support
and center of the drones frame. In order to deliver the rescue supplies, the
drones must be approached through the windows and the being rescued
person must reach the arm and bring the rescue supplies. Since the drones
are not fully accessible to the building, the length of the arm of the person
should be considered with the minimum length. In this case, it is necessary
to design the arm length so that it is not difficult for both a person with a
shorter arm and a person with a longer arm to approach, as shown in Figure
4. The design should be based on the average (498 mm) of the arm length
first decile (lowest quartile) for women between 10 and 80 years of age.
Height, width, and length are based on the minimum and maximum length.
The material called “Foam-max” was considered because of the weight
and the strength of the box itself. Foam-max is foamed and compressed by
using PVC as raw material and it is very easy to process while maintaining
the advantages of plastic and wood. It was to create boxes and to make
room for speakers in the middle. Speakers installed in the drones are able to
communicate with the stranded person who are in need. It is also possible
to reserve extra batteries in case of remaining space. In addition, a partition
was installed to prevent internal objects from moving during the flight, thus
preventing the center of gravity of the drones from being disturbed [11] [12]
.

Figure 4. Prototype of rescue material box.

 Development of Rescue Material Transport UAV (Unmanned ... 321

Type of Rescue Material and Delivery Method

In the case of storage space for rescue materials, it is possible to store about
1 kilogram of materials. It is possible to store items according to the purpose
and to greatly increase the survival rate in case of fire, such as oxygen
cans, masks, firefighting clothes, fire extinguisher, water, water towel, and
so on. Structural rope, radio, and emergency medicine can also be stored.
In addition, if a cardiac arrest patient is present, the attached box can be
removed and the cardiac defibrillator, portable AED, attached to the lower
part of the drone can be delivered. When a high-rise building is delivered
to a rescue target in the middle or upper floors, it approaches the building
window and draws the rope connected to the storage bag connected to
the storage box to deliver the rescue goods. As illustrated in Figure 5, the
operation procedure is as follows:
1) Pull out the line outside the box.
2) At the same time the door is opened, the rescue materials come out.
3) Hold the rescue supplies.
4) Pull the rescue supplies.

RESEARCH AND DEVELOPMENT RESULTS

The details of the developed drone is given in Figure 6. The drone was tested
and evaluated for its flight characteristics. Figure 7 shows the test flight of
takeoff and landing. The test was proceeded smoothly and the flight was
very stable.In this study, EDF was selected because of the safety problem of
the propeller, but it was not able to produce 4.5 kg of thrust as indicated in
the product specification. 4.5 kg When 4 EDFs were used, the specification
gave 18 kg of thrust.

Figure 5. Rescue material transport system.

322 Unmanned Aerial Vehicles (UAV) and Drones

Figure 6. Developed drone with structural box and relief items and its specifi-
cation.

Figure 7. Drones test flight: takeoff and landing.

Except for the weight of the body, 10 kg of thrust was expected, but
the actual thrust was only 10.5 kg. And a run-time of 1/10th compared to
propeller aircraft capable of flying over 30 minutes.
Thrust determination equation is as follows [9] [10] [11] [12] .

The equation to determine the propeller was as follows.

T = thrust [N]
D = propeller diameter [m]
ν = velocity of air at the propeller [m/s]
Δν = velocity of air accelerated by propeller [m/s]
ρ = density of air [1.225 kg/m3]
The thrust ratio was calculated by excluding the air velocity (v) related
factor through the thrust formula. And when using a propeller, it was calculated
that four times as much thrust was possible. This is because the diameter of the
propeller has the greatest influence on thrust. To develop an EDF that produces
 Development of Rescue Material Transport UAV (Unmanned ... 323

the same thrust as a 4× long propeller, it required 16 times the revolutions per
second, which drastically reduced battery life. In addition, it required a high
voltage of 30 V compared to a propeller requiring 22.2 V. In the case of a
commercially available battery, the maximum voltage was 6s (3.7 V × 4 =
22.2 V). Finally, the weight of the EDF itself was 900 g, totaling 3.6 kg. It
was about 9 times the weight of the propeller and I think it was the biggest
weight gain factor in the airframe design that we are trying to reduce 1g. In
order to compare the exact thrust between the propeller and the EDF, I think
it would be necessary to consider vortex elimination and air compression rise
through the duct. In order to actually introduce EDF, which has the advantage
of safety, we need three things: 1, lighter weight, 2. efficiency, and 3. propeller
size [13] [14] [15] [16] . Figures 8-12 shows the flight characteristics data, all
of which shows a very stable flight characteristics.

Figure 8. EDF and propeller comparison.

Figure 9. Z-axis acceleration data.

Figure 10. Bank angle data.

324 Unmanned Aerial Vehicles (UAV) and Drones

Figure 11. Pitch angle data.

Figure 12. Speed data.

CONCLUSION
UAVs capable of high-power, vertical takeoff and landing can complement
the disadvantages of conventional UAV’s lack of power and maximize
utilization throughout the industry. Through this technology development, we
localize the core technology of UAV that can control high power and vertical
flight stably. As the demand for UAV increases in the future, it will lay the
foundations for developing various types of UAVs (courier, transportation,
delivery, lifeguard, etc.) that meet customer demand. Expected effects of
meeting domestic demand, import substitution and export will be created. In
the future, we will complement the problems analyzed in this study and design
a new type of drone. In the end, we will develop more than 25 kg payload so
that it can be utilized for life-saving and transportation of rescue materials.

ACKNOWLEDGEMENTS
This work was supported by the Ajou University research fund.

CONFLICTS OF INTEREST
The authors declare no conflicts of interest.
 Development of Rescue Material Transport UAV (Unmanned ... 325

REFERENCES
1. Yu, S. and Kwon, Y. (2017) Development of VTOL Drone for Stable
Transit Flight. Journal of Computer and Communications, 5, 36-43.
https://doi.org/10.4236/jcc.2017.57004
2. Yu, S. and Kwon, Y. (2017) Development of Multi-Purpose, Variable,
Light Aircraft Simulator. Journal of Computer and Communications, 5,
44-52. https://doi.org/10.4236/jcc.2017.57005
3. Jo, D. and Kwon, Y. (2017) Analysis of VTOL UAV Propellant
Technology. Journal of Computer and Communications, 5, 76-82.
https://doi.org/10.4236/jcc.2017.57008
4. De la Torre, G.G., Ramallo, M.A. and Cervantes, E.G. (2016) Workload
Perception in Drone Flight Training Simulators. Computers in Human
Behavior, 64, 449-454.
5. Liu, Z. and Li, H. (2006) Research on Visual Objective Test Method
of High-Level Flight Simulator. Xitong Fangzhen Xuebao (Journal of
System Simulation).
6. Luan, L.-N. (2013) Augmenting Low-Fidelity Flight Simulation
Training Devices via Amplified Head Rotations. Loughborough
University.
7. Lee, S. and Lim, K. (2008) PCB Design Guide Book. Sehwa Publishing
Co., South Korea.
8. Kang, M. and Shin, K. (2011) Electronic Circuit. Hanbit Media, South
Korea.
9. Kim, D., Kim, J. and Yoon, S. (2016) Development and Validation
of Manned and Unmanned Aircraft Simulation Engine for Integrated
Operation in NAS. Journal of the Korean Society for Aeronautical &
Space Sciences, 44, 423-430.
10. Sponholz, C.-B. (2015) Conception and Development of a Universal,
Cost-Efficient and Unmanned Rescue Aerial Vehicle. Techische
Hochschule Wildau Technical University of Applied Sciences.
11. Guerrero, M.E., Mercado, D.A. and Lozano, R. (2015) IDA-PBC
Methodology for a Quadrotor UAV Transporting a Cable-Suspended
Payload. 2015 International Conference on Unmanned Aircraft
Systems (ICUAS), Denver, CO, 9-12 June 2015, 470-476. https://doi.
org/10.1109/ICUAS.2015.7152325
12. Gomez, C. and Purdie, H. (2016) UAV-Based Photogrammetry and
Geocomputing for Hazards and Disaster Risk Monitoring—A Review.
326 Unmanned Aerial Vehicles (UAV) and Drones

Geoenvironmental Disasters, 3, 1. https://doi.org/10.1186/s40677-

016-0060-y
13. Lee, E., Kim, S. and Kwon, Y. (2016) Analysis of Interface and Screen
for Ground Control System. Journal of Computer and Communications,
4, 61-66. https://doi.org/10.4236/jcc.2016.45009
14. Kwon, Y., Heo, J., Jeong, S., Yu, S. and Kim, S. (2016) Analysis of
Design Directions for Ground Control Station (GCS). Journal of
Computer and Communications, 4, 1-7. https://doi.org/10.4236/
jcc.2016.415001
15. Yu, S., Heo, J., Jeong, S. and Kwon, Y. (2016) Technical Analysis of
VTOL UAV. Journal of Computer and Communications, 4, 92-97.
https://doi.org/10.4236/jcc.2016.415008
16. Hong, J., Baek, S., Jung, H., Kim, S. and Kwon, Y. (2015) Usability
Analysis of Touch Screen for Ground Operators. Journal of
Computer and Communications, 3, 133-139. https://doi.org/10.4236/
jcc.2015.311021
 CHAPTER

15
Railway Transport Infrastructure
Monitoring by UAVs and Satellites

Sergey I. Ivashov, Alexander B. Tataraidze, Vladimir V. Razevig,

Eugenia S. Smirnova
Remote Sensing Laboratory, Bauman Moscow State Technical University,
Moscow, Russia

ABSTRACT
Improving the rail transport security requires development and implementation
of neoteric monitoring and control facilities in conditions of increasing speed
and intensity of the train movement and high level of terrorist threat. Use
of Earth remote sensing (ERS), permitting to obtain information from large

Citation: Ivashov, S., Tataraidze, A., Razevig, V. and Smirnova, E. (2019), “Railway
Transport Infrastructure Monitoring by UAVs and Satellites”. Journal of Transporta-
tion Technologies, 9, 342-353. doi: 10.4236/jtts.2019.93022.
Copyright: © 2019 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
328 Unmanned Aerial Vehicles (UAV) and Drones

areas with a sufficiently high resolution, can provide significant assistance

in solving the mentioned problems. This paper discusses the possibility of
using various means of remote sensing such as satellites and unmanned aerial
vehicles (UAV), also known as drones, for receiving information in different
ranges of the electromagnetic spectrum. The paper states that joint using of
both these means gives new possibilities in improving railroad security.
Keywords: Transport Infrastructure Monitoring, Remote
Sensing, Satellite, Unmanned Aerial Vehicle (UAV), Aerial
Photography, Radar Sensing, 3D Image Processing

INTRODUCTION
The railway transportation safety demands constant attention from the staff
and management of company that operates the facility. This applies to the
locomotive drivers’ condition monitoring, as well as to all services related
to ensuring the smooth operation of the railway: manifestation that is called
the human factor. These problems can be solved by a set of organizational
and technical measures that improve discipline and working conditions of
employees.
Unfortunately, there are still some technical factors, such as violation
of the railway line integrity (rail rupture, destruction of railway arrow, etc.),
which also require permanent monitoring. The another thing to mention is
a problem of countering terrorist acts with the use of improvised explosive
devices, which can lead to even more dire consequences. The use of remote
sensing means such as piloted helicopters and drones could render substantial
assistance in determining the scale of a railway catastrophe and ways of its
elimination, including assistance to possible victims [1] [2] [3].
There is one more task connected to the topic of improving the safety
of rail transportation. These are incidents related to natural phenomena
including mudflows and avalanches, heavy snowfalls in the plains, and some
others. Remote monitoring such phenomena could effectively help in its
prediction and consequences management. To monitor and control the state
of snow cover on the mountain slopes in avalanche prone areas, the UAV
could be successfully used for reconstructing three-dimensional images
of the terrain in the summer (without snow cover) and in the winter after
heavy snowfalls. After comparison with such images. it will be possible to
determine the height of the snow cover, and UAVs equipped with passive
radar (radiometer) operated in centimeter or decimeter wavelengths will
provide an opportunity to measure snowpack humidity [4]. These data could
 Railway Transport Infrastructure Monitoring by UAVs and Satellites 329

be added by measurements obtained from a ground-based radar network

[5]. The combination of these factors (stockpile and humidity of snow) will
allow creating a fairly accurate forecast of avalanches probability.
The railway is a transporter of dangerous consignments such as
explosives, perilous for environment and poisonous chemical compounds,
combustible and flammable materials. In case of an accident, it is necessary
to promptly assess the scale of the disaster and make decisions on the prompt
elimination of its consequences. The described examples widely show
possible applications of UAVs and other remote sensing tools, including
earth observation satellites, which can be used to ensure the safety of rail
transportation.The main goal of current paper is a comparison of the satellite
and drone effectiveness in recording images of railway infrastructure in
visual and microwave frequency range, and gives recommendation in their
joint using. This matter was not analyzed before and main emphasis was
mostly done on drones’ applications in infrastructure inspection [6].

ADVANTAGES AND DISADVANTAGES OF REMOTE-

LY PILOTED AIRCRAFTS
Recent years have been characterized by an increasing use of UAVs with
the expansion of tasks ranges, which are solved with their help. While in the
past UAVs were used exceptionally for military purposes, now their civilian
use is becoming extremely important. Military original interest to drones,
which arose in the late 60s, was stipulated by a desire to avoid the risk
for the crew lives during reconnaissance flights. The disadvantage of these
drones was the need of returning to “home” territory after the flight mission
for the treatment of the recorded photographic films and decryption of the
material. That reduced the relevance of obtaining information.
Hereinafter, the most successful development of UAVs was implemented
in Israel (tactical and operational UAVs) and the United States (tactical and
strategic UAVs). Progress in the designing a new generation of reconnaissance
UAVs is primarily associated with the emergence of digital photography,
including cameras functioning in different wavelength ranges, navigation
and communication through satellites. This significantly improved the
performance of on-board equipment, especially the weight and size, and
allowed improving the parameters of the UAV such as endurance, range and
height of flight. It should also be noted that the cost of the UAV has sharply
decreased, and access to satellite information has been improved in recent
years.
330 Unmanned Aerial Vehicles (UAV) and Drones

Modern drones for civil purposes are characterized by low operation

cost and no need for stationary airfields with paved runways. As a rule, they
are launched from simplest ground airfields, mobile catapults or directly
from the operator’s hand. Remarkable progress has been achieved in the
development of multi-rotor drone helicopters with electric power. UAVs of
this type are characterized by simplicity of design and affordability. This is
mainly due to the lack of such an essential for conventional helicopters a
structural element as rotor hinge system, and the use asynchronous control
of electric motors traction, that allows running the movement of the device
in flight. Figure 1 shows a six-rotor UAV helicopter, which was used to
obtain images of the Experimental Railway Ring (ERR) in Shcherbinka,
Moscow region, in current project. Advantage of this type of UAV is also
the fact that now their use in the exclusion zone of the railways at altitudes
of less than 100 m does not require special permission for flights.
The disadvantage of multi-rotor electrical UAVs is a relatively short flight
time usually less than an hour, and consequently restrictions on range and
altitude of the flight. However, this is not essential when the operator con-
trolling UAV flight is directly placed on the action place or near it. In this
case, information about objects and events happening on earth surface can
be obtained in real time and is compared with operator’s own vision, what
is essential, for example for making decisions in the elimination of railway
accident consequences. An increase of the drone flight time can be achieved
by replacing the batteries during intermediate landings, although the range
limit remains, and it is still sometimes a significant factor.

Figure 1. Helicopter-type UAV: (a) The UAV during the flight with a camera
mounted in the gimbal; (b) The operator with the remote control of the UAV on
the background.
 Railway Transport Infrastructure Monitoring by UAVs and Satellites 331

ANALYSIS OF SATELLITE RESOURCES FOR OB-

TAINING INFORMATION
In the 1960s and 1970s the main users of photographic and radar satellite
information were military, and this information itself was classified, but in
recent decades a significant segment of consumers is the highly profitable
civilian sector. In the first instance to the tasks to be solved by Earth
observation satellites the following items should be included:
• Research of natural resources;
• Prediction, analysis and control of emergencies and their
consequences;
• Geological exploration;
• Weather forecast;
• Control of the environment and its sources of pollution;
• Agriculture and prediction of harvest;
• Forestry and forest protection from fire and pests;
• Construction and other industries.
Remote sensing by satellites registers the radiation of the Earth’s surface
in different ranges of the electromagnetic spectrum. For above mentioned
tasks multispectral photography and side-looking radars are mostly used. In
some cases the lasers are used especially in surveying for the measurement of
Earth’s surface topography. Remote sensing data obtained from a spacecraft
equipped with visible-range photographic cameras is characterized by a high
degree of dependence on the atmosphere transparency and the times of day.
Therefore, in order to remove these restrictions, remote sensing satellites
equipped with side-looking radars have been widely developed. Radar
image has worse resolution in comparison with optical one, but it has an
ability to obtain images irrespectively of the weather and time of day, which
is crucial for example for the navigation of ships on the Northern Sea Route
during the polar night. Currently remote sensing satellites together with
communication and navigation satellites constitute a significant segment
of commercial space industry services. Often remote sensing satellites
are created by organizations associated with the military industry of their
own countries, which decreases the cost of their development due to the
continuity of design solutions with military reconnaissance satellites.
Remote sensing satellites are usually launched into the so-called sun-
synchronous orbit (sometimes called heliosynchronous), which has such
332 Unmanned Aerial Vehicles (UAV) and Drones

parameters that the satellite on this orbit, passes over any point of the
earth’s surface at about the same local solar time. Consequently, the angle
of sun over the earth’s surface is approximately the same at all paths of
the satellite above the same earth’s place. Permanent lighting conditions
are very convenient for use in satellites receiving optical images of the
earth’s surface including remote sensing and meteorological satellites. The
parameters of the sun-synchronous orbits fall within the range: the height
above the earth’s surface is 600 - 800 km, the period of rotation is 96 - 100
min, and the inclination of the orbit is about 98˚.
The main advantage of the remote sensing optical satellites is the ability
to obtain high-quality multispectral images, which resolution can reach 0.3
m or lesser. Nevertheless, visual spectrum satellites are not able to function
in night or cloudy weather. While the first shortcoming can still be potentially
overcome, for example, by using infrared optics, the cloudiness, which in
mid-latitudes in winter can hide the earth for weeks or even months, makes
impossible using even infrared satellites. All-weather satellites are radar
ones, even though they have the worst resolution. At the same time, radar
provides additional opportunities, for example, obtaining images of the
recorded signal in different polarizations or finding underground objects [7].

THE DATABASES OF SATELLITE INFORMATION

Currently, the results of satellite photography are widely distributed due to
the accessibility and simplicity of work with them. For example, several
sites in the Internet such as Google Maps (https://www.google.com/maps),
Yahoo! Maps (https://maps.yahoo.com/) and others provide free access to
photos recorded by satellites.
Satellite images and topographic maps based on these images are widely
used even for everyday household tasks as navigating private cars or calling
a taxi. Some websites provide only the results of satellite imagery allowing
working with databases of photographs: NASA World Wind, TerraServer-
US, Space image, Landsat Look Viewer (USGS).
Although free information services provide images with a sufficiently
high resolution, they still do not give an opportunity to obtain actual to the
moment information because data in them are rarely updated.
Photos for that websites were usually recorded in the warm season
without the snow cover. Such images can be used for some types of
analytical and theoretical works, but their drawbacks do not allow obtaining
 Railway Transport Infrastructure Monitoring by UAVs and Satellites 333

timely and relevant information about the event or object of interest. For
getting actual information, you should contact the specialized organizations
and place the order for recording satellite information. This is a chargeable
service but it allows the customer getting information according to his own
needs: time, place and conditions of photography, resolution of images and
other parameters.
The information may be provided as up-to-date one, i.e. recently received
but also can be extracted from the database accumulated from previous
years. The latter is important for information used in environmental studies,
where it is significant to compare changes in the landscape over long period.

COMPARISON OF INFORMATION RECEIVED

FROM DIFFERENT SOURCES
To compare satellite and airborne images of the railway transport
infrastructure, the Experimental Railway Ring in Shcherbinka was selected.
The choice of this object was related to the following circumstances.
Firstly, ERR is well studied and located close to Moscow. Secondly, there
were no difficulties with getting the permission for UAVs flights over
the ERR territory because of its experimental status. Figure 2 shows the
Experimental Ring image taken from the Google Maps database.
The center of the ERR ring is tagged with an arrow, its coordinates are
55.521373, 37.550190 (Northern latitude and Eastern longitude in degrees
accordingly). Inside the Ring, a lot of summer country houses are located.
The Experimental Railway Ring itself includes two double-tracks rings:
outer and internal.
The outer ring has the shape of a regular circle, and the inner ring
contains a straight section.
A rail depot is visible in the lower right corner of this picture. A rectangle
marks flight area of UAV which photos are used in this article. The another
arrow in the rectangle marks the location of a water tower, which served as
a landmark for UAV’s flights.
The enlarged image of the water tower obtained from the satellite is
shown in Figure 3.
This is maximal scale that the Google Maps site allows. Roof of the
tower has a conical shape and is covered with galvanized steel sheets. Joins
between the sheets have the form of vertical ribs. The ribbed shape of the
roof is visible even in the satellite image.
334 Unmanned Aerial Vehicles (UAV) and Drones

Figure 2. Image of the railway experimental railway ring from Google Maps
free database.

Figure 3. Large-scale satellite image of the water tower area on the experimen-
tal railway ring according to Google Maps.
The same area near the water tower was shot from the board of the
UAV equipped with a Canon EOS 5D Mark II camera with EF-S 17-55
f/2.8 IS USM lens. Camera was mounted on a gimbal controlled by an
operator, Figure 1. The result of the photographing from flight altitude about
70 m is shown in Figure 4. The difference between these two images consists
in the location of the rail wagons on the tracks near the water tower that
caused by different time of the shooting. The satellite image was recorded
in summer, and drone flights were performed in autumn. They also differ
 Railway Transport Infrastructure Monitoring by UAVs and Satellites 335

in size and direction of the local objects shadow, as the images were taken
at different positions of the sun on the celestial sphere. This fact stresses
the importance of sun-synchronous orbits, which were mentioned earlier
and are commonly used for remote sensing satellites. Another advantage of
using sun-synchronous orbits is the ability to determine whether an object
is being built or not, and at what speed it is being built, judging by the
length of the shadow at the images obtained from different orbit passes of
the satellite. Usually this method is used for the interpretation of images
recorded by military reconnaissance satellite, but it also can be used for
cadastral surveying, when it is necessary to allocate from a large array of
images new and illegally erected buildings. In that case, to determine the
newly erected buildings and their height change it is enough to subtract the
images.
If we compare the resolution of the two above photos, it is obvious that
the UAVs image has a much better resolution than satellite one. This is clear
since the altitude of the orbits of remote sensing satellites is about 600 km
in comparison with the same parameter for drones that equals a few hundred
meters or several kilometers. UAVs have another advantage over satellites
because they are less dependent on weather conditions. In the presence of
cloudiness, it is impossible to shoot in the visible spectrum from satellites.
While UAVs can hover below the lower cloud boundary and take pictures
of the terrain.

Figure 4. Photograph of the water tower area on the ERR recorded by the UAV.
336 Unmanned Aerial Vehicles (UAV) and Drones

As mentioned earlier, only radar imaging is weather-independent. Two

low-resolution images of the Experimental Ring district obtained from the
Google Maps website and the European Sentinel-1 radar satellite database
(Sentinel-1, https://sentinel.esa.int/web/sentinel/missions/sentinel-1 /
observation-scenario/archive) are presented in Figure 5.
At the lower part of both images Ostafievo airport (left in the pictures)
and ERR (right) are seen.
Both presented images have approximately the same scale, but the
visible-range image contains more detail than the radar one.
It is worth mention that radar image clearly depicts runways of the
airport. The scale of both images can be easily estimated, because a diameter
of the outer circle of the railway testing area is 1912 m.
The main advantage of radar, as it was mentioned earlier, is the ability to
record images during the whole day and in all-weather conditions. In some
cases, radar images may be the only source of information.
In this regard, it can be concluded that to solve various problems in the
interests of railway safety, it is important to have access to different sources
of remote sensing information.

RECONSTRUCTION OF 3D MODELS OF THE RAIL-

WAY INFRASTRUCTURE OBJECTS FROM A SET OF
2D IMAGES
Modern image processing software allows 3D environment model
reconstructing from a set of 2D images. This gives an ability to view objects
from different angles and heights.
Such technology can be used in designing complex objects of the railway
infrastructure (bridges, tunnels, etc.), as well as in solving problems of their
safety.
In the recent time, the increasing interest in using 2D aerial images,
collected by the UAV, in 3D model reconstruction is evident.
It is very important to note that the UAV simplifies the process of taking
images because operator can set the trajectory of the vehicle individually [8].
The multiple images are acquired by on board high resolution camera
from multiangle and multiposition during the flight of the UAV.
 Railway Transport Infrastructure Monitoring by UAVs and Satellites 337

Figure 5. Satellite images of the Experimental Railway Ring neighborhood: (a)

low-resolution visible spectrum image the according to Google Maps; (b) radar
image recorded by the Sentinel-1 satellite, EU.
The essence of an image is a projection from a 3D scene onto a 2D
plane, during which process the depth is lost. The 3D point corresponding to
a specific image point is constrained to be on the line of sight. From a single
image, it is impossible to determine which point on this line corresponds to
the image point. If two images are available, then the position of a 3D point
can be found as the intersection of the two projection rays. This process is
referred to as triangulation.
The key for this process is the relations between multiple views which
convey the information that corresponding sets of points must contain some
structure and that this structure is related to the poses and the calibration
of the camera. The task of converting multiple 2D images into 3D model
consists of a series of processing steps which brief overview is given here.
For a more extensive treatment the reader is referred to textbooks such as
[9] [10].
Camera calibration consists of internal (focal length, coordinates of the
principal point, distortion coefficients) and external (position and orientation)
parameters, without which at some level no arrangement of algorithms can
work Internal parameters are independent of outer conditions and do not
change from one snapshot to another snapshot.
In most situations, the internal parameters are calculated by shooting
calibration objects, the geometry of which is pre-known, but also it can
be computed without their use by applying a large number of images with
common points [9]. External calibration is performed by triangulating key
points common in two or more images.
338 Unmanned Aerial Vehicles (UAV) and Drones

To synthesize a 3D image, the problem of finding a match between key

points in different images has to be solved. This task is accomplished by
searching for points with minimum different descriptors-vector parameters
describing the point. Thereby the SIFT algorithm describes each key point
by a vector of 128 parameters identified on the basis of local gradients in the
vicinity of the point [11].
One of the most efficient algorithms for filtering false correspondences
of certain key points is RANSAC, Random Sample Consensus, [12]. The
essence of the mentioned method is iterative search for the best transformation
matrix (fundamental matrix) between N randomly selected key points in the
first image and the corresponding points in the second one. From the initials
set of key points RANSAC algorithm iteratively removes such points,
deviation of which as a result of the applying the found transformation from
the corresponding points in another image, exceeds the specified threshold.
Depth determination serves as the most challenging part in the whole
process, as it calculates the 3D component missing from any given image—
depth. The correspondence problem, finding matches between two images
so the position of the matched elements can then be triangulated in 3D space
is the key issue here. Once the multiple depth maps are determined they have
to be combined to create a final mesh by calculating depth and projecting out
of the camera. Camera calibration will be used to identify where the many
meshes created by depth maps can be combined together to develop a larger
one, providing more than one view for observation.
Complete 3D mesh may be the final goal, but applying the color from
the original photographs to the mesh is usually required. In this paper the
“mosaic” approach [13] was used to create the final texture which implies
two-step approach: it does blending of low frequency component for
overlapping images to avoid seamline problem (weighted average, weight
being dependent on a number of parameters including proximity of the pixel
in question to the center of the image), while high frequency component,
that is in charge of picture details, is taken from a single image—the one
that presents good resolution for the area of interest while the camera view
is almost along the normal to the reconstructed surface in that point.
As the initial data for the reconstructing a 3D model of the earth surface
with the objects of the railway infrastructure located on it, 176 drone photos
were used. A few examples of the original photos are shown in Figure 6.
The resulting 3D model of the surface under different camera angles and
from different heights is shown in Figure 7.
 Railway Transport Infrastructure Monitoring by UAVs and Satellites 339

A more complete picture of the obtained model is demonstrated in the video,

which could be downloaded from webpage [14]. Some limitations of the
data, which have been used for reconstruction of 3D image, should be noted:
• A relatively small number of images;
• Non-optimal flight path of the UAV;
• Small overlap between images;
• Absence of information about the orientation of the UAV and
camera, as well as data on GPS/GLONASS coordinates.

Figure 6. Examples of original images recorded by the UAV.

Figure 7. 3D reconstructed images of the railway depot.

Nevertheless, the resulting model sufficiently accurate reflects the features
of the area where the photography was made. The model was constructed
using the mathematical software Pix4D Mapper (Pix4Dmapper, https://
pix4d.com/product/pix4dmapper/).

CONCLUSION
The analysis of the possibilities of using remote sensing satellites and
UAVs equipped with various sensors showed that they could be an
340 Unmanned Aerial Vehicles (UAV) and Drones

effective monitoring tool for ensuring traffic safety on the railways. Their
information also can be used for solving other tasks, for example in the
design and construction of railways infrastructure. The paper shows that the
effectiveness of the Earth remote sensing can be increased by combining
information recorded by satellites and UAVs, and with the use of the modern
image processing means, it becomes possible to build synthetic three-
dimensional images of railway infrastructure.

ACKNOWLEDGEMENTS
The study was performed with financial support of the Russian Foundation
for Basic Research, Project No. 17-20-02086. The authors express their
gratitude to the staff of the Experimental Railway Ring for assistance in
conducting of UAV flights for obtaining the necessary information.

CONFLICTS OF INTEREST
The authors declare no conflicts of interest.
 Railway Transport Infrastructure Monitoring by UAVs and Satellites 341

REFERENCES
1. Flammini, F., Pragliola, C. and Smarra, G. (2016) Railway Infrastructure
Monitoring by Drones. International Conference on Electrical
Systems for Aircraft, Railway, Ship Propulsion and Road Vehicles &
International Transportation Electrification Conference, Toulouse, 2-4
November. https://doi.org/10.1109/ESARS-ITEC.2016.7841398
2. Eyre-Walker, R.E.A. and Earp, G.K. (2008) Application of Aerial
Photography to Obtain Ideal Data for Condition Based Risk
Management of Rail Networks. The 4th IET International Conference
on Railway Condition Monitoring, Derby, 18-20 June 2008. https://
doi.org/10.1049/ic:20080353
3. Francesco, F., Naddei, R., Pragliola, C. and Smarra, G. (2016) Towards
Automated Drone Surveillance in Railways: State-of-the-Art and
Future Directions. International Conference on Advanced Concepts
for Intelligent Vision Systems, Lecce, 24-27 October 2016, 336-348.
https://doi.org/10.1007/978-3-319-48680-2_30
4. Schwank, M. and Naderpour, R. (2018) Snow Density and Ground
Permittivity Retrieved from L-Band Radiometry: Melting Effects.
Remote Sensing, 10, 354. https://doi.org/10.3390/rs10020354
5. Skofronick-Jackson, G., Kim, M., Weinman, J.A. and Chang, D.-
E. (2004) A Physical Model to Determine Snowfall Over Land by
Microwave Radiometry. IEEE Transactions on Geoscience and Remote
Sensing, 42, 1047-1058. https://doi.org/10.1109/TGRS.2004.825585
6. Besada, J.A., Bergesio, L., Campaña, I., Vaquero-Melchor, D., López-
Araquistain, J., Bernardos, A.M. and Casar, J.R. (2018) Drone Mission
Definition and Implementation for Automated Infrastructure Inspection
Using Airborne Sensors. Sensors, 18, 1170. https://doi.org/10.3390/
s18041170
7. Koyama, C.N. and Sato, M. (2015) Detection and Classification of
Subsurface Objects by Polarimetric Radar Imaging. IEEE Radar
Conference, Johannesburg, 27-30 October 2015, 440-445. https://doi.
org/10.1109/RadarConf.2015.7411924
8. Bulgakov, A., Evgenov, A. and Weller, C. (2015) Automation of 3D
Building Model Generation Using Quadrotor. Procedia Engineering,
123, 101-109. https://doi.org/10.1016/j.proeng.2015.10.065
9. Hartley, R. and Zisserman, A. (2004) Multiple View Geometry
in Computer Vision. 2nd Edition, Cambridge University Press,
342 Unmanned Aerial Vehicles (UAV) and Drones

Cambridge. https://doi.org/10.1017/CBO9780511811685
10. McGlone, J.C. (2013) Manual of Photogrammetry. Sixth Edition,
American Society for Photogrammetry and Remote Sensing, Bethesda,
Maryland, USA, 1318 p.
11. Lowe, D.G. (1999) Object Recognition from Local Scale-Invariant
Features. Proceedings of the International Conference on Computer
Vision, Kerkyra, 20-27 September 1999, Vol. 2, 1150-1157. https://doi.
org/10.1109/ICCV.1999.790410
12. Fischler, M. and Bolles, R. (1981) Random Sample Consensus: A
Paradigm for Model Fitting with Applications to Image Analysis and
Automated Cartography. Communications of the ACM, 24, 381-395.
https://doi.org/10.1145/358669.358692
13. Maenpaa, T. and Pietikainen, M. (2005) Texture Analysis with Local
Binary Patterns. In: Handbook of Pattern Recognition and Computer
Vision, 3rd Edition, World Scientific, Singapore, 197-216. https://doi.
org/10.1142/9789812775320_0011
14. http://www.rslab.ru/pubdir/1.wmv
 CHAPTER

16
Assessment of UAV Based Vegetation
Indices for Nitrogen Concentration
Estimation in Spring Wheat

Olga S. Walsh1, Sanaz Shafian1, Juliet M. Marshall2, Chad Jack-

son2, Jordan R. McClintick-Chess1, Steven M. Blanscet3, Kristin
Swoboda4, Craig Thompson4, Kelli M. Belmont5, Willow L. Walsh6
Department of Plant Sciences, Southwest Research and Extension Center,
1

University of Idaho, Moscow, ID, USA.

Department of Entomology, Plant Pathology, and Nematology, Idaho Falls
2

Research and Extension Center, University of Idaho, Idaho Falls, ID, USA.
3
BASF-Chemical Co., Caldwell, ID, USA.
4
Take Flight UAS, LLC, Boise, ID, USA.
5
Seminis, Payette, ID, USA.
6
Vallivue High School, Caldwell, ID, USA.

Citation: S. Walsh, O., Shafian, S. , M. Marshall, J., Jackson, C., R. McClintick-Chess,

J., M. Blanscet, S., Swoboda, K. , Thompson, C., M. Belmont, K. and L. Walsh, W.
(2018) “Assessment of UAV Based Vegetation Indices for Nitrogen Concentration
Estimation in Spring Wheat”. Advances in Remote Sensing, 7, 71-90. doi: 10.4236/
ars.2018.72006.
Copyright: © 2018 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
344 Unmanned Aerial Vehicles (UAV) and Drones

ABSTRACT
Unmanned Aerial Vehicles (UAVs) have become increasingly popular in
recent years for agricultural research. High spatial and temporal resolution
images obtained with UAVs are ideal for many applications in agriculture.
The objective of this study was to evaluate the performance of vegetation
indices (VIs) derived from UAV images for quantification of plant nitrogen
(N) concentration of spring wheat, a major cereal crop worldwide. This study
was conducted at three locations in Idaho, United States. A quadcopter UAV
equipped with a red edge multispectral sensor was used to collect images
during the 2016 growing season.
Flight missions were successfully carried out at Feekes 5 and Feekes
10 growth stages of spring wheat. Plant samples were collected on the
same days as UAV image data acquisition and were transferred to lab for
N concentration analysis. Different VIs including Normalized Difference
Vegetative Index (NDVI), Red Edge Normalized Difference Vegetation
Index (NDVIred edge), Enhanced Vegetation Index 2 (EVI2), Red Edge Simple
Ratio (SRred edge), Green Chlorophyll Index (CIgreen), Red Edge Chlorophyll
Index (CIred edge), Medium Resolution Imaging Spectrometer (MERIS)
Terrestrial Chlorophyll Index (MTCI) and Red Edge Triangular Vegetation
Index (core only) (RTVIcore) were calculated for each flight event.
At Feekes 5 growth stage, red edge and green based VIs showed higher
correlation with plant N concentration compare to the red based VIs.
At Feekes 10 growth stage, all calculated VIs showed high correlation
with plant N concentration. Empirical relationships between VIs and plant
N concentration were cross validated using test data sets for each growth
stage. At Feekes 5, the plant N concentration estimated based on NDVIred
edge
showed one to one correlation with measured N concentration. At Feekes
10, the estimated and measured N concentration were highly correlated for
all empirical models, but the model based on CIgreen was the only model
that had a one to one correlation between estimated and measured plant N
concentration.
The observed high correlations between VIs derived from UAV and the
plant N concentration suggests the significance of VIs deriving from UAVs
for within-season N concentration monitoring of agricultural crops such as
spring wheat.
Keywords: Unmanned Aerial Vehicles and Systems (UAV), Vegetation
Indices (VIs), Plant Nitrogen Concentration
 Assessment of UAV Based Vegetation Indices for Nitrogen ... 345

INTRODUCTION
Nitrogen (N) is one of the essential factors for crop production in terms of
plant growth and development and crop quality [1] [2]. Adequate supply
of N is fundamental for optimizing wheat (Triticum aestivum L.) yield and
grain quality [3] [4]. Nitrogen regulates plant growth processes and plays a
vital role in chlorophyll (CL) production―the basis for the photosynthesis
process [5]. Insufficient N supply can negatively affect photosynthesis
process and result in crop yield and quality penalties [3]. On the other
hand, excessive N application to agricultural crops has been associated with
nitrate leaching, soil denitrification, ammonia volatilization, and nitrous
oxide contamination of aquifers and aggravating the climate change [6]
[7]. Dynamic and efficient fertilization (appropriate time and rate) is very
important for optimizing crop yield and maintaining environmental quality
[8]. Accurate estimation of crop N concentration is vital for developing
effective fertilizer-N recommendations.
There is a strong correlation between N concentration and CL content
at foliar and canopy scale because most of leaf N is localized within the
CL molecules [9] [10] [11] [12]. Chlorophyll content is the main elements
that govern the crop reflectance in the visible (VIS) and near infrared
(NIR) regions of spectrum [8]. Thus, vegetation reflectance in these parts
of spectrum is closely associated with N concentration. Remote sensing
enables to acquire crop reflectance and provide diagnostic information on
crop N concentration quickly and in a spatial context, compared to traditional
destructive sampling techniques [13]. During the last few decades, scientists
have proposed several vegetation indices (VIs) calculated from reflectance
data to assess CL content and N concentration [8] [13] [14] [15]. These
VIs are mostly a combination of NIR and VIS spectral bands, representing
radiation scattering by canopy and radiation absorption by CL respectively
[16]. Although these VIs accurately estimate CL and N concentration early
in the growing season at lower CL values, they become less sensitive as
the red spectral band is strongly absorbed by CL. Gitelson and Merzlyak
[17] showed that red edge region is sensitive to a wide range of CL content
values, and the use of this part of spectrum in VIs calculation can reduce
the saturation effect due to lower absorption of the red edge region by CL.
Several VIs based on this spectral region have been developed and used
successfully to estimate CL and N concentration.
Gitelson and Merzlyak [17] replaced the red spectral band (675 nm)
with red edge spectral band (705 nm) in Normalized Difference Vegetation
346 Unmanned Aerial Vehicles (UAV) and Drones

Index (NDVI) and developed a new index called Red Edge Normalized
Difference Vegetation Index (NDVIred edge). They showed that traditional
NDVI had a tendency to become saturated at higher CL level of senescing
maple and chestnut leaves while NDVIred edge continued to show strong linear
correlation with CL content and observed no saturation issue. In a similar
study, Gitelson et al. [18] showed that reciprocal of red edge spectral band
is closely related to the CL content in leaves of all species. They proposed
Red Edge Chlorophyll Index (CIred edge) and showed that CIred edge is highly
correlated with CL content (coefficient of determination R2 > 0.94). In another
study, Dash and Curran [19] developed MERIS Terrestrial Chlorophyll
Index (MTCI). They used data in three red/NIR wavebands centered at
681.25, 708.75 and 753.75 nm (bands 8, 9 and 10 in the MERIS standard
band setting) to develop MTCI. They determined the relationship between
MTCI and CL content using actual CL content for sites in the New Forest,
United Kingdom (UK) and for small plots in a greenhouse experiment.
Their results showed that MTCI is strongly and positively correlated to
actual CL content. Li et al. [5] evaluated red edge based spectral indices
for estimating plant N concentration and uptake of maize (Zea mays L.).
They calculated chlorophyll content index (CCCI), NDVIred edge, CIred edge and
MTCI from hyperspectral narrow bands, simulated Crop Circle ACS-470
active crop canopy sensor bands and simulated WorldView-2 satellite
broad bands. Their results showed that there is a positive strong correlation
between red edge based VIs and N concentration in maize. Their results
also indicated that CCCI performed the best across different bandwidths for
estimating maize plant N concentration at the V6 and V7 and V10 - V12
stages. In another study, Wang et al. [20] compared broad-band and red edge
based spectral VIs to estimate N concentration in corn, wheat, rice (Oryza
sativa L.) and soybeans (Glycine max L.). They calculated various VIs from
images acquired by the Compact Airborne Spectrographic Imager (CASI)
sensor. Their result showed that NDVI performed the best compared to other
VIs, and red edge based VIs did not show potential for accurate estimation
of leaf N concentration data due to spectral resolution.
Unmanned aerial vehicles (UAVs) are remote sensing systems that can
capture crop reflectance in the VIS-NIR region of spectrum and assess CL
and N concentration. The UAVs, which have recently gained tractions in
number of studies, acquire ultra-high spatial resolution images by flying at
low altitudes [21] [22]. Operational advantages such as low-cost systems,
high flexibility in terms of flight planning and acquisition scheduling, and
imaging below cloud cover make UAVs an appropriate tool to study crop
 Assessment of UAV Based Vegetation Indices for Nitrogen ... 347

biophysical parameters including N concentration [23]. In a few studies,

scientists have used reflectance data derived from UAVs to estimate CL or
N concentration. Lu et al. [24] mounted a Mini Multi-Camera Array (Mini-
MCA) imaging system on an octocopter UAV to estimate rice N status. They
examined various VIs to estimate N concentration at panicle initiation and
stem elongation growth stages in rice.
Their results showed that MTCI was best for estimating rice N
concentration with R2 = 0.48. In another study, Caturegli et al. [25]
compared the spectral reflectance of three turfgrasses (Cynodon dactylon ×
transvaalensis (Cdxt) “Patriot”, Zoysia matrella (Zm) “Zeon” and Paspalum
vaginatum (Pv) “Salam”) acquired with a UAV and using a ground based
instrument.
They also tested the sensitivity of the two data acquisition sources
in detecting induced variation among N application levels and for NDVI
calculation. Their results showed that NDVI obtained with the ground
based sensor was highly correlated with UAV based NDVI, with correlation
coefficient values ranging from 0.83 to 0.97.
They also showed that UAV based NDVI was strongly correlated with
N measured in the clipped plant biomass samples (correlation coefficient of
0.95). Similarly, Hunt et al. [26] used a UAV to monitor N status of irrigated
potato (Solanum tuberosum L.). They used a small parafoil-wing UAV to
acquire color-infrared images.
They showed that each of applied N treatments could be precisely
distinguished in the images. Their results also concluded that NDVI and
Green Normalized Difference Vegetation Index (GNDVI) were not useful
for in-season N management in potato because the above-ground changes in
leaf CL were not sufficiently large to be detected by remote sensing early in
the growing season.
So far, very few studies have investigated the potential of using red edge
based VIs from the UAV data for canopy CL or N concentration estimation.
To date, no studies on comparing red edge based VIs from the UAV data
for wheat canopy CL or N concentration estimation have been reported. The
main goal of this study was to evaluate the performance of UAV based VIs
in estimating plant N concentration at canopy scale.
Specifically, we analyzed and statistically compared the performance of
different red edge based VIs from UAV data to estimate spring wheat plant
N concentration.
348 Unmanned Aerial Vehicles (UAV) and Drones

MATERIALS AND METHODS

Study Area
The experimental studies were conducted at five different locations in Idaho
during 2016 growing season (Table 1 and Figure 1). The soil type, mean
annual temperature, and mean annual precipitation for each study site are
presented in Table 1.

Table 1. Latitude, longitude, soil type, mean annual precipitation, and mean
annual temperature for five locations in Idaho.).

Field Lat Lon Soil Type Mean annual

precipita- temper-
tion ature
Parma 43.80243 −116.94291 Green leaf-Owyhee silt loams 24 cm 10˚C
Ashton 44.07127 −111.39983 Greentimber-Marystown- 43 cm 5˚C
Robinlee silt loams
Rupert 42.72016 −113.50641 Sluka silt loam 28 cm 10˚C

Figure 1. Example of a figure caption (figure caption).

Hard red spring wheat (cv. Cabernet) was planted using a Hege 500 series
drill at Rupert and Ashton, and soft white spring wheat (cv. Alturas) was
planted at Parma using H&N Equipment research plot drill, at a density of
approximately 106.5 kg seeds ha−1. Row spacing was set at 17.78 cm using
double disk openers. The plots were 1.52 wide by 4.27 m long, and then
reduced to 3.05 m using glyphosate and tillage. Granular urea (46-0-0) was
surface broadcasted immediately after planting at five different rates (0, 84,
 Assessment of UAV Based Vegetation Indices for Nitrogen ... 349

168, 252, and 336 kg N ha−1). Each treatment was replicated four times in
a randomized complete block design, resulting in a total of 20 plots at each
location. Spring planting conditions were good for crop establishment. Soil
moisture March through April were above average, which resulted in excellent
early season growth and development. Early season precipitation provided
excellent growing conditions until irrigation became available in April when
all sites were irrigated every 7 to 10 days using sprinkle irrigation systems.
Timely planting dates resulted in excellent tillering and a long growing period.

Unmanned Aerial System

A quadcopter UAV 3DR Solo (3D Robotics, Inc., Berkeley, CA) shown
in Figure 2 was selected to carry camera payloads to acquire ultra-high-
resolution imagery. Solo is powered by four electric brushless 880 kV motors,
two that spin clockwise, and two counterclockwise. Solo’s arms are labeled 1
to 4 on the ends of the arms. Motors on arms #1 and #2 spin counterclockwise
and use clockwise-tightening propellers with silver tops. Conversely, motors on
arms #3 and #4 spin clockwise and use counter-clockwise tightening propellers
with black tops. Solo’s onboard computers control navigation, attitude, and
communications in flight while sending real-time telemetry and video output
and receiving control inputs over the 3DR Link secure WiFi network. A 14.8v
dc 5200 mAh lithium polymer (LiPo) rechargeable battery is located next to the
power button. The intelligent battery systems on the Solo tracks battery power
and informs the pilot when the battery needs to be recharged. Solo includes
a GoPro® the Frame (GoPro, Inc, San Mateo, CA) fixed mount to support a
GoPro® HERO camera (GoPro, Inc, San Mateo, CA). Alternatively, the fixed
mount could be replaced by the optional 3-Axis Gimbal (3D Robotics, Inc,
Berkeley, CA) (Figure 2). Empty weight of the quadcopter is 1.74 kg and weight
increased to 1.9 kg with compatible camera and Solo Gimbal.

Figure 2. The Unmanned Aerial System 3DR Solo.

350 Unmanned Aerial Vehicles (UAV) and Drones

A MicaSense Red Edge™ 3 Multispectral Camera (MicaSense, Inc, Seattle,

WA) with an integrated Global Positioning System (GPS) with an accuracy
of 2 - 3 meters, mounted on the UAV was used to obtain the imagery. The
camera was mounted on a Gimbal and as the camera’s weight was similar
to GoPro camera’s weight, there was no need to add balance weight. The
camera acquires 1.3-megapixel images in five spectral bands (red edge,
near infrared, red, green, and blue) with 12-bit Digital Negative (DNG) or
16-bit Tag Image File Format (TIFF) radiometric resolution. The ground
spatial resolution of resulting images from the narrowband imager is 8
cm (per band) at 120 m above ground level. The MicaSense Red Edge™
3 has a Downwelling Light Sensor (DLS) (MicaSense, Inc, Seattle, WA),
which measures the ambient light during a flight for each of the five bands
and records this information in the metadata of the images captured by the
camera. The gain and exposure settings are automatically optimized for each
capture and each band to prevent blurring or over-exposure, which results in
properly exposed images.

Multispectral Image Acquisition and Processing

The UAV images were captured within 2.0 hours of solar noon with flight
duration ranging from 15 to 20 minutes in sunny and cloud free conditions.
Mission planer software [27] was used to design the flight path and choose
the flight and sensor parameters to ensure there is an adequate overlap
between acquired images for mosaicking. Two flight missions performed at
each location to coincide with Feekes 5 and Feekes 10 spring wheat growth
stages resulted in six flight missions per season. These growth stages were
chosen because N fertilizer applied at these growth stages has potential to
maximize grain yield and quality.
Acquired multispectral images were imported to Micasense Atlas
software (MicaSense, Inc, Seattle, WA) for mosaicking, georeferencing
and radiometric calibration. Micasense Atlas has a partnership with Pix4D
mapper image analysis software (Pix4D SA, Lausanne, Switzerland) to
create aligned, mosaicked and georeferenced images from multispectral data
captured with the MicaSense red edge camera. The Pix4D finds hundreds
of tie-points between overlapped images and stiches the individual images
together to build one ortho-rectified image of the whole study area. The
accuracy of an outputted reconstructed images are usually 1-2 times the
ground spatial resolution. In this study, we had different treatments in the
adjacent rows (100 cm row spacing) (well separated plots), so a low level
of mosaicking error is allowable when using UAV images for our purpose
 Assessment of UAV Based Vegetation Indices for Nitrogen ... 351

[28]. Then, the mosaicked images were radiometrically calibrated using the
Red Edge Camera Radiometric Calibration Model in Atlas software. Atlas
software uses the calibration curve associated with a Calibrated Reflectance
Panel (CRP) to perform calibration model and convert the raw pixel values
of an image into absolute spectral radiance values. The CRP was placed
adjacent to the study area during each flight mission, and an image of the
CRP was captured immediately before and immediately after each flight.
The output of radiometric calibration model is a 5-layer, 16-bit ortho-
rectified GeoTIFF image.

Field Sampling and Measurements

To obtain a representative plant sample, aboveground biomass was
destructively sampled at Feekes 5 and Feekes 10 growth stages by cutting
three randomly selected plants in the middle of each plot immediately after
each UAV flight event. Plant samples were dried in the oven for 72 hours at
80˚c and then were transferred to the lab for N content analysis. Samples’
N content analysis was performed using the AOAC method 990.3 [29]
at Brookside Laboratories, Inc (New Bremen, OH, USA) with extended
uncertainty of ±5%.

Vegetation Indices
The UAV reflectance data were used for calculating eight VIs, many
of which have been proposed as surrogates for canopy N concentration
estimation. The VIs tested include the Normalized Difference Vegetation
Index, NDVI [30] , the Red Edge Normalized Difference Vegetation Index,
NDVIred edge [17] , the Enhanced Vegetation Index 2, EVI2 [31] , the Red Edge
Simple Ratio, SRred edge [32] , the Green and Red Edge Chlorophyll Indices,
CIgreen and CIred edge, respectively [18] , the MERIS Terrestrial Chlorophyll
Index, MTCI [19] , and the Core Red Edge Triangular Vegetation Index
(RTVIcore) [33] (Table 2). For each study plot, a region of interest (ROI) was
manually established by choosing the central two rows and mean of each VI
value corresponding to that plot was extracted.

Statistical Analysis
The study plots were randomly divided into test and training data sets. For
the training data sets, simple regression analysis was performed to find the
best relationship fit between N concentration and each UAV based VI. The
determination coefficient (R2) and Root Mean Squared Error (RMSE) were
352 Unmanned Aerial Vehicles (UAV) and Drones

used to evaluate the predictive accuracy of each model. These parameters are
widely used to evaluate the performance of empirical models. The RMSE
are computed as shown in Equations (1):

(1)
where, yˆi is predicted value of N concentration; yi is measured N
concentration, and M is total number of observations. In the next step, the
test data set was used to evaluate the performance of developed model
in the previous step. Predicted values of N concentration were plotted
versus corresponding values of N concentration measured in the lab. The
performance of regression models in estimating N for the training data set
were evaluated by calculating the R2 and

Table 2. Vegetation indices (VIs) tested in this study to estimate nitrogen (N)
content.

Vegetation Index Equation

Normalized Difference Vegetation Index (NIR − Red)/(NIR + Red)
(NDVI)
Red Edge Normalized Difference Vegetation (NIR − Red Edge)/(NIR + Red
Index (NDVIred edge) Edge)
Enhanced Vegetation Index 2 (EVI2) 2.5 × (NIR − Red)/(NIR + 2.4 ×
Red + 1)
Red Edge Simple Ratio (SRred edge) (NIR)/ (Red Edge)
Green Chlorophyll Index (CIgreen) (NIR/Green) − 1
Red Edge Chlorophyll Index (CIred edge) (NIR/Red Edge) − 1
Medium Resolution Imaging Spectrom- (NIR − Red Edge)/(Red Edge +
eter (MERIS) Terrestrial Chlorophyll Index Red)
(MTCI)
Core Red Edge Triangular Vegetation Index 100(NIR − Red Edge) − 10(NIR
(RTVIcore) − Green)
RMSE. In addition, Student’s t-tests were used to determine if the slope
and the intercept of the regressions were significantly different from 1 and
0, respectively. If the values of slopes were not significantly different from
1 and the values of intercepts were not significantly different from 0, then
it was concluded that the regression was not significantly different from the
1:1 line, and the empirical model could accurately predict N concentration.
 Assessment of UAV Based Vegetation Indices for Nitrogen ... 353

RESULTS AND DISCUSSIONS

Variation of Plant N Concentration

Based on different N application rates, a wide range of N concentration (%)
ranged from 0.76% to 1.58% were obtained at Feekes 5 (Table 3). In Parma
and Rupert sites, the highest N concentration (%) were obtained at fertilizer
N rate of 336 kg N ha−1 while in Ashton site, the highest N concentration
(%) was obtained at fertilizer N rate of 252 kg N ha−1. Similarly, at Feekes
10 a wide range of N concentration (%) ranged from 0.33% to 0.70%
were obtained. In all sites, the highest N concentration (%) were obtained
at fertilizer N rate of 336 kg N ha−1. Generally, observed variations in N
concentration were due to differences in climate, inherent soil fertility, and
N rate applications. When comparing N fertilizer rate applications, in Parma
and Rupert sites at both Feekes 5 and Feekes 10 growth stages, the rate of N
fertilizer significantly affected N concentration. In Ashton site at Feekes 5,
the rate of N fertilizer did not make significant changes in N concentration
while at Feekes 10, the rate of N fertilizer made significant changes in N
concentration.
Across growth stages, applied N rates and locations, plant N concentration
decreased at all locations (Table 3 and Table 4). These results indicate the
“dilution effect” as the crop matured, as described in previous studies [34]
[35]. Plant N concentration was lower for Ashton site at both Feekes 5 and
Feekes 10 growth stages, compared to other sites. As all study sites were
irrigated and had similar soil type (silt loam), the differences between plant
N concentration for the same N application rates may be associated with
differences in the amount of plant available water. Ashton site received 15
cm more precipitation compared to Rupert, and 17 cm more than Parma.
The similar result was obtained for the semiarid grassland by Lu [36]. In that
study, Lu showed that water additions significantly interacted to affect plant
N uptake and N concentrations at the community level.

Table 3. Plant N concentration at Feekes 5, as affected by the applied N fertil-

izer rate

N rate (kg N ha−1) Mean Plant N Concentration (%)

Parma Ashton Rupert
0 0.76d 0.76a 1.31.b
354 Unmanned Aerial Vehicles (UAV) and Drones

84 1.05c 0.80a 1.41b

168 1.18b 0.79a 1.44ab
252 1.17b 0.83a 1.39b
336 1.32a 0.82a 1.58a

Means within each column followed by the same letter are not
significantly different at p < 0.01, as determined by the Duncan’s multiple
range test.

Table 4. Plant N concentration at Feekes 10, as affected by the applied N fertil-

izer rate.

N rate (kg N ha−1) Mean Plant N Concentration (%)

Parma Ashton Rupert
0 0.33c 0.35b 0.33c
84 0.41bc 0.39ab 0.41bc
168 0.65a 0.45a 0.65a
252 0.49b 0.35b 0.49b
336 0.70a 0.45a 0.70a

Means within each column followed by the same letter are not
significantly different at p < 0.01, as determined by the Duncan’s multiple
range test.

Plant N Concentration Models Development and Validation

We used the training data set for establishing separate plant N content
predictive models using UAV based VIs for Feekes 5 and Feekes 10 separately.
The training data set includes a wide range of plant N content values due
to differences in N application rates. Eight models were developed using
the vegetation indices NDVI, NDVIred edge, EVI2, SRred edge, MTCI, CIgreen,
CIred edge and RTVIcore. Figure 3 and Figure 4 show the best relationship fit
between N concentration and each UAV based VI for Feekes 5 and Feekes
10, respectively. At Feekes 5 (Figure 3), the R2 of these models ranged from
0.69 to 0.88 and RMSE ranged from 0.096 to 0.16. At this growth stage, the
highest R2 between VIs and N concentrations was obtained for the CIgreen.
Also, the developed model based on this index showed the lowest RMSE
with N concentration. The best relationship fit between N concentration and
most VIs at Feekes 5 were quadratic (Figure 3). CIgreen and MTCI were the
 Assessment of UAV Based Vegetation Indices for Nitrogen ... 355

only VIs which their best relationship fit with N concentration were linear.
At Feekes 10 (Figure 4), the R2 of the developed models ranged from 0.82
to 0.88 and RMSE ranged from 0.06 to 0.08. At this growth stage, the largest
R2 between VIs and N concentrations again was obtained for CIgreen while
the developed model based on NDVIred edge had the lowest RMSE with N
concentration. The best relationship fit between N concentration and all VIs
(except EVI2) at Feekes 10 were quadratic (Figure 4). EVI2 was the only VIs
for which its’ best relationship fit with N concentration was linear.
All UAV based VIs used in this study showed strong positive correlations
with N concentration. In other words, all UAV based VIs used in this
study were good indicators of spring wheat plant N concentration for both
Feekes 5 and Feekes 10 growth stages. At Feekes 5 (Figure 3), red radiation
based UAV indices (NDVI and EVI2) had lower R2 and higher RMSE as
compared with other green and red edge based UAV indices. At this stage
(lower fractional vegetation cover), soil background influence on research
plots’ reflectance could be strong and could negatively affect the red based
VIs accuracy [37]. Red edge based VIs could minimize the soil reflectance
and isolate crop signal from soil reflectance as a function of canopy cover
changes [13]. This suggests that applying red edge based or green based VIs
from UAVs data can improve plant N concentration prediction compared
to the red based UAV VIs at Feekes 5 of wheat growth stage. At this stage,
CIgreen showed the highest R2 and lowest RMSE which suggests that the
green based VIs can be a better indicator of plant N concentration than the
red edge based VIs. This result is in consistent with the result of the previous
study conducted by Li [38] who showed that red edge based VIs were more
effective for N estimation at earlier growth stage of wheat. In addition,
CIgreen showed linear relationship with plant N concentration, which means
sensitivity of the model did not change due to the wide range of variation in
plant N concentration; it is straightforward to invert them between CIgreen and
plant N concentration to obtain a synoptic measure of N concentration.
All the red edge based VIs showed comparable performance with similar
R2 and RMSE values at Feekes 5. At Feekes 10 growth stage (Figure 4), all
UAV based VIs used in the study performed very well with similar R2 and
RMSE values. At this stage, the performance of NDVI and EVI2 for plant N
concentration improved. All other UAV based VIs had similar performance
compare to Feekes 5. Similar results have been reported by previous studies
as well [38]. At Feekes 10, the crop canopy had fully developed, and soil
background effect on research plots’ reflectance had been reduced, so red
based and red edge based VIs showed similar performance.
356 Unmanned Aerial Vehicles (UAV) and Drones

Figure 3. Relationships between measured plant N content (%) vs. (a) NDVI,
(b) NDVIred edge, (c) EVI2, (d) SRred edge, (e) MTCI, (f) CIgreen, (g) CIred edge and (h)
RTVIcore at Feekes 5 growth stage.
 Assessment of UAV Based Vegetation Indices for Nitrogen ... 357

Figure 4. Relationships between measured plant N content (%) vs. (a) NDVI,
(b) NDVIred edge, (c) EVI2, (d) SRred edge, (e) MTCI, (f) CIgreen, (g) CIred edge and
(h) RTVIcore at Feekes 10 growth stage.
The performance of the developed models in the previous step were
evaluated using test data set for Feekes 5 and Feekes 10, separately. For this
purpose, we used the developed models in previous step to estimate the N
concentration. Table 5 and Table 6 show the results of comparison between
358 Unmanned Aerial Vehicles (UAV) and Drones

measured N concentration and predicted N concentration retrieved from

developed model in previous step for Feeks 5 and Feekes 10, respectively.
At Feekes 5 (Table 5), the R2 of the relationship between measured and
predicted N concentration ranged from 0.67 to 0.84 and RMSE ranged from
0.147 to 0.196. At this growth stage, the lowest RMSE between measured and
predicted N concentrations were obtained using the developed models based
on RTVIcore and NDVIred edge (Table 5). At this growth stage, all the developed
models performed similar in terms of R2 (Table 5). The results of a Student’s t
test showed that for the developed models based on NDVIred edge, SRred edge and
CIred edge the slopes of those regression lines were not significantly different
from 1 (t = 0.0279; t = 0.43947; t = 0.43706 respectively) (Table 7), while
a similar test showed that the intercepts of those lines were not significantly
different from 0 (t = 1.026; t = 0.730; t = 0.731, respectively) (Table 7).
Thus, one could conclude that these regression lines were not significantly
different from the 1:1 line. These results indicated that the developed model
based on NDVIred edge can predict the plant N concentration at Feekes 5 with
best accuracy compared to other developed models (Figure 4(a)).

Table 5. The results of algorithm cross validation for estimating plant N con-
centration (N%) at Feekes 5 growth stage. Best fit functions, determination co-
efficients (R2) and root mean square errors (RMSE) of plant N concentration
estimation are given for eight vegetation indices.

VIs Estimated N% = a (mea- R2 RMSE (%)

sured N%) + b
NDVI 0.7265x + 0.3464 0.67 0.168
NDVIred edge 1.026x + 0.0815 0.84 0.148

EVI2 0.8167x + 0.2694 0.78 0.165

SRred edge 1.0419x + 0.0799 0.84 0.179
CIgreen 1.2495x − 0.1527 0.84 0.196
CIred edge 1.0417x + 0.08 0.84 0.179
MTCI 0.7932x + 0.3497 0.84 0.160
RTVIcore 0.8434x + 0.2811 0.84 0.147

Table 6. The results of algorithm cross validation for estimating plant N con-
centration (N%) at Feekes 10 growth stage. Best fit functions, determination
coefficients (R2) and root mean square errors (RMSE) of plant N concentration
estimation are given for eight vegetation indices.
 Assessment of UAV Based Vegetation Indices for Nitrogen ... 359

VIs Estimated N% = a R2 RMSE (%)

(measured N%) + b
NDVI 0.728x + 0.1509 0.72 0.10
NDVIred edge 0.769 + 0.143 0.74 0.091
EVI2 0.7778x + 0.107 0.73 0.095
SRred edge 0.756x + 0.147 0.75 0.091
CIgreen 0.891x + 0.03 0.76 0.090
CIred edge 0.752x + 0.141 0.75 0.091
MTCI 0.785x + 0.121 0.75 0.090
RTVIcore 0.778x + 0.163 0.72 0.104

Table 7. Regressions’ analysis parameter, t value, to determine if the slope and

the intercept of the regressions were significantly different from 1 and 0, respec-
tively at Feekes 5.

VIs NDVI NDVIred edge EVI2 SRred edge CIgreen CI- MTCI RTVIcore
red edge

Slop −2.53 0.0279 −2.2772 0.43947 2.1633 0.43 −1.9 −1.7958

856 706 776
Inter- 2.8066 1.026 2.7678 0.730 −1.1557 0.730 2.9279 2.8123
cept

All t values were determined at 46 df and α = 0.05.

Similarly, at Feekes 10, the R2 of the relationship between measured and
predicted N concentration ranged from 0.72 to 0.73 and RMSE ranged from
0.09 to 0.104. At this growth stage, the lowest RMSE between measured
and predicted N concentrations were obtained using the developed models
based on CIgreen and MTCI (Table 6). At this stage, all developed models
had weaker performance in term of R2, but all models had smaller RMSE
compared to Feekes 5. The results of a Student’s t test showed that only for
the model developed based on CIgreen, the slope of the regression line was not
significantly different from 1 (t = −1.15722, 46 df, α = 0.05) (Table 8), while
a similar test showed that the intercept of that line was not significantly
different from 0 (t = 0.6554, 46 df, α = 0.05) (Table 8). Thus, one could
conclude this regression line was not significantly different from the 1:1 line.
These results indicate that the model developed based on CIgreen can predict
the plant N concentration at Feekes 10 with greatest accuracy compared to
other models, and can be successfully used to estimate plant N concentration
of wheat crop in the future (Figure 4(b)).
360 Unmanned Aerial Vehicles (UAV) and Drones

In cross-validation process, results showed that at Feekes 5 developed

model based on NDVIred edge can predict the plant N concentration with
best accuracy compared to other developed models. The developed model
based on NDVIred edge slightly underestimated the plant N concentration at
lower values while overestimated the plant N concentration at higher values
(Figure 5(a)). One possible reason for this is the soil type that was slightly
different for each study site; it is possible that the total canopy reflectance
was affected by slightly different background reflectance values at each
location [39]. At Feekes 10, the developed model based on CIgreen predicted
the plant N concentration with highest accuracy compared to other models.
The developed model based on CIgreen at this growth stage did not have
over- or underestimation issue as the crop canopy had been fully developed,
which minimized the effect of different background reflectance soil values
(Figure 5(b)).
At different growth stages, UAV based VIs showed different behaviors
for estimating plant N concentration. Therefore, growth stage-specific
models would be preferred for estimating plant N concentration. Mid-season
plant N content estimation could significantly improve the opportunity
for farmers to intervene with strategic fertilizer management. Ideally, the
choice of an index for canopy N measurement should not depend on the
geographical location where measurements are made.
The VIs that minimize the soil reflectance, such as red edge based VIs,
have the best performance across different locations with different soil
type. These kinds of VIs could be used to map N content across farmer
fields without calibrations, allowing them to target N applications. The
results mean that adding red edge band to UAV sensors can improve plant N
concentration monitoring and estimation.
The main concern about developed models for plant N concentration
estimation is that these models are crop-specific. Each crop has its’ own
unique spectral signature; thus, the same models cannot be used for various
crops. Changes in plant characteristics (such as breeding improvements)
may require the development of additional algorithms.
Mixed results about the effect of wheat cultivars on spectral reflectance
have been reported in literature. Sembiring et al [40] found that wheat
varieties did not have significant effect on spectral measurements. On the
other hand, Sultana et al [41] documented that spectral reflectance (NDVI)
values have varied significantly for a wide range of cultivars treated with the
same N rates.
 Assessment of UAV Based Vegetation Indices for Nitrogen ... 361

Table 8. Regressions’ analysis parameter, t value, to determine if the slope and

the intercept of the regressions were significantly different from 1 and 0, respec-
tively at Feekes 5.

VIs NDVI ND- EVI2 SRred edge CIgreen CIred edge MTCI RTVIcore
VIred edge
Slop −2.9701 −2.9269 −2.7 −3.2624 −1.1572 −2.876 −2.6159 −2.4721
852
In- 2.8194 3.1016 2.2 3.3815 0.6554 2.809 2.5192 3.1225
ter- 996
cept

All t values were determined at 46 df and α = 0.05.

Figure 5. Cross validation of developed algorithms based on (a) NDVIred edge at

Feekes 5 and (b) CIgreen at Feekes 10 in estimating plant N concentration of
wheat crop in test data sets.
Remotely sensed VIs have been extensively used to quantify wheat
crop N status. The UAV technology appears to provide a good complement
to the current remote sensing platforms for N monitoring in wheat by
capturing low-cost, high resolution images. These UAV technologies can
bring a unique perspective to N management in wheat by providing valuable
information on wheat N status. Time, labor and money can be saved using
UAV data in crop monitoring.
Results presented in this paper show that high resolution images
acquired with UAVs are a useful data source for in-season wheat crop N
concentration estimation. At Feekes 5 growth stage, red edge and green
based VIs had higher correlation with plant N concentration compared to
red based VIs because red edge based VIs can reduce the soil background
effect on crop reflectance. At Feekes 10 growth stage, all calculated VIs
showed high correlation with plant N concentration, and there were no
significant differences between red and red edge based VIs’ performance.
At this stage, crop canopy has been fully developed, and soil reflectance did
362 Unmanned Aerial Vehicles (UAV) and Drones

not have strong effect on the reflectance of research plots. At Feekes 5, the
plant N concentration estimated based on NDVIred edge showed 1:1 correlation
with N concentration measured in the lab. At Feekes 10, the estimated and
measured N concentration were highly correlated for all developed models,
but the model based on CIgreen was the only model that had a 1:1 correlation
between estimated and measured plant N concentration. The observed high
correlation between UAV based VIs with plant N concentration indicates the
applicability of UAV for in-season data collection from agricultural fields.

CONCLUSIONS
Remotely sensed VIs have been extensively used to quantify wheat crop N status.
The UAV technology appears to provide a good complement to the current remote
sensing platforms for N monitoring in wheat by capturing low-cost, high resolution
images. These UAV technologies can bring a unique perspective to N management in
wheat by providing valuable information on wheat N status. Time, labor and money
can be saved using UAV data in crop monitoring.Results presented in this paper
show that high resolution images acquired with UAVs are a useful data source for in-
season wheat crop N concentration estimation. At Feekes 5 growth stage, red edge
and green based VIs had higher correlation with plant N concentration compared
to red based VIs because red edge based VIs can reduce the soil background effect
on crop reflectance. At Feekes 10 growth stage, all calculated VIs showed high
correlation with plant N concentration, and there were no significant differences
between red and red edge based VIs’ performance. At this stage, crop canopy
has been fully developed, and soil reflectance did not have strong effect on the
reflectance of research plots. At Feekes 5, the plant N concentration estimated based
on NDVIred edge showed 1:1 correlation with N concentration measured in the lab. At
Feekes 10, the estimated and measured N concentration were highly correlated for all
developed models, but the model based on CIgreen was the only model that had a 1:1
correlation between estimated and measured plant N concentration. The observed
high correlation between UAV based VIs with plant N concentration indicates the
applicability of UAV for in-season data collection from agricultural fields.

ACKNOWLEDGEMENTS
This study was supported in part by the Idaho Wheat Commission.

CONFLICTS OF INTEREST
The authors declare no conflicts of interest.
 Assessment of UAV Based Vegetation Indices for Nitrogen ... 363

REFERENCES
1. Biswas, D.K. and Ma, B.L. (2016) Effect of Nitrogen Rate and
Fertilizer Nitrogen Source on Physiology, Yield, Grain Quality, and
Nitrogen Use Efficiency in Corn. Canadian Journal of Plant Science,
96, 392-403. https://doi.org/10.1139/cjps-2015-0186
2. Jain, N., Ray, S.S., Singh, J.P. and Panigrahy, S. (2007) Use of
Hyperspectral Data to Assess the Effects of Different Nitrogen
Applications on a Potato Crop. Precision Agriculture, 8, 225-239.
https://doi.org/10.1007/s11119-007-9042-0
3. Walsh, O.S., Shafian, S. and Christiaens, R.J. (2018) Evaluation of
Sensor-Based Nitrogen Rates and Sources in Wheat. International
Journal of Agronomy, 2018, Article ID: 5670479. https://doi.
org/10.1155/2018/5670479
4. Ramoelo, A., Skidmore, A.K., Schlerf, M., Heitkönig, I.M., Mathieu,
R. and Cho, M.A. (2013) Savanna Grass Nitrogen to Phosphorous
Ratio Estimation Using Field Spectroscopy and the Potential for
Estimation with Imaging Spectroscopy. International Journal of
Applied Earth Observation and Geoinformation, 23, 334-343. https://
doi.org/10.1016/j.jag.2012.10.009
5. Li, F., Mistele, B., Hu, Y., Chen, X. and Schmidhalter, U. (2014)
Reflectance Estima- tion of Canopy Nitrogen Content in Winter Wheat
Using Optimised Hyperspectral Spectral Indices and Partial Least
Squares Regression. European Journal of Agronomy, 52, 198-209.
https://doi.org/10.1016/j.eja.2013.09.006
6. Reay, D.S., Davidson, E.A., Smith, K.A., Smith, P., Melillo, J.M.,
Dentener, F. and Crutzen, P.J. (2012) Global Agriculture and Nitrous
Oxide Emissions. Nature Climate Change, 2, 410. https://doi.
org/10.1038/nclimate1458
7. Ravishankara, A.R., Daniel, J.S. and Portmann, R.W. (2009) Nitrous
Oxide (N2O): The Dominant Ozone-Depleting Substance Emitted in
the 21st Century. Science, 326, 123-125.
8. Bao, Y., Xu, K., Min, J. and Xu, J. (2013) Estimating Wheat Shoot
Nitrogen Content at Vegetative Stage from in Situ Hyperspectral
Measurements. Crop Science, 53, 2063-2071. https://doi.org/10.2135/
cropsci2013.01.0012
9. Schlemmer, M., Gitelson, A., Schepers, J., Ferguson, R., Peng,
Y., Shanahan, J. and Rundquist, D. (2013) Remote Estimation of
364 Unmanned Aerial Vehicles (UAV) and Drones

Nitrogen and Chlorophyll Contents in Maize at Leaf and Canopy

Levels. International Journal of Applied Earth Observation and
Geoinformation, 25, 47-54. https://doi.org/10.1016/j.jag.2013.04.003
10. Baret, F., Houles, V. and Guérif, M. (2007) Quantification of Plant
Stress Using Remote Sensing Observations and Crop Models: The
Case of Nitrogen Management. Journal of Experimental Botany, 58,
869-880. https://doi.org/10.1093/jxb/erl231
11. Filella, I., Serrano, L., Serra, J. and Penuelas, J. (1995) Evaluating Wheat
Nitrogen Status with Canopy Reflectance Indices and Discriminant
Analysis. Crop Science, 35, 1400-1405. https://doi.org/10.2135/cropsc
i1995.0011183X003500050023x
12. Sage, R.F., Pearcy, R.W. and Seemann, J.R. (1987) The Nitrogen
Use Efficiency of C3 and C4 Plants: III. Leaf Nitrogen Effects on the
Activity of Carboxylating Enzymes in Chenopodium album (L.) and
Amaranthus retroflexus (L.). Plant Physiology, 85, 355-359. https://
doi.org/10.1104/pp.85.2.355
13. Cammarano, D., Fitzgerald, G., Basso, B., O’Leary, G., Chen, D.,
Grace, P. and Fiorentino, C. (2011) Use of the Canopy Chlorophyl
Content Index (CCCI) for Remote Estimation of Wheat Nitrogen
Content in Rainfed Environments. Agronomy Journal, 103, 1597-
1603. https://doi.org/10.2134/agronj2011.0124
14. Xuefeng, L., Qiang, L., Shaolan, H., Shilai, Y., Deyu, H., Zhitao, W.,
Lie, D., et al. (2016) Estimation of Carbon and Nitrogen Contents in
Citrus Canopy by Low-Altitude Remote Sensing. International Journal
of Agricultural and Biological Engineering, 9, 149.
15. Herrmann, I., Karnieli, A., Bonfil, D.J., Cohen, Y. and Alchanatis, V.
(2010) SWIR-Based Spectral Indices for Assessing Nitrogen Content
in Potato Fields. International Journal of Remote Sensing, 31, 5127-
5143. https://doi.org/10.1080/01431160903283892
16. Clevers, J.G. and Gitelson, A.A. (2013) Remote Estimation of Crop and
Grass Chlorophyll and Nitrogen Content Using Red-Edge Bands on
Sentinel-2 and-3. International Journal of Applied Earth Observation
and Geoinformation, 23, 344-351. https://doi.org/10.1016/j.
jag.2012.10.008
17. Gitelson, A. and Merzlyak, M.N. (1994) Spectral Reflectance Changes
Associated with Autumn Senescence of Aesculus hippocastanum L.
and Acer platanoides L. Leaves. Spectral Features and Relation to
Chlorophyll Estimation. Journal of Plant Physiology, 143, 286-292.
 Assessment of UAV Based Vegetation Indices for Nitrogen ... 365

https://doi.org/10.1016/S0176-1617(11)81633-0
18. Gitelson, A.A., Gritz, Y. and Merzlyak, M.N. (2003) Relationships
between Leaf Chlorophyll Content and Spectral Reflectance and
Algorithms for Non-Destructive Chlorophyll Assessment in Higher
Plant Leaves. Journal of Plant Physiology, 160, 271-282. https://doi.
org/10.1078/0176-1617-00887
19. Dash, J. and Curran, P.J. (2004) The MERIS Terrestrial Chlorophyll
Index.
20. Wang, Y., Liao, Q., Yang, G., Feng, H., Yang, X. and Yue, J. (2016)
Comparing Broad-Band and Red Edge-Based Spectral Vegetation
Indices to Estimate Nitrogen Concentration of Crops Using Casi
Data. International Archives of the Photogrammetry, Remote Sensing
and Spatial Information Sciences, 137-143. https://doi.org/10.5194/
isprsarchives-XLI-B7-137-2016
21. Maes, W.H., Huete, A.R. and Steppe, K. (2017) Optimizing the
Processing of UAV-Based Thermal Imagery. Remote Sensing, 9, 476.
https://doi.org/10.3390/rs9050476
22. Pádua, L., Vanko, J., Hruška, J., Adão, T., Sousa, J.J., Peres, E. and
Morais, R. (2017) UAS, Sensors, and Data Processing in Agroforestry:
A Review towards Practical Applications. International Journal of
Remote Sensing, 38, 2349-2391. https://doi.org/10.1080/01431161.2
017.1297548
23. Simelli, I. and Tsagaris, A. (2015) The Use of Unmanned Aerial
Systems (UAS) in Agriculture. HAICTA, Kavala, 17-20 September
2015, 730-736.
24. Lu, J., Miao, Y., Huang, Y., Shi, W., Hu, X., Wang, X. and Wan, J.
(2015) Evaluating an Unmanned Aerial Vehicle-Based Remote Sensing
System for Estimation of Rice Nitrogen Status. 4th International
Conference on Agro-Geoinformatics, Istanbul, 20-24 July 2015, 198-
203.
25. Caturegli, L., Corniglia, M., Gaetani, M., Grossi, N., Magni, S.,
Migliazzi, M., Raffaelli, M., et al. (2016) Unmanned Aerial Vehicle
to Estimate Nitrogen Status of Turfgrasses. PLoS ONE, 11, e0158268.
https://doi.org/10.1371/journal.pone.0158268
26. Hunt, E.R. and Rondon, S.I. (2017) Detection of Potato Beetle Damage
Using Remote Sensing from Small Unmanned Aircraft Systems.
Journal of Applied Remote Sensing, 11, Article ID: 026013. https://
366 Unmanned Aerial Vehicles (UAV) and Drones

doi.org/10.1117/1.JRS.11.026013
27. Mission Planner Home. http://ardupilot.org/planner
28. Shi, Y., Thomasson, J.A., Murray, S.C., Pugh, N.A., Rooney, W.L.,
Shafian, S., Rana, A., et al. (2016) Unmanned Aerial Vehicles for High-
Throughput Phenotyping and Agronomic Research. PLoS ONE, 11,
e0159781. https://doi.org/10.1371/journal.pone.0159781
29. Official Methods of Analysis of AOAC International. http://www.aoac.
org/aoac_prod_imis/aoac/publications/official_methods_of_analys
is/aoac_member/pubs/oma/aoac_official_methods_of_analysis.
aspx?hkey=5142c47 8-ab50-4856-8939-a7a491756f48
30. Rouse Jr., J., Haas, R.H., Schell, J.A. and Deering, D.W. (1974)
Monitoring Vegetation Systems in the Great Plains with ERTS.
31. Huete, A., Didan, K., Miura, T., Rodriguez, E.P., Gao, X. and Ferreira,
L.G. (2002) Overview of the Radiometric and Biophysical Performance
of the MODIS Vegetation Indices. Remote Sensing of Environment,
83, 195-213. https://doi.org/10.1016/S0034-4257(02)00096-2
32. Gitelson, A.A., Vina, A., Ciganda, V., Rundquist, D.C. and Arkebauer,
T.J. (2005) Remote Estimation of Canopy Chlorophyll Content
in Crops. Geophysical Research Letters, 32, L08403. https://doi.
org/10.1029/2005GL022688
33. Nicolas, T., Philippe, V. and Huang, W.J. (2010) New Index for
Crop Canopy Fresh Biomass Estimation. Spectroscopy and Spectral
Analysis, 30, 512-517.
34. Plénet, D. and Lemaire, G. (1999) Relationships between Dynamics
of Nitrogen Uptake and Dry Matter Accumulation in Maize Crops.
Determination of Critical N Concentration. Plant and Soil, 216, 65-82.
https://doi.org/10.1023/A:1004783431055
35. Muñoz-Huerta, R.F., Guevara-Gonzalez, R.G., Contreras-Medina,
L.M., TorresPacheco, I., Prado-Olivarez, J. and Ocampo-Velazquez,
R.V. (2013) A Review of Methods for Sensing the Nitrogen Status in
Plants: Advantages, Disadvantages and Recent Advances. Sensors, 13,
10823-10843. https://doi.org/10.3390/s130810823
36. Lü, X.T., Dijkstra, F.A., Kong, D.L., Wang, Z.W. and Han, X.G. (2014)
Plant Nitrogen Uptake Drives Responses of Productivity to Nitrogen
and Water Addition in a Grassland. Scientific Reports, 4, Article No.
4817. https://doi.org/10.1038/srep04817
37. Viña, A., Gitelson, A.A., Nguy-Robertson, A.L. and Peng, Y. (2011)
 Assessment of UAV Based Vegetation Indices for Nitrogen ... 367

Comparison of Different Vegetation Indices for the Remote Assessment

of Green Leaf Area Index of Crops. Remote Sensing of Environment,
115, 3468-3478. https://doi.org/10.1016/j.rse.2011.08.010
38. Li, F., Miao, Y., Hennig, S.D., Gnyp, M.L., Chen, X., Jia, L. and Bareth,
G. (2010) Evaluating Hyperspectral Vegetation Indices for Estimating
Nitrogen Concentration of Winter Wheat at Different Growth Stages.
Precision Agriculture, 11, 335-357. https://doi.org/10.1007/s11119-
010-9165-6
39. Todd, S.W. and Hoffer, R.M. (1998) Responses of Spectral Indices to
Variations in Vegetation Cover and Soil Background. Photogrammetric
Engineering and Remote Sensing, 64, 915-922.
40. Sembiring, H., Raun, W.R., Johnson, G.V., Stone, M.L., Solie, J.B. and
Phillips, S.B. (1998) Detection of Nitrogen and Phosphorus Nutrient
Status in Winter Wheat Using Spectral Radiance. Journal of Plant
Nutrition, 21, 1207-1233. https://doi.org/10.1080/01904169809365478
41. Sultana, S.R., Ali, A., Ahmad, A., Mubeen, M., Zia-Ul-Haq, M., Ahmad,
S., Jaafar, H.Z., et al. (2014) Normalized Difference Vegetation Index
as a Tool for Wheat Yield Estimation: A Case Study from Faisalabad,
Pakistan. The Scientific World Journal, 2014, Article ID: 725326.
https://doi.org/10.1155/2014/725326
 CHAPTER

17
Research and Teaching Applications of
Remote Sensing Integrated with GIS:
Examples from the Field

Laura C. Loyola1, Jason T. Knowles1,2, Andrew J. Marx1, Ryan

McAlinden1,3, Steven D. Fleming1
Spatial Sciences Institute, University of Southern California, Los Angeles,
1

CA, USA.
2
GeoAcuity, Los Angeles, CA, USA.
Institute for Creative Technologies, University of Southern California, Los
3

Angeles, CA, USA.

ABSTRACT
Remote sensing is used in the Spatial Sciences Institute (SSI) across the
full spectrum of the organization’s teaching and research initiatives. From

Citation: Loyola, L. , Knowles, J. , Marx, A. , McAlinden, R. and Fleming, S. (2019)

“Research and Teaching Applications of Remote Sensing Integrated with GIS: Ex-
amples from the Field”. Journal of Geographic Information System, 11, 670-682.
doi: 10.4236/jgis.2019.116041.
Copyright: © 2019 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
370 Unmanned Aerial Vehicles (UAV) and Drones

undergraduate to graduate classes that utilize unmanned aerial systems

(UAS) in data collection and graduate courses that incorporate remote
sensing for a variety of applications, including Earth observation, to applied
research via the Human Security and Geospatial Intelligence (HSGI) Lab
projects and work being done with One World Terrain (OWT) at the Institute
for Creative Technologies (ICT) to build a fully geo-referenced 3D planetary
model, SSI recognizes the need to educate students about remote sensing
techniques. This paper discusses faculty involvement in conducting a pilot
study for the Catalina Island Conservancy (CIC) using UAS to survey local
bison and deer populations. The team utilized an autonomous fixed-wing
UAS with a thermal payload to collect data for a semi-automated detection
workflow integrated within a GIS successfully identifying both deer and
bison. Additionally, graduate students participate in a weeklong experiential
learning opportunity on Catalina Island, CA during which they develop and
conduct a research project integrating UAS and other remotely sensed data
with primary data collection in a Geographic Information System (GIS). By
extension, the Institute then reinforces that these educational opportunities,
focused primarily on data acquisition, are instrumental in supporting the
geographic information systems, science, and technology experiences
in many diverse fields, including (but not limited to) human security,
humanitarian relief, sustainable urban and rural planning, and public health.
Keywords: Unmanned Aerial Systems (UAS), Integrated
Curriculum, Experiential Learning, Security, Sustainable Planning

INTRODUCTION
Within spatial sciences, the University of Southern California (USC)
Dornsife College of Letters, Arts and Sciences (Dornsife) Spatial Sciences
Institute (SSI) recognizes the importance of remotely sensed data as an
integral component of a Geographic Information System (GIS) and spatial
analysis to a variety of disciplines. As is evident from this special issue alone,
remotely sensed data integrated into a GIS can be applied to everything from
environmental analysis to human rights monitoring in far to reach locales.
We focus this paper on two research projects contributed by faculty and
affiliate faculty that incorporate remotely sensed data, acquired both from
satellite imagery and unmanned aerial systems, into GIS and virtual reality/
augmented reality (VR/AR) to promote the well being of military and the
monitoring of wildlife. These vastly differing projects underscore the variety
of applications for remotely sensed data and different research that the
 Research and Teaching Applications of Remote Sensing ... 371

Spatial Sciences Institute undertakes, which also can be used in teaching and
promoting the next generation of data acquisition and analysis specialists.
Reinforcing concepts and scientific theories is best accomplished through
active learning, when the creation of new knowledge occurs through the
transformation of experience [1] [2]. We discuss these projects not only
within the context of the integration of remote sensing with GIS, but also
the broader context of integrating remote sensing into curricular advances
within the SSI at both graduate and undergraduate levels. Situated in the
heart of Los Angeles, CA, the online graduate programs in Geographic
Information Sciences & Technology (GIST) at USC are currently the only
US programs that conform to UNIGIS standards of design and delivery
for distance learning in GIS and GISci. This places USC Dornsife SSI in a
unique position at the forefront of curricular development.
Below, we briefly describe some of the work that Director of Modeling,
Simulation, & Training at the Institute for Creative Technologies (ICT),
Ryan McAlinden oversees, and work that Professors Jason Knowles and
Andrew Marx have initiated with the Catalina Island Conservancy (CIC),
a non-profit organization that privately holds and manages over 88% of the
land on Catalina Island. The USC-ICT research utilizes novel geospatial
techniques and advances in the areas of collection, processing, storage and
distribution of geospatial data, while the pilot for wildlife monitoring of
local bison and deer populations via UAS-based high resolution visible
(RGB) and thermal imaging has not previously been conducted on Catalina
Island. We also describe how projects such as these have been incorporated
into specific spatial sciences courses over the years to provide support
for research and platforms for enhanced student learning. Aspects of the
pedagogical approaches were previously presented at the GIS-Pro & Cal-GIS
2018 conference [3], therefore, here we focus on newer research and faculty
capacity building. We also provide the workflow that we have undertaken to
increase the faculty capacity in operating UAS and working with remotely
sensed data. We acknowledge this may not work for all academic units, but
encourage academic departments to follow similar curricular advances.
The next section of this paper presents background, methodology, and
results of the wildlife monitoring pilot study on Catalina Island from 2018.
Section three describes the innovative research and application of remote
sensing integrated with GIS for AR/VR products and 3D planetary modeling
done at USC-ICT. Both research studies are presented within the context of
a Dornsife Spatial Sciences Institute graduate spatial data acquisition course
and the potential for building faculty capacity in integrating remote sensing
372 Unmanned Aerial Vehicles (UAV) and Drones

with GIS in this course. The fourth section focuses on novel integration
of remote sensing into the undergraduate and graduate curriculum within
spatial sciences and presents a potential workflow for other organizations to
build faculty capacity in this domain.

WILDLIFE MONITORING ON CATALINA ISLAND,

PILOT 2018

Background
The Dornsife Spatial Sciences Institute has long run an online spatial
data acquisition course that affords students the opportunity to experience
a weeklong field data excursion based at USC Wrigley Institute for
Environmental Studies (WIES) on Catalina Island. This course has evolved
from working solely with handheld GPS units to formally include unmanned
platforms. This evolution was bolstered when trained Remote License
Pilots (RPL) pilots, Professors Jason Knowles and Andrew Marx, joined the
faculty of SSI in late 2017. Additionally, spatial students and faculty have
intermittently previously worked with the Catalina Island Conservancy on
a variety of projects, mostly small-scale projects focused on areas near the
USC WIES campus or just beyond, during the week-long excursion, with
results remaining internal.
Having practical experience from previous work, both Knowles and
Marx were eager to continue their remote sensing work through SSI; the
already forged contacts at the CIC and the spatial data acquisition course
provided the opportunity to formally link their research with the student
curriculum and vice versa, to link the students with this practical application
of remote sensing. In April 2018, Knowles and Marx worked with the
CIC to develop and execute a pilot study for conducting wildlife surveys
of local bison and deer populations utilizing UAS-based high resolution
visible (RGB) and thermal imaging. The study looked at both the feasibility
of utilizing fixed wing UAS-based imagery to identify wildlife, but also at
workflow maximization; could gains be made in the efficiency and efficacy
of airborne counting and cataloging of wildlife in comparison to more
traditional on the ground field survey methods [4]. In addition, it was also
anticipated that this methodology would be less obtrusive and invasive to
the observed species due to the collection altitude of the UAS. This is a vital
component to wildlife monitoring and an especially high priority for ethical
organizations such as the CIC, which are the sole guardians of wildlife in
 Research and Teaching Applications of Remote Sensing ... 373

an area. The pilot study was done flying a fixed wing sense Fly Ebee Plus
with SODA (RGB) and Thermo MAP payloads (Figure 1) resulting in both
ultrahigh resolution aerial photography (at 1.12 in/2.84cm group sample
distance) and thermal data capture (at 9.49 in/24.11cm ground sample
distance). The CIC Director and two field biologists joined the collection
team for the duration of the project.

Methodology
The study area selected by the CIC was Middle Ranch Meadows on Catalina
Island (Figure 2), located near the center of the island in a valley surrounded
by rich topography making on the ground field visual observations difficult.
Two days of flying were completed on April 4 and 5, with both high-resolution
RGB and thermal cameras for a total of five flights. Both pilots (Knowles and
Marx) were holders of Civil Certificate of Waiver Authorization (COA), for
commercial operations via Federal Aviation Administration (FAA) Code of
Federal Regulations (14CFR) Part 107. The April 4 flights were for orientation
and equipment shakedown/calibration, while April 5 flights were for wildlife
data capture. On April 5, the first flight was completed with the thermal
package, taking off 30-minutes before sunrise (FAA earliest allowed). This
early flight was conducted to maximize the temperature differential between
the cold evening ground and the wildlife. Immediately after the first, one-
hour flight, a second flight was performed with a RGB sensor over the same
study area. This, along with ground truth performed by the CIC personnel
using binoculars from a vantage point to identify wildlife, was used to the
confirm the presence of wildlife and validate potential signatures identified in
the thermal imagery capture which is shown to be the most effective [5] [6].
Immediately following the flights, datasets were preprocessed in the field and
saved to secondary backup systems. Once back from the field, datasets were
processed overnight via commercial photogrammetry software (Pix4DMapper
v4.2) and the following datasets were produces within 48-hours of capture and
integrated within GIS (ArcMap v10.5):
• Ultra-high resolution RGB (visible) aerial photography (at 1.12
in/2.84cm ground sample distance);
• LAS Point cloud (Figure 3);
• Thermal surface model (at 9.49 in/24.11cm ground sample
distance);
• Digital surface model (DSM); and
• 3D Textured mesh (Figure 4).
374 Unmanned Aerial Vehicles (UAV) and Drones

After processing the thermal imagery, it was added to a GIS, for manual
analysis of the thermal imagery. Warm areas or literal “hotspots” from the
thermal imagery capture were identified and correlated to the visible imagery
and on the ground wildlife observations from the CIC field biologists.

Figure 1. SenseFly eBee Plus, field set up prior to data collection.

Figure 2. Study area: flight coverage at Middle Ranch, Catalina Island, CA and
Catalina Island with Califronia caostline.

Figure 3. LAS point cloud processing in photogrammetry software.

 Research and Teaching Applications of Remote Sensing ... 375

Figure 4. UAS collected products: Orthoimage integrated with a Streets base-

map (top left) and DSM integrated with Imagery basemap (bottom right) in
GIS.

Results
The preliminary results indicate that fixed wing UA appears to be a good
platform for wildlife surveys both for its ability to cover large areas and
host both RGB and thermal payloads. Initial surveys were a success with
both bison and mule deer identified in the thermal imagery captured by the
UAS survey (Figure 5 and Figure 6). Four mule deer and one bison were
ultimately counted over the 1.41 km2 study area.
The field crew and the RGB imagery flown immediately following the
thermal capture verified these signatures. The aerial collection methodology
was unquestionably more efficient in terms of being able to cover more
area (the eBee fixed wing has a flight time in excess of ~60-minutes and
depending on the collection altitude can cover vast areas in a single flight)
and visual observations of the wildlife during the flights indicated that there
was no disturbance, with the animals seemingly unaware of the UAS high
above them. In addition to the survey data, the derivative geospatial products
produced by the programming process (Figure 4 above) were found to be
extremely useful datasets for the CIC field biologists and GIS staff that
would normally not be available.
376 Unmanned Aerial Vehicles (UAV) and Drones

Summary
While we consider this initial pilot study a success, we believe that there
can further improvements to the workflow and collection methodology.
Collections and resultant thermal data capture would be significantly
improved by flying predawn (or at night) in order to get a larger wildlife
temperature differential signature between the environment and the wildlife.
Even at sunrise, the head distribution was already much greater on east-
facing slopes making detection of thermal signatures more difficult. Future
studies will see the submission of an FAA Waiver to allow for predawn
(or night) flying to maximize the temperature differential. Additionally,
manual analysis of the thermal imagery, while doable, is time consuming
and cumbersome. This process would benefit from the use of a scripted
automation or semi-automated routine for entity detection [7]. A detection
algorithm identify areas within the scenes where there are large temperature
differences would enable the user to more rapidly identify and catalog the
pertinent data from a large coverage area.

Figure 5. Thermal imagery from ThermoMap camera of bison (left) and RGB
imagery (right) with movement of bison from time 0615 to 0715.

Figure 6. Thermal imagery from ThermoMap camera of four mule deer (cir-
cled).
 Research and Teaching Applications of Remote Sensing ... 377

AUTONOMOUS TERRAIN MODELING

Prior to fully incorporating UAS data collection and integration into the
graduate level spatial data acquisition course, the Spatial Sciences Institute
collaborated with the Institute for Creative Technologies to provide
opportunities for UAS pilot training in difficult terrain and pilot workflow
presentations that mutually benefited ICT pilots and students on Catalina
Island. As such, we also describe the integration of aerial imagery performed
by USC-ICT.

Background
The USC-ICT’s Terrain efforts focus on researching and prototyping
capabilities that support a fully geo-reference 3D planetary model for use
in the Army’s next-generation training and simulation environments. USC-
ICT research exploits new techniques and advances in the focus areas of
collection, processing, storage and distribution of geospatial data to various
runtime applications.

Generalized Methodology
USC-ICT collects aerial images with Commercial off the Shelf (COTS)
UAS using the ICT autonomous UAV path planning and imagery collection
system. The software provides a user-friendly interface that encodes
photogrammetry best practices. Unlike other commercially available UAV
remote control software, the ICT solution was design for collecting aerial
images that cover a large area of interest with multiple flights. Parameters
that are required for data collections include a bounding box of the area
of interest, flight altitude, the desired overlap between images, and
camera orientation. An optimized flight path is then computed with these
parameters and the imaging task can be automatically accomplished. With
the acquired images, the 3D point clouds are reconstructed using commercial
photogrammetry software.
The photogrammetric-generated point clouds/meshes (Figure 7) from a
collection stage do not allow both user-level and system-level interaction, as
they do not contain the semantic information to distinguish between objects.
The workflow in previous works require either manually labeling the points
into different parts, or re-training a new model for each new scene. USC-
ICT has designed a fully automated process that utilizes deep learning to
automatically extract relevant features and segmentation from point clouds.
To train the feature attribution model, the points are first manually labeled
378 Unmanned Aerial Vehicles (UAV) and Drones

with the following labels: ground, man-made objects, vegetation, etc. These
point clouds are then adapted to 3D voxel grids to produce a representation
suitable for deep neural networks. ICT designed a simple yet effective 3D
encoding and decoding network architecture based on 3D U-Net for point
cloud segmentation.
During training, a straight and forward 3D data augmentation strategy
was designed to perform rotation, translation, and crop on the input data at
the same time.
This expands the amounts of data and allows for better generalization
capabilities of the model. The resulting pipeline is able to extract building,
ground, and vegetation in the raw point clouds automatically with high
accuracy and produce accurate 3D models (Figure 8).

Connection with Curriculum

With this technology and workflow in mind, Ryan McAlinden and pilot
trainees demonstrated the mission planning and other considerations to the
data acquisition classes. Pilot trainees were afforded the opportunity of a
non-traditional, low-stakes preparation, training, and flight time at WIES
on Catalina Island, while graduate students in the spatial acquisition course
experienced the mission planning and data collection, and were able to
utilize the aerial imagery collected and modeled in their project work. This
endeavor, while beneficial to USC-ICT and SSI, was taken over by the
general instructors in SSI, and is discussed below

Figure 7. Photogrammetric point cloud segmentation, University of Southern

California site.
 Research and Teaching Applications of Remote Sensing ... 379

Figure 8. 4-layer transition image of the USC campus.

INTEGRATING REMOTE SENSING AND UAS INTO

CURRICULUM
As an academic unit, the SSI continues to grow and develop curriculum at
the graduate and undergraduate levels. This is evident in our changing course
offerings such as Spatial Data Collection Using Drones, an introductory
undergraduate course that provides students with technical and practical
experience with UAS, and program development such our Graduate
Certificate in Remote Sensing for Earth Observation (RSEO). These courses
and programs have undergone curricular review by our respective internal
curriculum committees, and the appropriate Curriculum Offices at the
College and University levels.

Student Programs and Curriculum Development

The Spatial Sciences Institute initiated curricular updates concurrently
for both undergraduate education and graduate education that incorporate
remote sensing at a higher level than was previously encompassed. Some
changes were considered minor, modifications to content of existing courses
that meet learning objectives with regards to remote sensing and UAS data
collection, while other changes involved the creation of new courses and
programs. As discussed for this paper, students enrolled in the graduate level
Spatial Data Acquisition course who are also new to integrating remotely
sensed data, collected via instructor conducted UAS flights, gain experience
in pre-flight mission planning and post-flight processing of the data. They
also trouble shoot problems that may arise with integration of these data
380 Unmanned Aerial Vehicles (UAV) and Drones

with other field collected data; we encourage students to work through the
processes to solve issues such as mis-matched coordinate systems upon
projecting both UAS imagery (collected in UTM Zone 11N) with data
from high accuracy receivers (collected in WGS 1984 and projected in Web
Mercator) under guidance from faculty. The students can then related this
practical experience with real world situations in which they will be utilizing
these data and processes.
Additional curricular development culminated in the creation of a non-
major undergraduate course in spatial data collection utilizing drone, in which
students will, for the first time, work the UAS to plan, collect, and process
imagery. This course revolves around applied and active learning experiences
in which students can develop and demonstrate a deeper knowledge and
understanding of the technological sciences behind the UAS-based collections,
processing, and visualizations, and through germane examples. Additionally,
this is a one-off course, meaning students across all disciplines are welcome
to enroll in the course, no prior experience with GIS is required, and this is
not limited to majors or minors. We anticipate that this course likely will draw
students from the most diverse majors and academic disciplines. Lastly, we
have developed a new graduate certificate program in Remotes Sensing for
Earth Observations (RSEO), which leverages remotely sensed data from a
multitude of sources such as Location Based Services (LBS), social media,
and Internet-of-Things (IoT) devices for a variety of applications from weather
and environmental observations to disaster management and recover efforts.
The program focuses on the acquisition, management, and integration of these
data for the purpose of advanced trend analysis, with the aim that students and
professionals develop proficiency working with these data and are able apply
their use in decision-making processes.

Capacity Building of Faculty for Improved Student Outcomes

In order to create a sustainable program that incorporated UAS and remote
sensing technology at a greater level, SSI invested in building capacity of
current faculty (Figure 9) beyond the abilities of Knowles and Marx. This
entailed developing technical training and pilot certifications for the selected
faculty that are, or would be, responsible for spatial data acquisition via
UAS. Trainee faculty members worked in collaboration with USC-ICT and
experienced pilot faculty to develop study plans and to review materials for
the FAA RPL Part 107 exam. Faculty then created individual learning plans
and study schedules. Due to fiscal considerations, faculty were required to
sit for the Part 107 exam prior to the closing of fiscal year 2017.
 Research and Teaching Applications of Remote Sensing ... 381

In consultation with USC-ICT, Knowles, and Marx, SSI also invested in

the necessary equipment and photogrammetry software for educational
purposes. Equipment this included a quad copter UAS (DJI Phantom 4 Pro)
with a RGB payload, no thermal or multispectral payloads are mounted
currently. This was a budgetary consideration and we recommend that initial
investment also include a multispectral payload. This will greatly increase
the data collection possibilities to include vegetation and landscape analysis,
such as Normalized Difference Vegetation Index (NDVI) or others, which
can be applied to a variety of domains ranging from crop management in
agricultural to tracking people and objects for human security, search and
rescue, and other military operations. Additional equipment and accessories
included extra Intelligent Flight batteries, back-up propellers, carrying
case, and one tablet. Educational licensing for photogrammetry software
(Pix4D) was purchased and is renewed annually. Additional mission
planning software (AirMap, DJI-Go, etc) are available free of charge and
downloadable to any mobile device.

Figure 9. Workflow for capacity building and training of faculty for attaining
FAA RPL, Part 107.
Upon successful completion of the exam, all faculty pilots engaged in
physical training with Marx and Knowles. A pre-flight check list, in-flight
protocol, and post-flight image processing workflow were standardized for use
with the current graduate spatial data acquisition course and are available to
use for additional courses, such as the aforementioned undergraduate course
382 Unmanned Aerial Vehicles (UAV) and Drones

that specializes in spatial data collection using drones. Test flights, under
the direction of Marx and Knowles, were conducted in open, unpopulated
parks, in accordance with FAA regulations and recommendations that limit
flights overhead of people in public spaces. Faculty also conducted test
data collections, processing, and integration of outputs (DSM, 3D mesh)
incorporated into a GIS platform to visualize georeferenced data to hone
integration abilities.

DISCUSSION
Ideally, students will develop projects that test not only the utility of
remotely sensed imagery collected via UAS, but also efficiency and efficacy
of their workflows. The weeklong Catalina experience can be used as a
testing ground for implementation of this technology to new domains on
a small scale, prior to large-scale implementation. Our goal is to provide
students the opportunities to engage in active experimentation and improve
learning experience and outcomes. This is accomplished through active
collection of data, detailed data analysis, and product production via some
geo-visualization outcome.
Having experienced faculty that can train additional faculty, and act as a
resource throughout development of the programs and/or new student projects
and novel flight paths, is key to the success of our programs. Successful pilot
studies that apply remote sensing to wildlife tracking and monitoring, such
as the first study highlighted above, and 3D models of the natural and built
environment that are derived from UAS collected imagery and an automated
computing process, such as accomplished by USC-ICT, are exemplars of the
advantages of integrating remote sensing in a GIS. Additionally, faculty that
can effectively communicate and demonstrate the possibilities of remotely
sensed data acquired via UAS, are vital to improved student experiences and
outcomes. SSI does not aim to train pilots, and in fact some of the students
may already have experience working with UAS and product visualization
through current jobs. Rather, courses and programs focus on the utility of
remotely sensed data within a GIS, the science of photogrammetry and
production of geo-referenced 3D models, and the variety of geospatial
analyses that can be run for the diverse applications of remotely sensed data.
Lastly, while we have focused the case studies presented here on remotely
sensed data mainly collected via unmanned aerial systems, our students
interact with a variety of remotely sensed data (LiDAR, multi- and hyper-
spectral satellite imagery, etc) within a GIS during these courses and in the
 Research and Teaching Applications of Remote Sensing ... 383

progress of research projects that span the humanities and physical sciences.
Through this work, SSI reinforces that these educational opportunities,
focused primarily on data acquisition, are instrumental in supporting the
application of geographic information systems, science, and technology in
many diverse fields ranging from human security and humanitarian relief, to
sustainable urban and rural planning and public health.

CONCLUSION
We have presented two innovative research projects that integrate remotely
sensed data collected via UAS in GIS for distinct purposes, but that share
the common element of further integration of remote sensing into the
curriculum of spatial sciences courses. We successfully demonstrated the
potential for wildlife monitoring on Catalina Island utilizing UAS and
made tangible recommendations the future work in this domain. We also
presented the work of USC-ICT and development of an automated process
to build a fully geo-referenced 3D model of the earth for the training and
simulation needs as the impetus and model for faculty development at SSI.
In order to achieve the curricular renovations referenced above, faculty must
be properly equipped to guide student data acquisition, processing, and
analysis; we presented the workflow of how the Spatial Sciences Institute
achieved this and the importance for our student development.

ACKNOWLEDGEMENTS
We would like to thank the Catalina Island Conservancy for engaging with
the pilot study described here and permitting the public dissemination of the
results. We also thank the staff of the Wrigley Institute for Environmental
Studies on Catalina Island for their continue support of the field course.

CONFLICTS OF INTEREST
The authors declare no conflicts of interest regarding the publication of this
paper.
384 Unmanned Aerial Vehicles (UAV) and Drones

REFERENCES
1. Kolb, D.A. (2015) Experiential Learning: Experience as the Source of
Learning and Development. 2nd Edition, Pearson Education, Inc.
2. McLeod, S.A. (2017) Kolb-Learning Styles and Experiential Learning
Cycle. Simple Psychology. https://www.simplypsychology.org/
learning-kolb.html
3. Loyola, L.C., Marx, A.J. and Fleming, S.D. (2018) Combining
Teaching, Partnerships, and Research in the Field: Lessons from the
Spatial Data Acquisition Course (on Catalina Island). GIS-Pro &
CalGIS 2018, Palm Springs, 8-12 October 2018.
4. Kays, R., Sheppard, J., Mclean, K., Welch, C., Paunescu, C., Wang,
V., Crofoot, M., et al. (2019) Hot Monkey, Cold Reality: Surveying
Rainforest Canopy Mammals Using Drone-Mounted Thermal Infrared
Sensors. International Journal of Remote Sensing, 40, 407-419. https://
doi.org/10.1080/01431161.2018.1523580
5. Hodgson, J.C., Mott, R., Baylis, S.M., Pham, T.T., Wotherspoon, S.,
Kilpatrick, A.D., Koh, L.P., et al. (2018) Drones Count Wildlife More
Accurately and Precisely than Humans. Methods in Ecology and
Evolution, 9, 1160-1167. https://doi.org/10.1111/2041-210X.12974
6. Chrétien, L.P., Théau, J. and Ménard, P. (2016) Visible and Thermal
Infrared Remote Sensing for the Detection of White-Tailed Deer Using
an Unmanned Aerial System. Wildlife Society Bulletin, 40, 181-191.
https://doi.org/10.1002/wsb.629
7. Lhoest, S., Linchant, J., Quevauvillers, S., Vermeulen, C. and Lejeune,
P. (2015) How Many Hippos (HOMHIP): Algorithm for Automatic
Counts of Animals with Infra-Red Thermal Imagery from UAV.
International Archives of the Photogrammetry, Remote Sensing and
Spatial Information Sciences, 40, 355-362. https://doi.org/10.5194/
isprsarchives-XL-3-W3-355-2015
 CHAPTER

18
Development of Drone Cargo Bay with
Real-Time Temperature Control

Sedam Lee, Yongjin Kwon

Department of Industrial Engineering, College of Engineering, Ajou
University, Suwon, South Korea

ABSTRACT
In order to deliver medical products (medicines, vaccines, blood packs, etc.)
in time for needed areas, a method of transporting goods using drones is
being studied. However, temperature-sensitive medical products may decay
due to outside temperature changes. The time required to transport over the
distance may vary a lot as well. As a result, the likelihood of the goods
deteriorating is very high. There is a need for a study on cargo bay to prevent

Citation: Lee, S. and Kwon, Y. (2019), “Development of Drone Cargo Bay with Real-
Time Temperature Control”. World Journal of Engineering and Technology, 7, 612-
621. doi: 10.4236/wjet.2019.74044.
Copyright: © 2019 by authors and Scientific Research Publishing Inc. This work is li-
censed under the Creative Commons Attribution International License (CC BY). http://
creativecommons.org/licenses/by/4.0
386 Unmanned Aerial Vehicles (UAV) and Drones

this and to protect the medical goods. In this paper, in order to protect the
temperature sensitive medical goods, the inside cargo bay is equipped with
the cooling fan device and the electric heating elements. These elements
can be monitored and controlled according to the user’s discretion. By
using the web server built inside the cloud server, the temperature can be
controlled in real-time from anywhere without the limitation of distance.
We built the proposed device, and installed it on the drone cargo bay. The
test results show that the cargo bay can be temperature-controlled, and the
setting can be maintained over a great distance. The user can watch the
temperature variations during the transport and ascertain the goodness of
the medical supply with the data. It is expected that such development can
greatly enhance the utility of the drone operations, especially for the medical
supply transport applications.
Keywords: Real-Time Control, Cargo Bay Temperature, Cloud Server,
Medial Transport, Drones

INTRODUCTION
In many countries around the world, a fair accessibility to roads varies
according to the geographic characteristics and weather conditions. As a
result, it is difficult to transport items, such as medical supplies (medicines,
vaccines, blood packs, etc.) when needed [1]. Currently, there is a variety of
methods for transporting medical products by walking, transporting vehicles,
helicopters, and airplanes. However, these methods have limitations in terms
of distance, road accessibility, delivery costs, and geographic conditions
[1]. To overcome these limitations, methods for transporting medical
products using drones are being studied. The use of drones among various
methods has already been extended to many industries [2] [3]. In addition,
the transport of goods using drones has been steadily studied in recent
years [4]. According to the papers that studied the effects of transporting
blood by drones, the use of unmanned aircraft does not affect the quality
of the products, such as coagulation and deterioration of medicines and
biochemical samples [5]. However, the factor mostly affecting the goods in
transport is the external environment. Depending on various factors such as
region, season and weather, medical and various items stored in drone cargo
bay may deteriorate. Therefore, there is a need for a method for protecting
the medical supplies from outside temperature changes.
Currently, a common method for lowering the temperature inside the
drone cargo bay is by arranging dry ice, and ice packs. Those include the
 Development of Drone Cargo Bay with Real-Time Temperature Control 387

Rwanda’s blood transport drone operations and a few other countries. This
method does not have a means to raise or control the internal temperature
to the desired setting. The existing method is simple, but due to various
external requirements, it is difficult to avoid deterioration of stored items.
This is because the type, place, time, and storage temperature of the stored
items are all different. For this reason, there is a need for a method capable of
controlling the cargo bay temperature. To solve this problem, we developed
a technology to control the temperature inside the drone cargo bay in real-
time. The structure of this paper shows Chapter 2 as related works, the
configuration of temperature control system in Chapter 3, and hardware
development in Chapter 4. Chapter 5 describes software development, and
Chapter 6 describes the prototyping results. Finally, the last chapter includes
concluding remarks.

RELATED WORKS
In recent years, various approaches have been proposed for the delivery of
medical items (blood, medical supplies, defibrillators, life-rings, etc.) using
the drones [5]. One of these methods is to develop an efficient delivery
system to overcome the hurdles from mountains and buildings during the
flight, or to operate and manage drones using a modular design approach
to effectively accommodate the various cargo types [6] [7]. However, the
methods proposed or introduced in the papers have not considered the
goodness of the transport items during the flight. In addition, the paper
for improving the goodness of transport items is mainly dealt with cargo
bay insulation to maintain the internal temperature. In most cases, the
method focuses on maintaining a low temperature. It shows no result about
transporting items that need to be stored at high temperature. It is difficult
to find the related study that shows the flexible temperature control of the
drone cargo bay [8] [9]. In this regard, the method to control the temperature
of cargo bay by raising or lowering the temperature remotely can be viewed
as quite innovative [10] [11].

CONFIGURATION OF TEMPERATURE CONTROL

SYSTEM
In order to operate cargo bay module with real-time temperature control,
the system is configured as in Figure 1. System configuration is divided into
hardware and software sections. Hardware is composed of the cargo bay
388 Unmanned Aerial Vehicles (UAV) and Drones

control module that exchanges data with the database, mainly on Raspberry
Pi. This controller controls the attached devices and sensors. Software is
largely divided into cloud server and web browser. Client and database that
can send and receive data to and from the cloud server are configured. The
data can be transmitted to the UI (user interface) of web browser through
terminal. The real-time temperature control method proposed in this study
is the method of building various servers for network connectivity through
cloud sever, and controlling the cargo bay module using the deployed server,
which can check and command the data of the cargo bay module using the
UI.

HARDWARE DESIGN
The temperature-controlled drone cargo bay proposed in this study is
730-mm wide, 250-mm long and 200-mm high, as shown in Figure 2.
This is designed by considering the payload restrictions of the drone. In
order to effectively lower the internal temperature during the flight, the air
circulation port was fabricated. The cooling fan plays the role of taking the
air to the cargo bay, which can lower the internal temperature. Thermic ray,
which generates heat through the electricity supply, was used, and a barrier
membrane was designed to maintain the cargo bay temperature efficiently.
The barrier membrane also divided the cargo bay interior space into an
interior compartment for storing goods and an exterior compartment for
installing temperature control devices. Table 1 and Table 2 show the details.

Figure 1. Temperature control module system configuration.

 Development of Drone Cargo Bay with Real-Time Temperature Control 389

Figure 2. Temperature controlled cargo bay schematic.

Table 1. Controller specification.

390 Unmanned Aerial Vehicles (UAV) and Drones

Table 2. Temperature sensor, thermic rays, cooling fan specifications.

SOFTWARE DESIGN

Network Connection
It is implemented that a communication using RF sensor as a way to check
and control the temperature inside cargo bay in real time [12]. However,
RF communications have adopted a network connection method to address
distance constraints. A network connection using a specific IP address by
building a web server in the controller’s memory is not possible due to the
nature of the drone when the network is disconnected due to geographic
distances. Therefore, we designed a static IP that plays an intermediate
bridge role by building a web server through a cloud server. The web server
was installed using “Apache, PHP, MySQL” on the cloud server, and data
were running with the use of interlocks to receive data from the controller
and send commands. Table 3 shows the data interlocking list for network
connections. All six data linkage lists are implemented in PHP language
through Terminal. It is designed to display and command data on the UI
through interlocked data.
The control UI is designed to prevent users from receiving unnecessary
information through the controller’s kernel window or complicating the
controller’s command statements. It is designed to receive the necessary
information quickly or issue commands. The first box of each configuration
in Figure 3 shows the current temperature, as well as the upper and lower
limits of the temperature. The second is the upper and lower limit input
fields of temperature. The third is the operating status of thermic ray and
cooling fan, and the on/off control switch of the controller power. The fourth
box is a graph showing the cargo bay’s internal temperature change, updated
continuously over time to show the changes in temperature.
 Development of Drone Cargo Bay with Real-Time Temperature Control 391

Control Algorithm Design

The goods stored in cargo bay for the transport have different storage
temperatures. Since the different storage temperatures must be considered
for each item, the cargo bay internal temperature must be freely set. When
a certain value is set for internal temperature control and there is no margin
value of the specified temperature, the temperature control device adjusts
frequently even with a minor temperature change of only 0.1˚C. Therefore,
the temperature should be set by specifying the range. In this study, the
method of setting and controlling the temperature using upper and lower
limits was used.

Figure 3. Temperature control user interface.

Table 3. Network data interlock list.

392 Unmanned Aerial Vehicles (UAV) and Drones

The temperature control algorithm set the basic operation of the thermal
ray and cooling fan to zero as shown in Figure 4. After that, the upper and
lower limit values are entered and the algorithm is repeated or terminated
according to each condition. For example, in order to store an article having
a storage temperature of 25˚C to 26˚C, the upper limit value is set to 26˚C
and the lower limit is set to 25˚C. After inputting the upper and lower limit
and the internal temperature is higher than the upper limit, the cooling fan
becomes operated. If the internal temperature is lower than the lower limit,
the Thermic ray becomes activated. If the internal temperature is between
the upper and lower limits, the current temperature is maintained.

Figure 4. Temperature control algorithm pseudo code.

PROTOTYPE DESIGN
Figure 5 shows the VTOL (vertical take-off and landing) drone equipped
with a prototype temperature controlled cargo bay. Since the payload is about
2 kg, the weight of the medical product as a transport item is considered.
The cargo bay prototype was built with a weight of around 1.5 kg. To check
whether the temperature inside the cargo bay is controlled and maintained,
two temperature sensors were used to measure the current temperature. The
temperature inside the cargo bay was checked for 15 minutes in every 10
seconds. As shown in Figure 5, the upper and lower limits of the cargo bay
 Development of Drone Cargo Bay with Real-Time Temperature Control 393

temperature were adjusted to 25˚C and 24˚C, respectively, and the cargo bay
temperature was measured to be maintained at the desired setting.
It took about three minutes for the cargo bay internal temperature to
reach the set point at 25.4˚C. After that, when the internal temperature is out
of the set temperature, the cooling fan operates to maintain 25˚C. In order to
check whether the temperature is maintained by thermic ray, it is verified by
setting the cargo bay internal temperature to 26˚C - 27˚C, as shown in Figure
6. Compared to the cooling fan, the set temperature was reached in about 40
seconds. If the temperature goes out of the set temperature, the thermic ray
is operated to increase the temperature to maintain the internal temperature,
and there is an error of about 0.1˚C.
The controller can effectively recognizes and controls the setting
measured by the temperature sensors. However, an error occurs between the
set temperature and the holding temperature.
The error in temperature is due to the delay in the data and must react
immediately when the temperature changes. In fact, the precise measurement,
control, and data processing require the use of high-performance computers,
sensors, and devices. In this study, however, due to the limitations of weight
and payload of the drone, the use of high performance precision device was
not feasible (Table 4).

Figure 5. Temperature module prototype.

394 Unmanned Aerial Vehicles (UAV) and Drones

Figure 6. External temperature vs. cargo bay internal temperature. (a) Graph of
internal temperature and external temperature; (b) Graph of internal tempera-
ture and external temperature.

Table 4. Erro value with a 40 second interval.

The temperature measured for a total of 15 minutes, and the interval

was set as 40 seconds. Each error value is an absolute value that is the
maximum error. In order to reduce error from the data delay, there are ways
to increase the electronic device load capacity by using the high performance
computers, sensors and devices, or to increase the network performance of
the controller. Also, there is a way to use the LTE modem to increase the
controller and network performance. If a modem to make the controller an
access point (AP) is used to reduce the data delay, one can effectively reduce
the error [13].

CONCLUSION
This paper deals with the development of temperature controlled cargo bay
to protect goods from decay due to external environmental factors. A small
 Development of Drone Cargo Bay with Real-Time Temperature Control 395

controller, Raspberry Pi, was used to control the temperature control device,
and a cloud server was used to solve the communication distance limitations.
As a result of detecting the change of the cargo bay internal temperature
for 15 minutes, it was confirmed that the set temperature was effectively
maintained and monitored, despite some small errors. The developed device
can be used to protect and safely transport items that are easily corrupted
and deteriorated, such as medical blood products that are very sensitive to
outside temperature variations. Even though the proposed device showed a
small range of errors during the operation, this problem can be easily solved
by using more accurate devices. The method provides a new approach to
either raising or lowering the inside temperature of drone cargo bay, and such
can be done remotely in real-time. By doing so, the temperature variations
during the flight can be verified, and the end user can be assured that the
transport items have arrived safely.

ACKNOWLEDGEMENTS
This work was supported by the Hyundai-NGV Future Technology Research
Fund.

CONFLICTS OF INTEREST
The authors declare no conflicts of interest regarding the publication of this
paper.
396 Unmanned Aerial Vehicles (UAV) and Drones

REFERENCES
1. Laksham, K.B. (2019) Unmanned Aerial Vehicle (Drones) in Public
Health: A SWOT Analysis. Journal of Family Medicine and Primary
Care, 8, 342-349. https://doi.org/10.4103/jfmpc.jfmpc_413_18
2. Griffiths, F. and Ooi, M. (2018) The Fourth Industrial Revolution-
Industry 4.0 and IoT. IEEE Instrumentation & Measurement Magazine,
21, 29-43. https://doi.org/10.1109/MIM.2018.8573590
3. Li, B., Fei, Z. and Zhang, Y. (2019) UAV Communications for 5G and
Beyond: Recent Advances and Future Trends. IEEE Internet of Things
Journal, 6, 2241-2263. https://doi.org/10.1109/JIOT.2018.2887086
4. Klinkmueller, K.M., Wieck, A.J., Holt, J.K. and Valentine, A.W. (2019)
Airborne Delivery of Unmanned Aerial Vehicles via Joint Precision
Airdrop Systems. AIAA Scitech 2019 Forum, San Diego, CA, 7-11
January 201, 2285. https://doi.org/10.2514/6.2019-2285
5. Amukele, T., Ness, P.M., Tobian, A.R., Boyd, J. and Street, J. (2016)
Drone Transportation of Blood Products. Transfusion, 57, 582-588.
https://doi.org/10.1111/trf.13900
6. Scott, J.E. and Scott, C.H. (2017) Drone Delivery Models for
Healthcare. Proceedings of the 50th Hawaii International Conference
on System Sciences, Hilton Waikoloa Village, HI, 4-7 January 2017,
3297-3304. https://doi.org/10.24251/HICSS.2017.399
7. Lee, J.H. (2017) Optimization of a Modular Drone Delivery System.
2017 IEEE International Systems Conference, Montreal, 24-27 April
2017, 1-8.
8. Erdos. D., Erdos. A. and Watkins, S.E. (2013) An Experimental
UAV System for Search and Rescue Challenge. IEEE Aerospace and
Electronic Systems Magazine, 28. 32-37. https://doi.org/10.1109/
MAES.2013.6516147
9. Xiang, G., Hardy, A., Rajeh, M. and Venuthurupalli, L. (2016) Design
of the Life-Ring Drone Delivery System for Rip Current Rescue. 2016
IEEE Systems and Information Engineering Design Symposium,
Charlottesville, VA, 29April 2016, 181-186. https://doi.org/10.1109/
SIEDS.2016.7489295
10. https://www.hankookilbo.com/News/Read/201903201101062687
11. https://fortune.com/2019/01/07/delivery-drones-rwanda/
12. Lee, S.-H., Yang, S.-H. and You, Y.-M. (2017) Design and Development
of Agriculture Drone Battery Usage Monitoring System Using
 Development of Drone Cargo Bay with Real-Time Temperature Control 397

Wireless Sensor Network. The International Journal of Advanced

Smart Convergence, 6, 38-44.
13. Krishna Chand, G.S.L., Lee, M. and Shin, S.Y. (2018) Drone Based
Wireless Mesh Network for Disaster/Military Environment. Journal
of Computer and Communications, 6, 44-52. https://doi.org/10.4236/
jcc.2018.64004
 INDEX

Symbols automatic cardioverter defibrillator

(AED) 319
3D cartesian paths 81, 86 automaticmanual switching 131
A Autonomous planning 41
autonomous takeoff 4
acceleration 80, 81, 83, 86, 90, 91, autonomous unmanned aerial sys-
93, 95, 96, 101, 108 tems (AUAS) 38
access point (AP) 394 autonomy 39
actuators 65
Adaptive control algorithms 25 B
Aerial Vehicle Control Language Backstepping 20, 25, 32, 33, 34
(AVCL) 80 Backstepping control 25
aerodynamic forces 122 battery 318, 319, 323
aircraft 322 blade element theory (BET) 270
algorithm 392 Bluetooth 4, 6, 7, 8, 9, 12, 13, 15,
ammonia volatilization 345 16, 17
angle of attack (AOA) 271 broadband wireless access 5
aquadcopter 61 Brushless motors 65
attitude control 115, 131, 134, 136
Attitude Controller block 67 C
400 Unmanned Aerial Vehicles (UAV) and Drones

Calibrated Reflectance Panel (CRP) 330

351
F
Catalina Island Conservancy (CIC)
370, 371 Federal Aviation Administration
cell decomposition 143 (FAA) 373
chlorophyll 345, 346 feedback control algorithm 220
Civil Certificate of Waiver Authori- feedback linearization (FBL) 26
zation (COA) 373 Finite-Difference Time-Domain
clockwise (CW) 215 (FDTD) Method 147
cluster heads (CHs) 202 finite state machine (FSM) 235
Commercial off the Shelf (COTS) fire detection 3
377 Fixtures 319
communication 390, 395 flight mission 4
Compact Airborne Spectrographic forest surveillance 3
Imager (CASI) 346 formation leader (FL) 202
computer systems 114 Fuselage 122
coordinate system 21, 22 fuselage body 9
Coriolis terms 262, 265 fuzzy logic 28, 29
D G
Data acquisition 127 Gaussian process 244
deconfliction 38, 39, 43, 49, 51, 52 genetic algorithm 28
Disturbance 69, 71 Geographic Information Sciences &
Downwelling Light Sensor (DLS) Technology (GIST) 371
350 geographic information system
Dubins waypoint navigation (DWN) (GIS) 232
209, 211, 224 Global Positioning System receivers
(GPS) 113
E
graphical user interface (GUI) 303
Earth remote sensing (ERS) 327 gravity 122, 123
Electronic Speed Controllers (ESC) Green Normalized Difference Veg-
65 etation Index (GNDVI) 347
elementary motion detector (EMD) guaranteed time slot (GTS) 6
296
H
environments 142
Euler angles 62, 64 helicopter dynamics 114, 117
Euler-Lagrange method 263
Evolutionary algorithms 114 I
Experimental Railway Ring (ERR) inertial measurement unit (IMU) 9,
 Index 401

297, 305 Nitrogen 343, 344, 345, 363, 364,

Institute for Creative Technologies 365, 366, 367
(ICT) 370, 371 nonlinear aerodynamic effects 181
interface Babelfish 305 Normalized Difference Vegetation
Internet Communications Engine Index (NDVI) 381
(ICE) 305
O
iterative feedback tuning (IFT) 284
iterative learning control (ILC) 283 Obstacle Field Navigation (OFN)
82
L
One World Terrain (OWT) 370
landing maneuvers 4 Optimization algorithms 27
Linear Quadratic Regulator (LQR) orientation vector 186, 187
20 Ostafievo airport 336
logical architecture 10, 12
P
M
packet loss ratio (PLR) 14
machine learning 28 partial least squares (PLS) 283
maneuverability 20 Path Command block 67
maneuvers 115, 127, 131, 134, 136, PHP language 390
137, 139 Position Controller block 67
marine pollution 230 principal component analysis (PCA)
Maritime activity 230 283
mass moments of inertia 22 probability density 244
master node (MM) 9 propeller 317, 318, 321, 322, 323
Mathematical models 117 propellers 61, 65
Micasense Atlas software 350 proportional Integral Derivative
microaerial vehicles (MAVs) 296 (PID) 60
Mission Planner (MP) 84, 105 pseudorandom binary sequence
mobile cellular networks 5 (PRBS) 277
mobile robot formation control 180 pulse width modulation (PWM) 20
model free adaptive control (MFAC)
Q
283
momentum theory (MT) 270 Quadcopter Control mixing block
Motion detection 296 67
Quadcopter Dynamics block 67
N
Quadcopters 60
neural networks 28, 29 quadrotor 255, 256, 257, 258, 259,
Newton-Euler formalism 21, 167 260, 261, 262, 263, 264, 265,
nitrate leaching 345 266, 267, 268, 269, 270, 271,
402 Unmanned Aerial Vehicles (UAV) and Drones

272, 273, 274, 275, 276, 277, T

278, 279, 280, 281, 282, 283,
Tag Image File Format (TIFF) 350
284, 285, 286, 287, 288, 289,
Thermic ray 388, 392
290, 291, 292, 293, 294
Torque 118, 122
quadrotor aerial robot 163, 164, 165
translational rotation matrix 182
quadrotor mass 166, 173
quadrotor system 297 U
quadrotor unmanned aerial vehicle
19, 20 Unmanned Aerial Systems (UAS)
79
R Unmanned Aerial Vehicles (UAVs)
344
radar antenna 147, 148
Radar Cross Section (RCS) 144 V
radar model 146
radars 142, 143, 148, 149, 152, 155 vegetation indices (VIs) 344, 345
radar satellite information 331 vertical take-off and landing (VTOL)
railway transportation 328 20, 164, 256
Reichardt model 297, 300 virtual machine (VM) 9
relay nodes (RN) 202 virtual reality/augmented reality
Remote License Pilots (RPL) 372 (VR/AR) 370
robots 229, 230, 231, 234, 236, 237, virtual reference feedback tuning
240, 244, 245, 246 (VRFT) 284
Robust control algorithms 26 W
Root Mean Squared Error (RMSE)
351 watershed pollution 3
rotor 20, 21, 22 waypoint navigation (WN) 209, 210
WiFi network 349
S wireless local networks (WLAN) 5
satellite networks 5 wireless personal networks (WPAN)
satellite photography 332 5
scatternet 12 Wrigley Institute for Environmental
Sliding Mode Control (SMC) 24 Studies (WIES) 372
snowfalls 328 Z
soil denitrification 345
Spatial Sciences Institute (SSI) 369, zone surveillance 3
370
state vector 63
swarm robotics 229, 230, 253, 254