1.

INTRODUCTION

Navigation means the ability to wonder in the environment without colliding

with obstacles, the ability to determine one's own position, and the ability to
reach certain goal locations. The robot also can construct an internal
representation to its environment on the form of a map. This map can be used
in planning paths towards its goals locations. So, navigation system may
imply the following components: robot positioning system, path planning and
map building.

In order to achieve this objective, the robot needs to be equipped with

sensors suitable to localize the robot throughout the path it has to follow.
These sensors may give overlapping or complementary information and may
also sometimes be redundant [1,2]. These are four popular positioning
systems:-

1- Odometry (dead reckoning) – based navigation which uses encoders to

measure the rotation of the wheels and the steering orientation. The
vehicle can estimate its relative position using these measures.
2- Active beacons- based navigation system where three or more actively
transmitted beacons are located at known positions in the environment,
and a receiver on the vehicle. The absolute position of the vehicle is
computed from the measurements of the distances or angles with
respect to the beacons.
3- Landmark-based navigation system which is similar to active beacons
system but instead of active beacons a natural or artificial landmarks
are defined.
4- Map-based navigation system: In this system, the vehicle contains a
map to the environment. The vehicle estimates its position by matching
the acquired information from its sensors with the environment map.

Mobile robots generally carry dead reckoning sensors such as wheel

encoders and inertial sensors, also landmark and obstacle detecting and map
making sensors such as time of flight (TOF) ultrasonic sensors. Sensors
measurements in this case are to be fused to estimate the robot's position.

In this paper, the navigation system built on a mobile robot operating in a

warehouse is presented. Hybrid navigation system that combines the
perception and dead reckoning was found to be complementary and gives a
satisfactory operation of the mobile robot. The position estimate provided by
dead reckoning is corrected by matching the perception against a stored map.
Landmark-based navigation depends mainly on the agent's perception to its
environment. If the environment contains confusing information or few
perceptually distinguishable landmarks, the performance of these systems
decline. The perceptual aliasing problem can be solved by including the
odometry data to discriminate between the similar places.

2. Odometry and Odometry Errors

Odometry is the most widely used navigation method for mobile robot
positioning. It is well known that odometry provides good short term accuracy,
is inexpensive and allows very high sampling rates. However, the
fundamental idea of odometry is the integration of incremental motion
information over time, which leads inevitably to the accumulation of errors.
Particularly, the accumulation of orientation errors will cause large position
errors which increase proportionally with the distance travelled by the robot.
Odometry is used in almost all mobile robots, for various reasons: Odometry
data can be fused with absolute position measurements to provide better and
more reliable position estimation [3,4];- Odometry can be used in between
absolute position updates with landmarks;- Many mapping and landmark
matching algorithms assume that the robot can maintain its position well
enough to allow the robot to look for landmarks in a limited area.

2.1 Systematic and Non-systematic Odometry Errors

The correct functioning of a mobile robot requires no faults( fault means that
the robot is functioning outside its specification limits, being unable to
accomplish normally its tasks). An error in orientation during the robot
movement may lead to deviations ( the amplitude of the deviations depends
on the gravity of the error). If these deviations are causing the incapacity of
the system to realize its task, then the error became a fault.

Odometry is based on simple equations that are easily implemented and

that utilize data from inexpensive incremental wheel encoders. However,
odometry is also based on the assumption that wheel revolutions can be
translated into linear displacement relative to the floor. This assumption is of
limited validity. One extreme example is wheel slippage :- if one wheel was to
slip on, say an oil spill, then the associated encoder would register wheel
revolutions even though these revolutions would not correspond to a linear
displacement of the wheel. There are also, several other subtle reasons for
inaccuracies in the translation of wheel encoder readings into linear motion.
All of these error sources fit into one of two categories: systematic errors and
non-systematic errors.

a- Systematic Errors:-
- Unequal wheel diameters.
- Average of actual wheel diameters differs from nominal wheel
diameter.
 - Actual wheelbase differs from nominal wheelbase.
- Misalighment of wheels.
- Finite encoder resolution.
- Finite encoder sampling rate.

b- Non-Systematic Errors:-

- Travel over uneven floors.

- Travel over unexpected objects on the floor.
- Wheel-slippage due to: slippery floors, over acceleration fast
turning (skidding), external forces (interaction with external bodies),
internal forces (castor wheels) and non-point wheel contact with the
floor.

Systematic errors are particularly grave because they accumulate

constantly. On most smooth indoor surfaces, systematic errors contribute
much more to odometry errors than non-systematic errors. However, on rough
surfaces with significant irregularities, non-systematic errors are dominant.
The problem with non-systematic errors is that they may appear unexpectedly
(for ex., when the robot traverses an unexpected object on the ground, and
they can cause large position errors.

The majority of researches are focusing on the systematic odometry errors

using offline techniques based on calibrations. Also, it has to be taken into
account that the mobile robot is moving in dynamic environments, where the
trajectory is never the same [5].

To correct the errors in positioning resulting from the odometry system and
for safe navigation and obstacle avoidance, the robot needs to be equipped
with sensors suitable to localize the robot throughout the path it has to
follow[6]. Because ultrasonic sensors can provide good range information
based on the time of flight(TOF) principle for rather low expense, they have
been widely used in mobile robots applications[7-9].

3. Ultrasonic Sensors

Ultrasonic transducers are preferably used to obtain three- dimensional

information of the environment. Time-of-flight (TOF) ranging systems
measure the round-of-trip time required for a pulse of emitted energy to
travel to a reflecting object , then echo back to a receiver. Ultrasonic is
typically employed, they have many advantages:- Measure and detect
distance to moving objects;- Impervious to target materials, surface and
color;- Solid- state units have virtually unlimited , maintenance –free lifespan
and are not affected by dust, dirt or high– moisture environments.
 But some problems appear in sonar response. Ultrasonic sensors suffer
from unreliable sonar responses from the environment. For sonar – based
mobile robot in confined space, special attention should be paid to these
problems .The space is normally a closed environment.

As our concern is the navigation in confined spaces using multi- sonars,

we must understand why there are such unreliable readings in ultrasonic
sensor responses .Two major problems are discussed in the following [10]:

3.1 Angular uncertainty

The angular uncertainty means the uncertainty means the uncertainty in

the angle information of a sonar response from a detected object. Figure (1)
conveys the idea. when an ultrasonic sensor gets a range response of R
meters, the response simply represents a cone within which the object may
be present .There is no way to pin – point exactly where the position of the
object is . As shown in the figure, the opening angle of the ultrasonic sensor
is 2α and the object can be anywhere in the shaded region for the response
R.

Fig (1) Angular error of an ultrasonic sensor α is the half opening angle of
sonar cone, R is a sonar response.

3.2 Specular reflection

Specular reflection refers to the sonar response that is not reflected back
directly from the target object .In specular reflection , the ultrasound is
reflected away from the reflecting surface , which results in longer range
reporting or missing the detection of object all together [ 11,12 ] .

The specular reflection is due to different relative positions of the ultrasonic

transceiver and the reflecting surfaces. Figure (2) shows sonar responses in
two different situations. In figure (2a) the sensor transceiver axis is
perpendicular to the reflection surface, so most of the sound energy is
reflected directly back to the ultrasonic sensor.

However, in figure( 2b), because the sonar transceiver is not perpendicular

to the surface , much energy is reflected away . The amount of reflected
sound energy depends strongly on the surface structure of the obstacle and
the incidence angle [13]

Fig (2) Specular reflections

4. Robot Description

The mechanical design for the robot plays a critical role in the success of
the robot facility. The requirements of the mechanical facility of the robot is
the major question before beginning the mechanical design. The following
specifications should be met:-

1- Moving forward and backward without rotation.

2- Moving asides (right and left) without rotation.
3- Have the facility to rotate in a complete circle.
4- The robot will use a battery assembly with total volt 48V.
5- The control unit and battery charger should be on the robot itself.

The driving system of the robot is composed of 4 wheels each of them

equipped with a separate electric motor. A front and rear steering system
were added to give flexibility in the motion planning for smooth navigation.
The mobile robot configuration is shown in figure (3).
 Fig (3) Mobile Robot Configuration.

4.1 Robot Positioning

Methods for robot positioning can be roughly categorized into groups:

relative and absolute position measurements. Because of the lack of a single
good method, developers of mobile robots usually combine two methods, one
from each category.

In this work, the relative method used is odometry. This method uses
encoders to measure wheel rotation and/or steering orientation. Odometry is
totally self-contained and it is always capable of providing the vehicle with an
estimate of its position, but the position error grows without bound unless an
independent reference is used periodically to reduce the error. Natural
landmark recognition was used as an absolute positioning measurement
system to correct periodically the right position of the robot.

4.2 Encoder, Decoder and Motion Controller Used

Optical incremental encoders are a mean for capturing speed and travelled
distance on a motor. Incremental encoders output square pulses as they
rotate. Counting the pulses tells the application how many revolutions, or
fractions of, the motor has turned. Rotation velocity can be determined from
the time interval between pulses, or by the number of pulses within a given
time period. Because they are digital devices, incremental encoders will
measure distance and speed with perfect accuracy. Quadrature encoders
have dual channels, A and B, which are electrically phased 90 o apart. This,
direction of rotation can be determined by monitoring the phase relationship
between the two channels. In addition, with a dual-channel encoder, a four
times multiplication of resolution can be achieved by counting the rising and
falling edges of each channel (A&B).

In this work, the HEDS5540, 3 channels high performance optical

incremental encoders, shown in figure (4), are used. This IC consists of
multiple sets of photo detectors and the signal processing necessary to
produce the digital waveforms. The digital output of channel A is in quadrature
with that of channel B (90 degrees out of phase). The encoder standard
resolution is 1024 counts per revolution.

The general purpose motion control IC, HCTL-1100, is employed. It frees

the host processor for other tasks by performing all the time intensive
functions of digital motion control. The HCTL-1100 provides position and
velocity control for DC, DC brushless and stepper motors. It receives its input
commands from a host processor and position feedback from an incremental
encoder with quadrature output.

Fig (4) Block Diagram of Encoder

4.3 ULTRASONIC SENSOR

Sensor model is the mathematical description of the sensory data obtained

from the physical sensing units. Ultrasonic sensors are used in this work to
build a map of the environment. The map contains information of the
boundaries of the environment and obstacles inside which are to be used in
the navigation. The ultrasonic sensor provides range information based on the
time-of-flight (TOF) principle as given in equation (1),

d=ν t (1)

where d is the round-trip distance, v is the speed of propagation of the pulse

and t is the elapsed time.
 4.3.1 WHY USE ULTRASONIC SENSORS?

According to [8,14], ultrasonic TOF ranging system is today the most

commonly used technique employed on indoor mobile robotics systems,
primarily due to the following reasons,

1. Low cost
Ultrasonic sensors are widely available at very low prices. In this aspect,
the ultrasonic sensor has great advantage over the laser scanner and other
sensors such as the camera.

2. Easy maintenance
For practical use, maintenance is an important issue. Ultrasonic sensors
are compact in design, light in weight and very reliable. It is also easy to
interface these sensors with other subsystems of the robot.

3. High range detection accuracy

The range detection from the ultrasonic sensor is very accurate.

4.3.2 SRF04 Ranger

The most commonly used sonar device for mobile robots is the well known Polaroid
ultrasonic ranging system.

After studying some types of ultrasonic sensors, we selected the [Devantech SRF04
Ranger Compact High Performance Ultrasonic Ranger] for use in this work. The description
and specifications of the SRF04 are as follows :

Description: This is a fantastic ultrasonic ranger that has an approximate range of 3cm to
3m. This ranger has a logic line used to trigger a pulse and the echo is returned on a second
line. Minimal power requirements and a compact, self contained design make this one of the
most popular detectors.

This block anodized aluminum housing can hold one SRF04 range finder. All necessary
hardware is included. Figure (5) represents the SRF04 ultrasonic sensor.

.Fig. (5) Devantech SRF04 Range

 SOFTWARE DEVELOPMENT

ATMEL AVR microcontrollers is easily programmed using the ATMEL

integrated development environment AVR studio 4, where assembly
programs can be edited, simulated, debugged and downloaded to the
microcontroller.

In our case to make our programs easy-made, efficient, portable and

readable, we decided to program in C. We’ve used IAR imbedded
workbench for ATMEL which is an integrated development environment
(IDE) provided by IAR corporation for C-programming of ATMEL
microcontrollers. By IDE is meant the full software development cycle
including source code editing, debugging, compiling and linking are
supported in one user friendly operator interface

Time Base Generation

It is clear that a microcontroller always depends on some sort of precise

clock (or oscillator) for its normal operation. The oscillator used in our
application is a 8 MHz crystal. By using one of the internal time/counter
modules of the microcontroller itself, this 8 MHz clock can be stepped
down to any desired value, the timer/counter module is fully controllable by
the software regarding its start/stop. This feature is particularly useful in
generating the time base only when desired and switching it off to reduce
HF noise when the time measurement task is completed.

URFS H/W BUILDING BLOCKS

As shown schematically in figure (6), the main building block beside the Atmel AVR
microcontroller is a programmable TTL counter chip 8254 which has 3 independent
counters, the gate of each counter is controlled by a dual input OR gate, a counter is
enabled as long as its gate is in high logic level.
 Fig. (6) URF board H/W building blocks

4.3.5 Ultrasonic Sensors Positioning

The main sensory system used to detect obstacles is the ultrasonic

sensors. From the specifications of the SRF04 sensor and to minimize the
dead zones during the navigation of the robot, the distribution of the sensors
on the robot body was designed as shown in figure (7).
 Fig.(7) Sensors positioning around the robot.

Although dead reckoning navigation is easy to implement but it suffers

from the drift problem which is serious in some navigation tasks.

So navigation to be reliable shouldn’t depend on one mechanism only.

Hybrid navigation system that combines the perception and dead reckoning
is better. Information from different sources can be fused to help the robot to
take the decision of what is the next step. The position estimate provided by
dead reckoning is corrected by matching the perception against a stored
world map. Landmark-based navigation depends mainly on the agent’s
perception to its environment. If the environment contains confusing
information or few perceptually distinguishable landmarks the performance of
these systems decline. The perceptual aliasing problem can be solved by
including the odometry data to discriminate between the similar places.

5. PROPOSED ROBOT NAVIGATION

5.1 AUTONOMOUS CONTROLLER APPROACHES

There are two main control approaches used in designing robot controller
which are:

1. Traditional approach
2. Behavior-based approach.

5.1.1 Traditional Approach

Traditional approach [15, 16] structuring a robot’s control into functional

modules: perception, planning, learning etc., and constraining as much as
possible the environment where the robot will operate. Creating a model of
the environment and the robot preprocess sensor information into abstracted
internal representations that are acted on by a central planner, then
instantiated the results to become actions that can be executed by the robot
to reach a specific goal. This can be represented in horizontal control
architecture as shown in figure (8).
 Fig.(8) Horizontal control architecture of the robot.

Control systems using this architecture solve their task in several steps. First,
the sensors input is used to modify the internal representation of the
environment. Second, based on the internal representation planning is made.
This results in a series of actions for the robot to take to reach a specified
goal. Third, this series of actions is used to control the motors of the robot.
This completes the cycle of the control system and it’s restarted to achieve
new goals.

General planning approach has several problems. Maintaining the model is in

many cases difficult because of sensor limitation or imperfection. The plans
produced by the planner often don’t give the effects in the real world that is
anticipated.

5.1.2 Behavior-Based Approach

In Behavior-based approach, instead of decomposing the task based on the

functionality, the decomposition is done based on task-achieving modules, are
called behaviors on top of each other as shown in figure (9), this is called
vertical control architecture.

Fig. (9) Vertical control architecture of the robot

Each behavior calculates a mapping from sensor inputs - the sensor inputs
relevant for the task of that behavior are used - to motor outputs. The
suggested motor outputs from the behavior with highest priority are used to
control the robot’s motors, or summed to generate one motors’ output. These
architectures are called behavior-based control approaches and represent
methodologies for endowing robots with a collection of intelligent behaviors.
Behavior-based approaches are an extension of reactive architecture, their
computation is not limited to lookup table and executing a simple functional
mappings. Behavior-based systems are typically designed so that the effects
of the behaviors interact in the environment rather than internally through the
system. We used this controller architecture in designing our controller

5.2 ROBOT'S MOTION TRAJECTORY

The vehicle will move inside and outside the warehouse as shown in the
following block diagram figure (10). Navigation of the vehicle is divided into
main three parts:

1- Inside warehouse

2- Outside warehouse

2- Maneuvering at the warehouse door.

Robot can figure out from the odometer system if it is inside or outside
warehouse. There are several rules bases according to sensors readings and
odometer that make the robot will switch from one controller to the other.
 Fig.(10) Motion Trajectory and Assembly Lines

5.3 ROBOT'S MOTION TYPES

The robot vehicle is designed to perform only two distinct kinds of motion in
the warehouse:

1) straight-line motion, where both motors are running at the same speed
and in the same direction,
2) rotation about the vehicle's center-point, where both motors are running
at the same speed but in opposite directions.

This approach is advantageous for several reasons:

1. Wheel slippage is minimized because of the simultaneous action or rest of

both wheels and because of the "on-the-spot" rotation action for turns.

2. A relatively simple control system may be used, since in either case the
only task of the controller is to maintain equal angular velocities,

3. The vehicle path is always predictable, unlike other motion strategies which
smooth sharp corners by an unpredictably curved path. A predictable path
is advantageous when global path planning, to avoid obstacles, is
employed.

4. The vehicle always travels through the shortest possible distance (straight-
line or "on-the spot" rotation).
5.4 THE PROPOSED CONTROLLER

At designing a motion controller for an autonomous mobile robot there is a

main problem that must be handled which is the obstacle avoidance problem.

The robot senses the obstacles using its sonar sensors.

5.4.1 Controller Inside The Warehouse

In this location of the warehouse, it is required from the robot to go to certain
shelf to pick up some components and then move towards the door as a first
step to drive these components to the required assembly lines. So, this task
can be interpreted as, it is required from the robot to move from initial (x,y)
position to certain target (x,y) position avoiding collision with any obstacles in
his way, like human, furniture,..etc, the robot can perform this behavior by one
of two strategies:

1- The angle between the robot's head and the line connecting between the
center of the robot and the target point is calculated as shown in figure (11)
then the robot rotates as shown in the following equation. Assume the
vehicle has to travel from a known present location (xo , yo, θo) to a new
location (xf , yf , θf ), the following procedure is performed to determine a
trajectory. First, the distance L and the slope θ of the straight line connecting
the present and final locations are calculated:

y f − yo
φ=arctan
x f −x o (2)

L= √(x f −x o )−( y f − y o ) (3)

.Fig. (11) Vehicle traveling from initial position to a final position

2- As the mission of the robot in the inner warehouse is to pick up

components so the target point will be always next to one of the walls. So we
suggest the robot will move parallel to the walls until arrive to the target point
to pick up the required components.
In our implementation we used the second strategy. It is more convenient
than the first one as the robot may need to pickup components from different
faraway shelves.

The readings of the grouped sensors are assigned one of two labels which are : far and
dangerous. Each group has different interpretation to the meaning of far and dangerous. For
example for the front sensors group the readings of the group ranging from 1m to 3m is far
while the readings from 3cm to 1m is dangerous. The expert system rule base is presented
in Table(1).

The difference between the readings of the two sides sensors are used to help the robot to
align to the left wall. We want the robot to keep certain distance to the wall which is in our
work 50 cm. the selection of this value to help the robot to turn smoothly at the corners
without getting stuck with the walls. While the other grouped sensors allow the robot to
detect the obstacles in his way and avoid by rotating around.

Table (1) Expert system's rule base

Right
Wheel Left Wheel FR-LT FR L∆ #

Forward Forward x Far 0 1

Forward Back x Far ve+ 2

Backward Forward x Far ve- 3

Backward Forward x Dang x 4

Backward Forward dang Dang x 5

5.4.2 Outside the warehouse

Over the outer-warehouse, the robot will move in one of three

predetermined paths to reach one of the three assembly lines. These
paths are saved in the robot. Each path is saved as a consequent number
of (x, y) points as shown in figure(10). In order that the robot will follow a
certain path, it has to trace its points in sequence.

References

[1] MOSHE KAM, XIAOXUN ZHU, PAULKATA, "Sensor Fusion for Mobile
Robot Navigation", Proceeding of the IEEE, Vol. 85, No.1, January 1997,
pp 108 -119.
[2] I. Ashokaraj, A. Tsousdos, P. Silsonand B. White, m" Sensor Based Robot
Localisation and Navigation : Using Interval Analyris and Extended Kalman
Filter" Proc of the MISC´99 Workshop on Applications of Internal Analysis to
systems And control, criona, Spain , 1999, pp158-165.

[3] F. Chenavier and J. Crowley, " Position Estimation for a Mobile Robot
using Vision and Odometry ", .proc. of IEEE inter, Conf. on robotics and
Automation , ICRA´92, Nice, France, May 12-14, pp2588-2593.

[4] J.M.Evans, " HelpMate: An Autonomous Mobile Robot Courier For

Hospitals " Proc. of the Int. Conference on Intelligent Robots and Systems,
IROS´94, Munich, Germany, Sept `12-16,pp1695-1700.

[5] Johann Borenstein and Liqiang Feng, "Measurement and Correction of

Systematic Odometry Errors in Mobile Robots", IEEE Trans on Robotics and
Automation, Vol12 , No. 6, December 1996, pp 869-880.

[6] Adrian Korodi, Toma L. Dragomir, "Correcting Odometry Errors for Mobile
Robots Using Image Processing", Proc. Of the International Conference of
Engineers and Computer Scientists 2010, Vol. п, IMECS 2010, March17-19,
2010, HongKong.

[7] Alberto Elfes, " Sonar–based Real- world Mapping and Navigation " IEEE
Journal of Roboties and Automation, Vol. RA-3, IV'3, 1987,PP 249-265.

[8] John J. Leonard and Hugh. F. Durrant – Whyte, "Directed Sonar

Sensing for Mobile Robot Navigation" Kluwer Academic Publishers,1992.

[9] S. Borthwick and H. Durrant–Whyte, "Dynamic Localization of Autonomous

Guided Vehicle", Proc. 1994 IEEE Int. Conf on Multi sensor Fusion, Las
Vegas , NV, 1994, pp92-97.

[10] Zou Yi, "Multi-Ultrasonic Sensor Fusion for Mobile Robots in Confined
Spaces", M.Sc thesis, School of Electrical & Electronic Engineering, Nanyang
Technological University, 2001.

[11] Michael Drumheller, "Mobile Robot Localization Using Sonar", IEEE

Transactions on Pattern Analysis and Machine Intelligence, Vol. PAMI-9,
No.2, pp.325-332, 1987.

[12] John Hwan Lim and Dong Woo Cho, " Specular Reflection Probability in
the Certainty Grid Representation", Transactions of the ASME, Vol.116,
pp.512-520, 1994.

[13] Roman Kuc and Billur Barshan, "A Spatial Sampling Criterion for Sonar
Obstacle Detection", IEEE Transaction on Pattern Analysis and Machine
Intelligence, PAMI-12(7), pp.686-690, 1989.
[14] Raschke U. and Johann Borenstein, "A Comparison of GridType Map-
building Techniques by Index of Performance", Proc. of IEEE Inernational
Conference on Robotics and Automation, Vol.3, pp.1828-1832, 1990.

[15] Arkin R., “Behavior-Based robots”, MIT Press, 1998.

[16] Russell S., Norvig P., “Artificial Intelligence: A Modern
Approach”, Prentice-Hall International, Inc., 1995.