Design and Research of UAV Autonomous Grasping System

Juntong Qi, Jinan Kang, Xiang Lu

Abstract—High-speed development of unmanned aerial environment obviously. If UAV equipped with the mobile
vehicles extends the applications of robots from the manipulator, the scope of its application will be more
ground to the sky, which has brought great changes to extensive. However, the research in this area is still relatively
people’s social life. However, most commercially poor at present.
available UAVs are not equipped with robotic arms. The According to the type of rotor, UAV can be divided into
UAV with robotic arm has the ability to manipulate fixed-winged aircraft and Rotary-wing aircraft. Nowadays,
surrounding objects, which will greatly expand the fixed-winged aircraft technology is relatively mature.
working scope of the UAVs. In our research, we designed However, Rotary-wing aircraft has the characteristics of
and completed a practical UAV autonomous grasping flexible flight posture and small requirements of the landing
system in this experiment. UAV autonomous grasping point, which help to complete tasks such as vertical lift, low
system consists of five sub-systems, including flight hover and low speed flight. Therefore, it is favored by
control system, motion capture system, wireless research institutions and university.
communication system, ground station control system Compared with traditional aerial robots, UAV autonomous
and Single-DOF Gripper System. The main difficulty lies grasping system (UAVAGS) possesses perfect application
in the coupling of the dynamics between the four-rotor prospect. Typical applications could be diffusing an
unmanned aerial vehicle and the manipulator on it, explosive device in a hazardous area, performing
mathematical modeling and etc. In addition, the maintenance on a tree or building, or clearing rubbish in the
four-rotor UAVs need to generate the appropriate flight air. The UAV system with robotic arms is different from the
path according to the actual pose of the target object in control of traditional UAVs. It is faced with lots of challenges
order to ensure the efficiency of the task. Based on this because of the introduction of robotic arms. The main
system, we have carried on the preliminary research on difficulty lies in the coupling of the dynamics between the
flight control and trajectory planning of the UAV system four-rotor unmanned aerial vehicle and the manipulator on it,
equipped with the manipulator. The results of the mathematical modeling [1]
research will be helpful to implement and improve the Although there are lots of difficulties in the UAVAGS,
system in the future. more and more attentions are paid to the development of
rotary-wing aircraft system which is capable of grasping
Keywords—UAV; Robotic arm; Autonomous Grasping ; objects both at home and abroad. Harbin Institute of
Trajectory planning ; Technology has designed a four-rotor UAV equipped with a
mechanical arm, which is used to clean up dangerous objects
in hazardous areas [2].Shenyang Institute of Automation also
I. INTRODUCTION designed a Hex-Rotor Aerial Robot equipped with a
With the continuous development of microelectronics five-degree-of-freedom manipulator ,which is capable of
technology, GPS, autonomous control and digital grasping objects outdoors with the help of GPS and other
communication technology, UAV has gradually become a auxiliary equipment [3]. Drexel Autonomous Systems Lab
new type of air power, which has been widely used in designed an UAV called MM-UAV [4], which is able to open
homeland security, military operations, and civilian-sector or close the valve in hazardous area. In 2011, Utah State
applications. UAV has many advantages, such as strong University conducted a research on the four-rotor UAV
viability, flexibility, suitable for harsh environment, etc. Due crawling objects indoors. The main contents of the research
to these advantages, UAVs have the potential to perform include accurate positioning, recognition and operation of
many tasks in locations that would be impossible for an target object and flight stability of the UAV [5].In 2013,
unmanned ground vehicle to reach. However, most tasks Vijay Kumar and other scientists developed a flying-type
performed by UAVs are only passive tasks, such as four-rotor UAV that can accurately catch objects during
environmental monitoring, shooting and avoidance, which high-speed flight [6]. This flying-type UAV can reduce the
lacks the ability to interact with objects in the surrounding hover time to increase the battery life as a result of high
speed.
JUNTONG QI is professor of the School of Electrical and Information Generally, UAVAGS consists of two parts: flight platform
Engineering, Tianjin University, Tianjin, China (e-mail: [email protected]). and grasping mechanism. The flight platform includes
JINAN KANG is master student of the School of Electrical and multi-rotor vehicle, coaxial rotorcraft [7] and ducted-fan
Information Engineering, Tianjin University, Tianjin, China (e-mail: aerial vehicle [8]. The grasping mechanism can be divided
[email protected]).
XIANG LU is engineer of the EFY intelligent control Co. Ltd, Tianjin, China into two categories: single-DOF grippers [9] and multi-DOF
(e-mail: [email protected]). manipulator system [10]. Nowadays, multi-rotor aircraft

978-1-5386-3107-2/17/$31.00 ©2017 IEEE

126
equipped with multi-DOF manipulator is widely used. The barometer. The software structure diagram for PIXHAWK is
reason is that the simple structure of multi-rotor aircraft and shown in Fig.2.
relative low coupling are beneficial to designing controller.
Multi-DOF manipulator can help flight platform to achieve
much more flexibility when performing the tasks
[11].However, in order to improve the stability of the system, Mavlink Radio Modem

this experiment chooses to use the combination of four-rotor Onboard Computer

UAV and single-DOF grippers. Flight Management

For the research of UAVAGS, the team in the EFY Micro SD Card
Mission Management System Health
technology company has done a lot of preparatory work.
Finally, as is shown in Fig.1, an actual physical UAVAGS is
built. It comprises four-rotor aircraft and single-DOF gripper,
which are introduced in Section II. RC input

Actuator Interface VSHHG

Trajectory Control

Mixer SRVLWLRQ
5DWHVHWSRLQW
Position Control

5DWHVHWSRLQW
Attitude Control

Anglar Rate Control

Attitude Estimation

Sensor Readout Position Estimation

Fig.1. UAVAGS in EFY Technology
Fig.2. Software structure diagram for PIXHAWK

Motion capture system is an accurate system for indoor

II. SYSTEM COMPOSITION
positioning. It is used as the positioning system where GPS is
A. Four-Rotor Aerial Robot invalid, as shown in Fig.3. There are many kinds of motion
capture systems. In this experiment, we choose OptiTrack
The four-rotor aircraft frame is made of Nylon and fiber. Its vision-based localization system. It mainly consists of
structure parameters are shown in TABLE I. following parts: 12 Prime41 infrared cameras, a Gigabit
switch, workstation and optical motion capture software
TABLE I. PARAMETERS OF FOUR-ROTOR AIRCRAFT
Motive.
Parameter Value Effective capture space is about 10 m long, 10 m wide and
5 m high, separated with safety net and large enough for flight
Tip-to-tip wingspan 450cm crawling experiment. The motion capture system provides a
Arm length 18cm real time positioning for ground station control system at the
Propeller 10cm precision of millimeter.
Frame height 9cm
Total mass 1.5kg
3cells battery 0.3kg

The four–rotor aircraft system is mainly composed of four

sub-systems, including flight control system, motion capture
system, wireless communication system, and ground station
control system.
In the experiment, PIXHAWK, an open source controller
is chose as master of flight control system. PIXHAWK runs a Fig. 3. Motion capture system
very efficient real-time operating system, which provides a
POSIX-style environment. The function of PIXHAWK is Ground station control system is installed in the PC and
powerful, including Mission Management, Flight runs in the ROS environment. The main function of ground
Management, Attitude Control, Position Control, Trajectory station system are real-time data acquisition and transmission,
Control, etc. Besides, PIXHAWK also integrates several integration of rotorcraft control algorithm, sending control
sensors, such as gyroscope, accelerometer, magnetometer and command timely, visual information processing, etc.
Wireless communication system is an important part of the

127
experiment, which is mainly responsible for mutual Several avionics sensors, such as gyroscope, accelerometer
communication between ground station and flight control and barometer, are integrated on the aircraft autopilot.
system. In the experiment, to ensure the stability of the Therefore, in order to receive accurate attitude, position and
four-rotor aircraft, flight control system need receive velocity information of four-rotor aircraft, PIXHAWK should
real-time four–rotor aircraft location and attitude information, be placed at the center of vehicle. Then, the battery is fixed
which results in the communication frequency up to above the mechanical claw, making the center of gravity
100hz.Due to the phenomenon of large delay and data packet locating near the symmetrical center of the whole structure.
loss, ordinary wireless communication module will not meet As the height of the mechanical claw is higher than that of
the needs of experiment. After a lot of comparative tests, we the aircraft landing gear, the four-rotor aircraft is equipped
choose RS9113WiFi module as the wireless communication with high density sponge landing gear, which will reduce the
system in the end. impact force when aircraft landing. The height of landing gear
is shorter than that of fully closed mechanical claw. Thus, the
B. Single-DOF Gripper System
claw is in the open state when the aircraft takes off.
Considering the actual needs of the experiment and the Generally, the UAVAGS consists of five sub-systems,
control complexity of the robot arm, we choose a single-DOF including flight control system, motion capture system,
gripper as operating device. It is attached below the four-rotor wireless communication system, ground station control
aircraft base. As is shown Fig.4, mechanical claw is made of system and Single-DOF Gripper System. Besides, to ensure
aluminum alloy and glass fiber, so its quality is very light. The the safety of the experiment, we prepared a remote control
inner edge of claw is designed in a wavy shape, which module which is used to control the UAV manually. The
increases contact area and friction force with the object. It control structure diagram is shown in Fig.6
weighs 218g (including actuator) and will grab about 1kg item. XKSUZKIUTZXUR 9OTMRK*5,-XOVVKX
As the control of Single-DOF Gripper is relatively simple, it LROMNZIUTZXURY_YZKS 9_YZKS
SUJ[RK
can be directly controlled by PIXHAWK. Therefore, the
hardware structure will be simplified greatly. Besides, to
ensure the safety of experiment, it also can be remotely
controlled with handle by ground control station.
]OXKRKYYIUSS[TOIGZOUT
Y_YZKS

MXU[TJYZGZOUTIUTZXUR
SUZOUTIGVZ[XKY_YZKS
Y_YZKS
Fig. 6. Control structure diagram of UAVAGS

III. TRAJECTORY PLANNING OF GRASPING MOVEMENT

Fig. 4. Single-DOF gripper
In the experiment, we need to plan the flight trajectory
while UAV is grasping object. In general, trajectory planning
C. Overall System algorithm is based on the optimal control theory, and the
The overall hardware and software of UAVAGS have been optimal flight trajectory will be obtained by building a
introduced, and the hardware structure is shown in Fig.5. minimum cost function. There are many UAV trajectory
planning algorithms, such as genetic algorithm, artificial
potential field method and fast expansion random tree (RRT)
algorithm [12]-[14], but these algorithms are very complex
and very difficult to use.
Therefore, this experiment chooses a very simple and
practical method. Firstly, we extract a few flight key points
and divide complete path into multiple straight path. Then,
four-rotor aircraft flies to the top of target object step by step,
and lands to the grasping point slowly. Finally, mechanical
claw grasps the object. The advantage of this method is simple
planning, the control of the UAV and robotic gripping
separated, small workload, easy to achieve. However, this
Fig. 5. UAV Autonomous Grasping System hardware trajectory planning method does not impose a corresponding
kinetic constraint on the track point, there is no guarantee that

128
all trajectories will be connected together smoothly. The Fig. 9. Position of UAVAGS (normal state)
desired flight trajectory is shown in the Fig.7.
flight trajectory
The position change of UAVAGS when mechanical claw is
in movement is shown in Fig.10. UAVAGS is hovering at a
Desired

0.8
desired position [0.65 m; -0.48 m; 0.6 m].
0.7 X--axis
0.09

0.6 Realistic
0.08
Desired

0.5 0.07

X(m)
0.4 0.06
Z(m)

0.3
0.05
0 500 1000 1500 2000 2500

time(ms)
0.2
2
Y-axis
-0.03
0.1 1

Realistic
0
0 -0.04 Desired
-1
1.5

Y(m)
1
0.5 -2
0 -0.05
-0.5
-1
-1.5 -3 X(m)
˄ m˅ -2
Y
-0.06
0 500 1000 1500 2000 2500

time(ms)
Z-axis

Fig. 7. Desired flight trajectory 0.61

Realistic

0.6 Desired

Z(m)
0.59

0.58
0 500 1000 1500 2000 2500

IV. HOVERING EXPERIMENT AND GRASPING

time(ms)

EXPERIMENT Fig.10. Position of UAVAGS (mechanical claw movement)

A. Hovering Experiment By comparing Fig.9 and Fig.10, it can be seen obviously

With mechanical claw attached below four-rotor aircraft that the amplitude of position during mechanical claw
base, the force in the Z-axis direction will change a lot, which movement is larger than that in normal state. The hovering
is harmful to the stability and accuracy of UAVAGS. After position mean error and root-mean-square error of UAVAGS
adjusting the PID parameters in the Z-axis direction [15], the in normal state and during manipulation movement are
control effect has been significantly improved and meets the provided in TABLE II, which are calculated by Matlab.
needs of test. TABLE II. HOVERING POSITION Error OF RAMS
The movement of mechanical claw will influence the flight
stability of four-rotor aircraft. The state change of UAVAGS
Hovering mean error(m) root-mean-square
is researched by the hovering experiment, when mechanical
Position error(m)
claw is open and closed. The movement of mechanical claw is
of Normal During Normal During
presented in Fig.8. In Fig.8-(a), single-DOF gripper is
UAVAGS state movement state movement
absolutely open, which is normal state. As is shown in
Fig.8-(b), when the trigger command is received from ground
x 0.0010 0.0040 0.0062 0.0094
control station, the single-DOF gripper closes automatically
y 0.0009 0.0045 0.0053 0.0077
to grasp the object.
z 0.0021 0.0071 0.0079 0.0091

Through a lot of experiments, numerical results show that the

position errors during the manipulation movement are larger
than those in normal state, which demonstrates that the effect
of the mechanical claw movement cannot be ignored. By
adjusting the PID parameters in the x, y and z-axis directions,
the influence of single-DOF gripper movement can be
reduced but not eliminated.
Fig. 8. Hovering experiment of UAVAGS

The position change of UAVAGS in the normal state is B. Grasping Experiment

shown in Fig.9, where the sampling frequency is 100 Hz. In this experiment, three 11cm long, 3 cm wide and 6cm
UAVAGS is hovering at a desired position [0.035 m; –0.029 high cuboid target objects where visual markers are installed
m; 0.6 m]. are placed randomly in effective capture space. Therefore,
objects position can be obtained as operating target position.
To transmit accurate position and attitude data of UAVAGS
X--axis
0.06

Realistic

0.04 Desired

and target objects from motion capture system, two steps are
X(m)

0.02

needed to be accomplished by ground station control system:

0
0 500 1000 1500 2000 2500

time(ms)
Y-axis

firstly, receiving data from localization system; then, sending

-0.01

Realistic
-0.02
Desired

data to flight control system by Wireless communication

-0.03
Y(m)

-0.04

-0.05

system
0 500 1000 1500 2000 2500

time(ms)
Z-axis
0.62

0.6
Realistic
Desired
We intend to make the UAVAGS grasp three target objects
Z(m)

0.58

0.56
with markers in order and fly back to original position. Firstly,
the UAVAGS with markers takes off and automatically
0 500 1000 1500 2000 2500

time(ms)

129
follow the set trajectory to approach target object, with the
distance about 10 cm over the object. Next, single-DOF
gripper realizes the grasping of tube, while the UAVAGS is
hovering with the help of flight control system.
In actual grasping experiment, we found that the distance
between adjacent key points is too far, which leads to aircraft
shock seriously even out of control. Therefore, we choose to
add a straight line between the key points. The experiment
results vary depending on the number of track points between
adjacent key points.
a. No track point
In the case of no track point, as is shown in Fig 10 (flight Fig. 12. Flight data in x, y, z-axis with 20 track point
data in x, y, z-axis), the aircraft has a serious shock, which is
harmful to stability. After a number of comparative tests, we found that adding
2
X--axis
Realistic
Desired
track points between the key points will improve the stability
1

0
and success rate of UAVAGS grasping object.25 track points
X(m)

-1

-2
is best choice in this experiment.
0 1000 2000 3000 4000 5000

time(ms)
Y-axis
6000 7000 8000 9000 10000

The whole grasping process is shown in Fig.13. First, the

2

1
Realistic
Desired aerial manipulator takes off and approaches the target object,
0
which is shown in Fig.13-(a) and (b). Then, as is shown in
Y(m)

-1

-2
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Fig.13-(c), the UAVAGS is hovering in the desired position
and getting ready to grasp the object. Next, the target is
time(ms)
Z-axis
1
Realistic

grasped and placed in desired position, which can be seen in

Desired

0.5
Z(m)

Fig.13-(d). Finally, UAVAGS returns and lands in the

original position, which is shown in Fig.13-(e) and (f).
0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000

time(ms)

Fig. 10. Flight data in x, y, z-axis with no track point

b. 40 track point
In the case of 40 track point, as is shown in Fig.11, aircraft
flight tracking effect is better and flight trajectory is relatively
smooth. However, the time required to complete the task is
also lengthened, which reduces the efficiency of grasping
objects.

Fig. 11. Flight data in x, y, z-axis with 40 track point

c. 20 track points
In the case of 20 track point, as is shown in Fig.12, aircraft
flight tracking effect is better and flight trajectory is relatively
smooth. Besides, the time required to complete the task is
relatively short.

Fig. 13. Grasping experiment of UAVAGS

130
 C. Conclusion from experiment [2] Wangshuai, Zhouyang, “Four-rotor flight catcher for cleaning up
dangerous area items,” Ordnance Automation, vol.3, pp. 78-80, Augˈ
Hovering and grasping experiment research is to 2011
demonstrate the application of UAVAGS, analyze the [3] X Meng, Y He, F Gu,L Yang,B bai, “Design and implementation of
problem included and also verify the feasibility of this rotor aerial manipulator system," Proceedings of the IEEE International
system. Conference on Robotics and Biomimetics, Qingdao,2016, pp.673-678.
[4] M. Orsag, CM. Korpela, S. Bogdan, and PY. Oh, “Hybrid adaptive
The work of UAVAGS autonomous grasping and placing control for aerial manipulation,” Journal of intelligent & robotic
target objects is basically completed in the motion capture systems, vol. 73, no. 1-4, pp. 693–707, Sep. 2014.
system. From this experiment, the following essential [5] V Ghadiok , J Goldin, W Ren. “Autonomous indoor aerial gripping
problems are found and should be paid attention to in the using a quadrotor,” Proceedings of the IEEE/RSJ International
Conference on Intelligent Robots and systems. San Francisco, 2011, PP.
future work. 4645-4651.
(1) Grasping mechanism. A single-DOF gripper is chosen as [6] D Mellinger, M Shomin, V Kumar. “Control of Quadrotors for Robust
operating mechanism, which results in grasping form Perching and Landing,” Proceedings of the International Powered Lift
monotonous. In the future, we should increase the freedom of Conference, Philadelphia, 2010, pp. 205-225.
[7] M. Kobilarov, “Nonlinear trajectory control of multi-body aerial
robotic arm. manipulators,” Journal of Intelligent & Robotic Systems, vol. 73, no.
(2) Stable control of UAVAGS. As the quality of target object 1-4, pp. 679–692, Oct. 2013.
is very light, the effect of mechanical claw movement and [8] F. Forte, R. Naldi, A. Macchelli and L Marconi, “Impedance control of
object gravity is ignored in the process of grasping, which is an aerial manipulator,” American Control Conference, Montréa, 2012,
pp. 3839-3844.
harmful to stable control of UAVAGS. In the future, we may [9] PEI. Pounds, AM. Dollar, “Stability of Helicopters in Compliant
design variable PID control algorithm, one for stable flight, Contact Under PD-PID Control,” IEEE Transactions on Robotics, vol.
the other for grasping target. 30, no. 6, pp. 1472–1486, Oct. 2014.
(3) Positioning System. At present, motion capture system is [10] K Kondak, F Huber, M Schwarzbach and M Laiacker. “Aerial
manipulation robot composed of an autonomous helicopter and a 7
used as positioning system. However, its working area is degrees of freedom industrial manipulato,” IEEE International
limited and it is difficult to apply outdoors, which greatly Conference on Robotics and Automation, Hong Kong, 2014, pp.
limits the working scope of UAVAGS. In the future, we 2107-2112.
intend to use GPS as positioning system outdoors. [11] B Yang, Y He, J Han, G Liu and G Zhang, “Survey on Aerial
Manipulator Systems,” ROBOT, vol. 37, no. 5, pp. 628–640, Sep. 2015.
[12] T Mclain, R Beard. “Trajectory planning for coordinated rendezvous of
unmanned air vehicles,” Fire Control & Command Control, vol. 34, no.
V. CONCLUSION AND FUTURE WORK 2, pp. 1247-1254, Jul, 2009.
In this paper, we give a thorough introduction about flight [13] G Vachtsevanos, T Liang, G Drozeski, L Gutierrez. “From mission
planning to flight control of unmanned aerial vehicles: Strategies and
control system, motion capture system, wireless implementation tools,” Annual Reviews in Control, vol. 29, no. 1-2, pp.
communication system, ground station control system, 101-115,Sep, 2005.
Single-DOF Gripper System and flight path planning [14] X Yang, MY Ding, CP Zhou, “Fast marine route planning for UAV
algorithm. A UAVAGS consisting of a four-rotor aircraft and using improved sparse A* algorithm,” Fourth International Conference
on Genetic and Evolutionary Computing, Shenzhen, 2010, pp. 190-193
a single-DOF gripper is designed, implemented and
[15] PE Pounds, DR Bersak, AM Dollar, “Stability of small-scale UAV
researched in indoor environment, which lays a foundation helicopters and quadrotors with added payload mass under PID
for autonomous grasping work in outdoor environment. The control,” Autonomous Robots, vol.33, pp.129-142, Aug, 2012
whole system has been verified through hovering experiment
and grasping experiment. However, to achieve ultimate goal
of autonomous grasping in outdoor environment, there are a
lot of work to do in the future. We will concentrate on the
overall system dynamics modeling, stable control of flight,
positioning system, grasping mechanism and aerial
autonomous mission planning in the future work.

ACKNOWLEDGMENT
This paper is supported by The Major Program of the
National Natural Science Foundation of China (Grand No.
61433016) and The Science and Technology Program of
Tianjin (Grand No. 15ZXZNGX00280)

REFERENCES
[1] A. Chovancova, T. Fico, L. Chovanec, P. Hubinsk, “Mathematical
modelling and parameter identification of quadrotor,” Procedia
Engineering, vol. 96, pp. 172–181, Oct, 2014.

131