Motion Control of Omnidirectional Mobile Robot

Based on Fuzzy PID

Qian Jia1, Chao Chang2,Shuqing Liu1, Luhao Zhang2 ,Sidi Zhang1
1. Industrial Center, Nanjing Institute of Technology, Nanjing 211100

E-mail: [email protected], [email protected], [email protected]

2. Postgraduate Department, Nanjing Institute of Technology, Nanjing 211100

E-mail: [email protected], [email protected]

Abstract: The motion model of omnidirectional mobile robot is created, the fuzzy PID control method is designed,
and the motion control parameters are adjusted online. Matlab is used to simulate the fuzzy PID control effect of the
robot, and the actual test of the fuzzy PID controller on the robot path tracking control effect. Computer simulation and
actual test results show that in the path tracking control of the robot, the fuzzy PID controller has faster response,
smaller overshoot, less fluctuation, stable change and stronger system stability than the conventional PID controller.
Key Words: Mecanum wheel, Omnidirectional mobile robot, Fuzzy PID

mobile robot without steering mechanism, skidding will

1 INTRODUCTION occur in the movement process, which inevitably affects
the tracking accuracy of the robot. Therefore, how to
Omnidirectional mobile mechanism has 3 complete eliminate or compensate the influence of wheel skidding
degrees of freedom of movement in the plane, namely, the and improve tracking accuracy is one of the important
movement along the X and Y axes and the rotation around contents of motion control research of omnidirectional
the Z axes. Mecanum wheel is compact in structure, mobile robot. At present, some domestic scholars have
flexible in movement and able to achieve all-round made relevant researches on Mecanum wheel. In literature
movement stably. At present, the research on wheeled [1], the application of fuzzy sliding mode trajectory
omnidirectional mobile mechanism represented by tracking control algorithm in Mecanum wheel
Mecanum wheel has become a hot topic. Compared with omnidirectional mobile platform was studied. Constraint
traditional wheeled mobile mechanism, the Mecanum conditions and singular configuration of Mecanum wheel
wheel omnidirectional mobile mechanism can realize omnidirectional motion system are discussed in literature
in-situ lateral, zero turning radius rotating movement, such [2]. In literature [3], the parametric modeling method of
as its high space utilization, the narrow operating Mecanum using API was studied. In literature [4], the
environment or need frequent to situations with the principle structure of Mecanum wheel is introduced and
movement of the flexibility, especially in the storage and the motion simulation based on ADAMS software is
transportation, service life, and other fields has a broad carried out. However, when the robot studied in the past
application prospect[1-4]. works in the actual environment, its motion trajectory is
Omnidirectional mobile robot is a typical multi-input quite different from the theoretical requirements, and there
multi-output nonlinear system. Compared with ordinary is a lack of comparison and verification between the
wheeled mobile robot, omnidirectional mobile robot has theoretical analysis and the actual measurement of the
more degrees of freedom and is difficult to control. In prototype. In this paper, the motion model of
addition, due to the wheel structure of omnidirectional omnidirectional mobile robot is established, and the fuzzy
PID control method is designed for the motion and control
characteristics of the robot. In the mathematical software
This work is supported by Natural Science Foundation of Jiangsu
Province(Grant No. BK20170766), the Opening Project of Jiangsu Key
Matlab/Simulink environment, the fuzzy PID control
Laboratory of Advanced Numerical Control Technology (Grant method of the robot is simulated and compared with the
No.SYKJ201605 ) and Natural Science Foundation of the Jiangsu Higher simulation results of the traditional PID control method.
Education Institutions of China(16KJB460011) Finally, the fuzzy PID control robot is applied to the
mobile robot to collect and analyze experimental data and

978-1-7281-0106-4/19/$31.00 2019
c IEEE 5149
 verify the actual effect of the fuzzy PID control robot on
path tracking and correction.
between expected pose (x 0ˈ y 0ˈ θ 0 )T and actual pose
2 OMNIDIRECTIONAL MOBILE ROBOT
(x sˈ y sˈ θ s )T is compared, and the desired speed of
KINEMATICS MODELING the robot is obtained by using fuzzy PID control method.
According to the inverse kinematics equation of the robot,
The Mecanum wheel comprises a hub and a set of rollers the expected rotational speeds of four wheels
mounted on the outer edge of the hub, with the roller axis
situated at an angle to the hub axis and the roller being (ω1ˈ ω 2ˈ ω3ˈ ω4 )T are calculated.
rotatable about its own axis. Robots using Mecanum
wheels can travel freely in tight working spaces without
3 Design of motion controller based on fuzzy
the need for large turning radii. Through the combined use PID
of the wheels and the coordinated control of rotational Omnidirectional mobile robot is a typical multi-input
speed and direction of each wheel, the vehicle body can be multi-output nonlinear system, and it has one more degree
moved and rotated in any direction in the plane [5], of freedom than ordinary wheeled mobile robot, so it is
considerably improving the transportation efficiency. difficult to control. In addition, due to the wheel structure,
Fig. 1 shows the wheel layout of the omnidirectional the motion process of the omnidirectional mobile robot
mobile robot. Let the generalized speed of the robot be based on Mecanum wheel will slip, which will inevitably
T
˄v x , v y , wz˅ T
㸪 ( w1 , w2 , w3 , w4 ) to indicate the affect the tracking accuracy of the robot. Combining the
actual operating conditions, the motion control system
rotational speed of each wheel. The angle between the axis design which combines fuzzy control and PID control
of the roller and that of the hub is the half-length and structure parameter self-tuning fuzzy PID algorithm to
half-width of the robots a and b, respectively, whereas the control the motor, the controller has the advantages of
wheel radius is R. fuzzy control is flexible and adaptable, and has the
characteristics of high precision of PID control, make the
motion control system for the high real-time performance,
strong robustness and stability design key points, and
through the fuzzy control rule base expansion, for the
motion control system is convenient to add other functions
[6].
Using linear CCD is installed on the robot to the robot car
positioning, actual output robot pose, and then compared
with expected position, get the distance deviation and the
deviation change rate ec, e input deviation to fuzzy
controller, through fuzzy reasoning, output by adjusting
Fig. 1 Omnidirectional mobile robot wheel layout the parameters of the ratio of the driven motor, by
adjusting the motor speed and direction of complete
According to kinematics analysis, the positive kinematics control the robot posture .
equation of the Mecanum wheel robot is : 3.1 Structure of fuzzy PID controller
ܸ௫ െ‫ ߙ݊ܽݐ ߙ݊ܽݐ‬െ‫ ߙ݊ܽݐ‬െ‫߱ ߙ݊ܽݐ‬ଵ
ͳ  ͳ ቏ ൦߱ଶ ൪  (1) The fuzzy self-tuning PID parameter control system is
൥ ܸ௬ ൩ ൌ ቎ ͳଵ ͳ
ோ
ସ ଵ ଵ ଵ ߱ଷ shown in Fig. 2, which is mainly composed of two parts:
߱௭ െ െ
௟బ ௟బ ௟బ ௟బ ߱ସ the PID with adjustable parameters and the fuzzy
–Š‡‹˜‡”•‡‹‡ƒ–‹ •‡“—ƒ–‹‘‹•˖ controller. Parameters can be adjusted PID to complete the
control of the controlled object. Fuzzy controller is used to
߱ଵ െܿ‫݈ ͳ ߙݐ݋‬଴
߱ଶ
ܸ௫ conduct online self-tuning of PID parameters kp, ki and kd
ଵ ܿ‫ߙݐ݋‬ ͳ െ݈଴
൦߱ ൪ ൌ ൦ ൪ ൥ ܸ௬ ൩ (2) by fuzzy reasoning method, so as to meet the requirements
ଷ ோ െܿ‫ͳ ߙݐ݋‬ ݈଴
߱ସ ߱௭ of system control for different e(n) and ec(n), so that the
ܿ‫ ͳ ߙݐ݋‬െ݈଴
controlled object has good dynamic and static performance
[7].
Among the equationsˈ l0 = l + L cot α Ǥ From Equation The PID controller adopts the classical PID incremental
(1) and (2)ˈthe different steering and rotational speeds of control, and its expression is as follows:
the four wheels can be used to move the robot to any —ሺሻ ൌ —ሺ െ ͳሻ ൅  ୮ ሾ‡ሺሻ െ ‡ሺ െ ͳሻሿ ൅  ୧ ‡ሺሻ ൅
position and direction. This condition serves as the ݇ୢ ሾ‡ሺሻ െ ʹ‡ሺ െ ͳሻ ൅ ‡ሺ െ ʹሻሿ (3)
theoretical basis for controlling the omnidirectional mobile ‡ሺሻ ൌ ”ሺሻ െ ›ሺሻ (4)
robot; in the robot path tracking, firstly, the difference

5150 The 31th Chinese Control and Decision Conference (2019 CCDC)
Where, u(k) is the output of the controller; r(k) is the 3.3 self-tuning rules of parameters
expected pose given; y(k) is the actual pose of the robot;
In order to comprehensively guarantee the dynamic and
e(k) is the deviation signal of the two.
static performance of the system, different PID control
r e u v
Parameter adjustable
PID
Controlled
object parameters can be considered in different stages of the
+ - dynamic process of the system. According to the influence
d/dt
KP KI KD of various control performance indexes, the self-tuning
requirements of PID parameters for the controlled process
ec Fuzzy controller under various errors and error change rates are
summarized [9]:
Fig. 2 Fuzzy self-tuning PID parameter control system 1) Under a large deviation, to achieve a better
fast-tracking performance in the system response and
3.2 Variables and membership functions avoid the differential saturation that may be caused
In this paper, the design of the fuzzy controller for two by the instantaneous variation in the initial deviation,
input three output structure, its input difference in the larger KP and the smaller KD should be
actual trajectory and target trajectory e and difference considered. The KI should be limited to avoid a large
rate of change of ec, its output to the adjustment of PID overshoot of the system response.
parameters kp, ki, kd value,namelyǻkp, ǻki, ǻkd. 2) Under a medium-level deviation, to achieve a
smaller overshoot in the system, a smaller KP and
Their linguistic variables, basic domains, fuzzy subsets,
appropriate KD and KI should be considered to
fuzzy domains and quantitative factors are shown in Table
ensure the response speed.
1.
3) When the deviation is relatively small, large KP and
Table1. Fuzzy input and output KI should be considered for desirable steady-state
performance of the system. The value of KD, which
is usually considered as a medium value, can be
larger when the variation rate of deviation is small.
The KD value should be smaller when the deviation
rate is larger.
The values mentioned in the above self-tuning
requirements are all relative to the initial value of the
parameter, and the PID parameter applied in practice is
self-tuning on its initial value, that is:
 ୔ ൌ  ୔଴ ൅ ο ୔
In E and EC and the output fuzzy variables KP, KI,  ൝  ୍ ൌ  ୍଴ ൅ ο ୍                  (5)
 ୈ ൌ  ୈ଴ ൅ ο ୈ
KD language variables and the theory of domain is
Note: where KP0, KI0, and KD0 are the initial values of P, I,
determined, must first determine the membership degree
of fuzzy language variables [8]. The commonly used and D, respectively, Δ KP, Δ KI and Δ KD are the
membership functions are b-spline basis function, gaussian self-adjusted PID parameters obtained by fuzzy reasoning.
membership function, triangle membership function, etc. The fuzzy relation model adopts a set of if-then fuzzy rules
Considering the requirements of simple design and to approach the input and output relations of the system;
real-time performance, triangle membership function is its fuzzy rules are as follows: If e is PB, and ec is NM, then
adopted. As shown in Fig. 3. KP is NS, KI is ZO, and KD is NS. The following is a list of
fuzzy control rules.
Table2. KP control rule table

Fig. 3 Input and output variable membership function graph

The 31th Chinese Control and Decision Conference (2019 CCDC) 5151
 Table3. KI control rule table 3 Simulation and experiment
In order to illustrate the advantages of fuzzy PID control,
we compare it with ordinary digital PID control.
MATLAB Simulink function module is used to simulate
the fuzzy control PID, and verified by the real object.
In the MATLAB/Simulink toolkit, select the mathematical
operation modules and tool modules required by the robot
fuzzy control algorithm and drag them to the simulation
design window. Connect each module according to the
requirements of the fuzzy control algorithm, set the
Table4. KD control rule table parameters of each module, and finally complete the
simulation model of the fuzzy controller, as shown in
Fig.5.

The output obtained by fuzzy inference is a fuzzy set, and Fig. 5 Robot fuzzy controller simulation model
U(t) is a distribution combination of different values of the
output variable domain. The output is generally expressed The conventional PID control simulation model is built, as
as an irregular, segmented membership function and shown in Fig. 6.
cannot be directly used. In the physical control process,
the amount of control applied to the control object must be
an exact value. Thus, the fuzzy output must be sharpened
to a meaningful value. In this paper, the weighted average
method (center of gravity method) is used to clear the
fuzzy output, that is, the elements in the output and the
corresponding membership are combined to obtain the
average value[10]. For the discrete universe, the Fig. 6 Robot conventional PID controller simulation model
expression is as follows: 3.1 MATLAB simulation process and result analysis
σ೙
೔షభ ௨೔ ௭೔
‫ݖ‬଴ ൌ σ೙
(6) When the simulation is set, the sampling time of the
೔షభ ௨೔

Note: where zi denotes the various elements in the output computer is 0.02‫ݏ‬, the robot speed is 0.5݉Ȁ‫ݏ‬, and the
fuzzy set U, ui represents the corresponding membership robot position deviation at the initial moment is 0.1m,
degree, and z0 is the judgment result obtained by applying angle deviation is 0o, the target motion curve is y=0.1, and
the weighted average method to the fuzzy set. the fuzzy PID controller and the conventional PID
controller are respectively used for simulation.
Fig. 4 shows the fuzzy PID control program flow chart.
In PID control, according to the existing kinematics model,
Program entry
the parameters are obtained by trial and error method: kp
Cuurent sampling Caluculate current
=5, ki=1.2, kd=1.
y(k) kpȽkiȽ kdȽ
Fig. 7 is the simulation diagram of the motion trajectory of
e=r(k)-y(k)
the fuzzy controller and the conventional PID controller.
PID operation
ȴe=e(k)-e(k-1)
Fig. 8 shows the simulation diagram of motion deviation
Controlled variable
of fuzzy controller and conventional PID controller.
e˂ȴe fuzzification
Output

Fuzzy turning ȴKPǃ ȴKIǃȴKD Controlled objective

Ambiguity-resolvin Returns

Fig. 4 Fuzzy PID control program flow chart

5152 The 31th Chinese Control and Decision Conference (2019 CCDC)
 direction along the X-axis and to the left in a positive
direction along the Y-axis. Establish the data transmission
between the camera and PC, record the robot's movement,
and save the track information in the form of coordinates.
The position deviation between each trajectory point and
the straight black belt is calculated. The robot driving
direction is positive, left and right deviation is positive,
and right and left deviation is negative.

Fig. 7 Motion trajectory simulation diagram

Fig. 9 Robot experiment scene Fig. 10 Kinect depth camera

The image sampling frequency is 0.1‫ݏ‬, the robot velocity

Fig. 8 Motion deviation simulation diagram V=0.5 ݉Ȁ‫ ݏ‬, the robot position deviation at the initial
moment is set to 0.1݉, and the deviation Angle is 0o. The
It can be seen from the figure that when the initial position robot control system successively writes the conventional
deviation of the robot is the same, the PID controller PID control program and the fuzzy PID control program,
responds in time, and the deviation is rapidly reduced. respectively runs under the two control programs, obtains
However, under the control of the PID controller, the the test data, and draws the robot path tracking position
positional deviation fluctuates 3-4 times repeatedly, and it deviation curve according to the data (Fig.11)
takes about 6.5 seconds, the robot can converge to a
predetermined trajectory. In contrast, under the control of
the fuzzy controller, the positional deviation not only
drops faster, but also experiences only one fluctuation,
which takes about 4 seconds before the robot quickly
converges to the predetermined trajectory.
According to the simulation results of the above two
controllers, it can be clearly seen that in the robot
trajectory tracking control, the fuzzy PID controller has
faster response, smaller overshoot, less fluctuation, stable
Fig. 11 Path tracking position deviation curve
change and stronger system stability than the conventional
PID controller. Table5. Path tracking position deviation data statistics

4.2 verification and analysis of physical experiments

The robot path tracking test scenario is shown in Fig. 9.
The laboratory size is 12*10*4.5݉, and the ground is flat,
which meets the basic requirements of robot operation.
The robot is placed under a depth camera to record track
deviations. Depth camera device used Kinect V2.0, as
shown in Fig. 10, equipped with a full 1080p HD wide At the same initial position deviation of the robot, the PID
Angle camera, and adapt to the low light environment of controller responds in time, and the deviation decreases
active infrared camera, the microphone array has more rapidly. However, under the control of the PID controller,
function of noise reduction and voice commands, the the position deviation fluctuates twice with the maximum
camera can achieve 2 GB of data per second recognition deviation of 0.048݉ and the average deviation of 0.014݉.
ability, so as to more accurately determine the robot's tiny After about 3.5 ݉ ǡ the robot converges on the
action, can detect the range of 0.5-8݉. In the path tracking predetermined trajectory. By contrast, under the control of
test of the robot, a straight black belt laid on the ground is the fuzzy controller, the positional deviation not only
used as the guiding path. The linear black belt, as the decreases faster, but also fluctuates once, with the
X-axis, indicates that the robot moves in a positive maximum deviation of 0.021݉ and the average deviation

The 31th Chinese Control and Decision Conference (2019 CCDC) 5153
of 0.006 ݉ . Within about 1.1m, the robot quickly [2] W. Yizhi, C. Gegong, Motion restricted conditionand singular
configuration for mecanum wheeled omni-directional motion
converges to the predetermined trajectory.
system, Journal of Shanghai Uni-versity: Natural Science, Vol. 15 ,
According to the comparison of the actual test results of No.2 , 181㸫185, 2009.
the above two controllers and the trajectory tracking [3] Z. Yunan, Tian Peng, Wang Shuangshuang, Parametric solid
control of the robot, the fuzzy PID controller exhibits a modeling of omnidirectional wheelbased on Solid Works API,
faster response, less overshoot and fluctuation, and stable Modern ManufacturingEngineering㸪2013( 2) , 59㸫62㸬
change and system than the conventional PID controller. [4] L. Zhongxin, Z. Chunliang, W. Zhanfu, Omnidirectional mobile
platform motion analysis andsimulation, Machinery and
4 CONCLUSION Electronics, 2013 ( 8), 16㸫 19㸬
[5] Z. Liming, W. Zhenyu, W. Xu, Visual AGV two-wheel differential
On the basis of establishing the kinematics model of steering fuzzy control, Journal of Kunming University of Science
omnidirectional mobile robot and introducing the fuzzy and Technology (Science and Technology Edition), Vol.4, 28-32,
control principle, this paper designs the fuzzy PID 2007.
controller of the robot and conducts simulation analysis on [6] W. Xingsong, Mecanum wheel all-round mobile robot technology
the MATLAB platform. Finally, the control effect of fuzzy and its application, Machinery Manufacturing and Automation,
Vol.43, No.3, 1-6, 2014.
PID controller on AGV path tracking is tested. Computer
[7] F. Feng, D. Zhiliang, and Z. Xu, Research on self-organizing fuzzy
simulation and actual test results show that in the AGV controller of AGV automatic navigation vehicle, Microcomputer
trajectory tracking control, the fuzzy PID controller has Information, Vol.24, No.4-1, 78-80, 2008.
faster response, smaller overshoot, less fluctuation, stable [8] W. H. Huang, Fuzzy Control Systems and Applications, Publishing
change and stronger system stability than the conventional House of Electronics Industry, Beijing, China, 2012.
PID controller. [9] H. X. Li, S. K. Tso, Quantitative design and analysis of fuzzy PID
control: A step towards autotuning, International Journal of
REFERENCES Systems Science, Vol.31, No.5, 337-341,2000.
[10] S. Doudou and F. Khaber, Direct adaptive fuzzy control of a class of
[1] W, Shuangshuang㸪Z. Yunan㸪Wang Heyuan, Trajectory tracking MIMO non.affine nonlinear systems, International Journal of
based on fuzzy sliding mode controlfor omnidirectional platform, Systems Science, Vol.43, No.6, 1029-1038, 2012.
Drive and Control㸪Vol.40, No.8 , 63㸫66, 2012.

5154 The 31th Chinese Control and Decision Conference (2019 CCDC)