Home Search Collections Journals About Contact us My IOPscience

Differential-Drive Mobile Robot Control Design based-on Linear Feedback Control Law

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2017 IOP Conf. Ser.: Mater. Sci. Eng. 190 012001

(http://iopscience.iop.org/1757-899X/190/1/012001)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 72.198.203.96
This content was downloaded on 20/04/2017 at 01:54

Please note that terms and conditions apply.

International Conference
IAES International on Recent
Conference Trends inEngineering,
on Electrical Physics 2016Computer
(ICRTP2016)
Science and Informatics IOP Publishing
Journal of Physics:
IOP Conf. Conference
Series: Materials Seriesand
Science 755 (2016) 011001
Engineering doi:10.1088/1742-6596/755/1/011001
190 (2017) 012001 doi:10.1088/1757-899X/190/1/012001
1234567890

Differential-Drive Mobile Robot Control Design based-on

Linear Feedback Control Law

Siti Nurmaini1, Kemala Dewi2, Bambang Tutuko3

1,2,3
Robotic and Control Research Lab. Computer Engineering Department,
Universitas Sriwijaya
1
[email protected], 2 [email protected], 3 [email protected],

Abstract. This paper deals with the problem of how to control differential driven mobile robot
with simple control law. When mobile robot moves from one position to another to achieve a
position destination, it always produce some errors. Therefore, a mobile robot requires a
certain control law to drive the robot’s movement to the position destination with a smallest
possible error. In this paper, in order to reduce position error, a linear feedback control is
proposed with pole placement approach to regulate the polynoms desired. The presented work
leads to an improved understanding of differential-drive mobile robot (DDMR)-based
kinematics equation, which will assist to design of suitable controllers for DDMR movement.
The result show by using the linier feedback control method with pole placement approach the
position error is reduced and fast convergence is achieved.

1. Introduction
According to various studies that have been conducted, mobile robot control system can be classified
into three categories. The first category is namely the sensor-based control-based approach. Such
control system is emphasized on how to model the motion of a robot in a dynamic environment [1].
The control process to produce estimation and predictions of the mobile robot movement is based on
information from sensor detection [2]. The intelligent control scheme is an approach that is most
widely used [3][4][5]. However, its results produce sub-optimal response, because the motion is only
around the trajectory detection [5]. The second category is the approach of decomposed execution
process using a path planning [6][7]. The control system regulates the movement of the mobile robot
through the planned path, therefore it can move according to the target that has been set up. The
environmental mapping is created for producing collision-free path. Such scheme of control is based
on minimal distance, energy and time. The third category presenting the optimization algorithm is
developed for controlling the mobile robot with accurate trajectory. The controller design is based on
mathematical model of mobile robot. The approach is for tracking the mobile robot errors between
reference and actual trajectory [8][9].
However, the whole categories of the control system only operate when the condition of linear
velocity is not zero. Therefore, a mobile robot is difficult to control, especially in the case of following
the reference of trajectory in a short time with minimal errors. Nonlinear control approaches have been
employed to solve this problem [10][11][12][13]. Although the regulation problem is solved to track
the mobile robot move to desired trajectory, but it found to yield slow asymptotic convergence [13]. In
order to obtain faster convergence, an alternative approach must be proposed. This paper proposes the
linier feedback control law to overcome such limitation with pole placement approach to regulate the

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
IAES International Conference on Electrical Engineering, Computer Science and Informatics IOP Publishing
IOP Conf. Series: Materials Science and Engineering 190 (2017) 012001 doi:10.1088/1757-899X/190/1/012001
1234567890

polynom desired. By using the scheme, the controller performance can be regulated and fast
asymptotic convergence is achieved.

2. Differential Drive Mobile Robot Kinematic Model

The differential-drive mobile robot (DDMR) is a class of computer-controlled vehicles whose motion
can be described or transformed into the following model with constrained movement in a plane.
Figure 1 describes the DDMR on 2D-cartesian plane, it means the mobile robot move only in the x y
axis, while the contour and elevation of the z-axis is ignored [13].

Figure 1. DDMR on 2D cartesian plane

From Figure 1, the representation of 𝑥̇ 𝑄 , 𝑦̇ 𝑄 are x, y coordinate relation to center of the robot Q, 𝑣𝑄
is a linier velocity, 𝑣𝐿 are linier velocity of left wheels and 𝑣𝑅 linier velocity of right wheel, 𝑟 is a
radius of wheel, 2𝑏 is a distance from right to left wheel, 𝑄 is a center point of the robot, 𝐺 is a center
of gravitation and 𝑏 is a distance between 2 wheels with G. By ignoring the analysis of the castor-free,
the configuration of the mobile robot can be described into three general variables 𝑞(𝑡) to describe the
position of the robot and the control input 𝑢(𝑡). It can be defined in matrix:
𝑇
𝑞(𝑡) = (𝑥(𝑡), 𝑦(𝑡), 𝜃(𝑡))𝑇 , and 𝑢(𝑡) = (𝜃̇𝑅 (𝑡), 𝜃̇𝐿 (𝑡)) (1)

The angular velocity equations for the right wheel and left wheel such as,

1 1
𝜃̇𝑅 (𝑡) = . 𝑣𝑅 (𝑡) ; 𝜃̇𝐿 (𝑡) = . 𝑣𝐿 (𝑡) (2)
2𝜋𝑟 2𝜋𝑟
𝑣𝑅 (𝑡) = 𝑣(𝑡) + 𝑏𝜃̇ (𝑡) ; 𝑣𝐿 (𝑡) = 𝑣(𝑡) − 𝑏𝜃̇ (𝑡) (3)

From equations (2) and (3) above, a linear velocity equation is 𝑣𝑄 (𝑡) and an angular velocity
equation is ̇𝜃̇ (𝑡) as follow,
1 𝑣𝑅 (𝑡)−𝑣𝐿(𝑡)
𝑣𝑄 (𝑡) = 2 (𝑣𝑅 (𝑡) + 𝑣𝐿 (𝑡)); 𝜃̇ (𝑡) = (4)
2𝑏

However, the DDMR has a limitation in terms of movement, due to the non-holonomik constraint.
The constrained are obtained by two main assumptions. The first assumption no lateral slip motion.
It’s mean the robot can move only in a curved motion (forward and backward) but not sideward. The
velocity equation at the midpoint of the both wheels (𝐺) is obtained by using Figure 2(b), the equation
is described as follows,

𝑣(𝑡) = 𝑥̇ (𝑡) cos 𝜃(𝑡) + 𝑦̇(𝑡) sin 𝜃(𝑡) (5)

2
IAES International Conference on Electrical Engineering, Computer Science and Informatics IOP Publishing
IOP Conf. Series: Materials Science and Engineering 190 (2017) 012001 doi:10.1088/1757-899X/190/1/012001
1234567890

In the robot frame, this condition means the velocity of the wheel in the centre point 𝐺 is zero
along the lateral axis and it’s only move along 𝑥 axis and 𝑦 axis [7]. Therefore, the velocity equation
gives,

𝑥̇ (𝑡) sin 𝜃(𝑡) − 𝑦̇(𝑡) cos 𝜃(𝑡) = 0 (6)

Figure 2. No lateral slip

The second assumption is the pure rolling constraint, due to the wheels do not slip on the floor and
has limitations of pure rolling (Figure 3). The velocities of the contact points in the robot frame are
related to the wheel velocities for the right wheel and left wheel gives,

𝑣𝑅 (𝑡) = 𝑟𝜃̇𝑅 (𝑡) and 𝑣𝐿 (𝑡) = 𝑟𝜃̇𝐿 (𝑡) (7)

Figure 3. Pure rolling motion constraint

By looking at the starting position of the wheel to the Cartesian field 𝑥𝑦 can be obtained the
equation of position wheeled robot, 𝑥(𝑡) = 𝑣(𝑡) sin 𝜃(𝑡), 𝑦(𝑡) = 𝑣(𝑡) cos 𝜃(𝑡), and 𝜃(𝑡) = 𝑣𝜃 (𝑡). Thus,
the transformation matrix equations from the initial position can be defined as,
𝑟 𝑟
cos 𝜃 (𝑡) cos 𝜃(𝑡)
2 2
𝑟 𝑟
𝑇𝑁𝐻 (𝑞 ) = sin 𝜃 (𝑡) sin 𝜃 (𝑡) (8)
2 2
𝑟 𝑟
[ − 2𝑏 ]
2𝑏

Furthermore, the angular velocity 𝑢(𝑡) as input to the robot have three parameters such as, 𝑢1 (𝑡) =
̇𝜃𝑅 (𝑡), 𝑢2 (𝑡) = 𝜃̇𝐿 (𝑡), and 𝜃̇ (𝑡) = 𝜔(𝑡). From equation (8) and three parameters of input equation, the
model of DDMR kinematic equation is, 𝑞̇ (𝑡) = 𝑇𝑁𝐻 (𝑞 ) 𝑢(𝑡) or,

𝑥̇ (𝑡) cos 𝜃(𝑡) 0 𝑣(𝑡)

[𝑦̇(𝑡)] = [ sin 𝜃(𝑡) 0] [ ] (9)
𝜔(𝑡)
𝜃̇ (𝑡) 0 1

3
IAES International Conference on Electrical Engineering, Computer Science and Informatics IOP Publishing
IOP Conf. Series: Materials Science and Engineering 190 (2017) 012001 doi:10.1088/1757-899X/190/1/012001
1234567890

3. The Error Model of DDMR Movement

The error model of DDMR is obtained by reducing the reference position 𝑞𝑟 (𝑡) = (𝑥𝑟 (𝑡), 𝑦𝑟 (𝑡), 𝜃𝑟 (𝑡))𝑇
with the actual position of the robot 𝑞𝑐 (𝑡) = (𝑥𝑐 (𝑡), 𝑦𝑐 (𝑡), 𝜃𝑐 (𝑡))𝑇 . The visualisation of DDMR
movement is presented in Figure 4. Thus, the error of mobile robot movement is represented by
𝑞𝑒 (𝑡) = (𝑒1 (𝑡), 𝑒2 (𝑡), 𝑒3 (𝑡))𝑇 . 𝑞𝑒 (𝑡 ) is the transformation result between the reference position 𝑞𝑟 (𝑡) in
a local coordinate system with the previous position 𝑞𝑐 (𝑡) and 𝑋 axis which is the direction toward the
robot is 𝜃𝑐 (𝑡) or 𝑞𝑒 (𝑡) = 𝑇𝑒 (𝑞𝑟 (𝑡) − 𝑞𝑐 (𝑡)). It can be realized in equation (10) as follows:

𝑒1 (𝑡) cos 𝜃𝑐 (𝑡) sin 𝜃𝑐 (𝑡) 0 𝑥𝑟 (𝑡) − 𝑥𝑐 (𝑡)

[𝑒2 (𝑡)] = [−𝑠𝑖𝑛 𝜃𝑐 (𝑡) cos 𝜃𝑐 (𝑡) 0] [ 𝑦𝑟 (𝑡) − 𝑦𝑐 (𝑡) ] (10)
𝑒3 (𝑡) 0 0 1 𝜃𝑟 (𝑡) − 𝜃𝑐 (𝑡)

By lowering the equation (10) and involving the equations that exist in the equation of non-
holonomic DDMR constraints, the error equation is obtained as equation (11) below:

𝑒̇1 (𝑡) cos 𝑒3 (𝑡) 0 −1 𝑒2 (𝑡)

𝑣 (𝑡) 𝑣 (𝑡)
[𝑒̇2 (𝑡)] = [ sin 𝑒3 (𝑡) 0 ] [ 𝑟 ] + [0 −𝑒1 (𝑡)] [ 𝑐 ] (11)
𝜔𝑟 (𝑡) 𝜔𝑐 (𝑡)
𝑒̇3 (𝑡) 0 0 1 0 −1

Figure 4. Transformation of error movement

From equation (11), 𝑣𝑟 (𝑡) as a reference linear velocity, 𝜔𝑟 (𝑡) as a reference angular velocity,
𝑣𝑐 (𝑡) as an actual linear velocity, and 𝜔𝑐 (𝑡) as an actual angular velocity. In this model the combination
between feed forward control and linier feedback control based on pole placement approach is
proposed, therefore a new input is created below,

𝑣𝑐 (𝑡) = 𝑣𝑓𝑓 (𝑡) − 𝑣1 (𝑡)

𝑣𝑐 (𝑡) = 𝑣𝑟 (𝑡) cos 𝑒3 (𝑡) − 𝑣1 (𝑡) (12)
𝜔𝑐 (𝑡) = 𝜔𝑓𝑓 (𝑡) − 𝑣2 (𝑡)
𝜔𝑐 (𝑡) = 𝜔𝑟 (𝑡) − 𝑣2 (𝑡) (13)

By substituting equation (12) and (13) into equation (11) nonlinear equation of error control
velocity is obtained,

𝑒̇1 (𝑡) 0 𝜔𝑐 (𝑡) 0 𝑒1 (𝑡) 0 1 0 𝑣 (𝑡 )

[𝑒̇2 (𝑡)] = [−𝜔𝑐 (𝑡) 0 0] [𝑒2 (𝑡)] + [sin 𝑒3 (𝑡)] 𝑣𝑟 (𝑡) + [0 0] [ 1 ] (14)
𝑣 (𝑡 )
𝑒̇3 (𝑡) 0 0 0 𝑒3 (𝑡) 0 0 1 2

By using the linearization approach around its operating point (OP : 𝑒1 = 𝑒2 = 𝑒3 = 0, 𝑣1 = 𝑣2 = 0).
Thus, the linear model of error in equation (14) become,

4
IAES International Conference on Electrical Engineering, Computer Science and Informatics IOP Publishing
IOP Conf. Series: Materials Science and Engineering 190 (2017) 012001 doi:10.1088/1757-899X/190/1/012001
1234567890

0 𝜔𝑟 (𝑡) 0 1 0
∆𝑒̇ (𝑡) = [−𝜔𝑟 (𝑡) 0 𝑣𝑟 (𝑡)] . ∆𝑒(𝑡) + [0 0] . ∆𝑣(𝑡) (15)
0 0 0 0 1

4. Control System Design

In this research, the controller design has three states, namely 𝑥(𝑡), 𝑦(𝑡), and 𝜃(𝑡) and two inputs,
namely 𝑣(𝑡), and 𝜔(𝑡). The general equation linear state space controller. 𝑣(𝑡) = 𝐾. 𝑒(𝑡), therefore the
feedback control law becomes

𝑒1 (𝑡)
𝑣1 (𝑡) −𝑘 0 0
[ ]=[ 1 ] [𝑒2 (𝑡)] (16)
𝑣2 (𝑡) 0 −𝑠𝑖𝑔𝑛 (𝑣𝑟 (𝑡))𝑘2 −𝑘3
𝑒3 (𝑡)

From equation (16) a linear velocity of robot 𝑣(𝑡) replaced by 𝑣1 (𝑡), an angular velocity of robot
𝜔(𝑡) replaced by 𝑣2 (𝑡), some values of 𝑘1 , 𝑘2 , and 𝑘3 are a gain in the controller. The diagram block
of the control system design can be seen in Figure 4. Moreover, the value of k control law
corresponding will be determined. This problem is solved by optimizing some function value. In this
research, the determination of the k value is determined by comparing the characteristics of the real
polynoms and desired polynoms of control system characteristics [5]. These polynomials take the
following form as follow,
2
(𝑠 + 2𝜁𝜔𝑛 )(𝑠 2 + 2𝜁𝜔𝑛 𝑠 + 𝜔𝑛 2 ) and 𝑠 3 + (4𝜁𝜔𝑛 )𝑠 2 + (4 𝜁2 𝜔𝑛 2 + 𝜔𝑛 )𝑠 + 2𝜁𝜔𝑛 3 (17)

Where, two parameters are selected such as, the damping coefficient 𝜁 ∈ (0,1) and the natural
frequency 𝜔𝑛 > 0. At the additional pole 𝑠 = −2𝜁𝜔𝑛 which is useful to increases the rise time and to
reduce the overshoot system. The polynomial characteristic of the closed loop control law in equation
(15) with input to the state space controller in equation (16) are the two important parameters design.
All equations from (8) to (16) can be visualized as a block diagram with linier feedback control law as
shown in Figure 5.

Figure 5. Block diagram control law of DDMR system

5. Simulation Results
The purpose of this research is to find the parameters of error value including x(t), y(t) and θ(t) based
on polynomials approach to produce good control performance such as, rise time, and steady state
error parameters. Some simulations are conducted to show the DDMR performance in terms of error
and delta error values by using the proposed control law. The starting position of the robot at
coordinates is set at (0; 0; 0.78) and initial linear velocity is set at 0,01 m/s and angular velocity is set
at 0,02 rad/s. Therefore, it can be calculated to obtain the value of the average error and time to steady

5
IAES International Conference on Electrical Engineering, Computer Science and Informatics IOP Publishing
IOP Conf. Series: Materials Science and Engineering 190 (2017) 012001 doi:10.1088/1757-899X/190/1/012001
1234567890

state. Initial performance of controller are the damping factor and natural frequency. The damping
factor is set variable from 0,2 < 𝜁 < 0.8 and natural frequency are set at 𝜔𝑛 = 5 , 𝜔𝑛 = 2, and 𝜔𝑛 = 1
respectivelly. The error position of DDMR based on selected parameter as shown in Figure 6.

Figure 6. The DDMR error position control

The summary all experiments for DDMR movement in simulation process is described in Table 1
and Table 2.

Table 1. (a) Comparison data of average error value with 0,2 < 𝜁 < 0.8 and 𝜔𝑛 is fixed
(b) Comparison data of the average interval time with 0,2 < 𝜁 < 0.8 𝜔𝑛 is fixed
(a) (b)

In table 1(a) and 1(b), it can be seen the value of the error and the smallest travel time obtained by
setting the value of ζ = 0.6. In Table 2, it compared with the test data in value 𝜁 = 0.6, 𝜔𝑛 = 2, and
𝑔 = 1 before, the delta error value and the smallest interval time are obtained by assigning a value
𝜔𝑛 = 2.

Table 2. Comparison data of the average error value and time interval with
𝜔𝑛 = 5 , 𝜔𝑛 = 2, and 𝜔𝑛 = 1 and 𝜁 is fixed

6. Conclusion
From the experimental results, it is concluded that the method of linearization error with feed-forward
and feedback control law with poles placement approach produces small value of the average error
6
IAES International Conference on Electrical Engineering, Computer Science and Informatics IOP Publishing
IOP Conf. Series: Materials Science and Engineering 190 (2017) 012001 doi:10.1088/1757-899X/190/1/012001
1234567890

position, and it achieved in a short time to steady state condition. From Seven values of ζ and two
values of 𝜔𝑛 , are selected 𝜁 = 0.6, 𝜔𝑛 = 2, and 𝑔 = 1, due to by using the values it can be obtained
the average value of delta error is 4x10-4 m on x position; the average value of delta error is 10-5 m on
y position and the average value of delta error of direction side of the robot is -4x10-4 rad with an
average time interval is 2.5x10-2 s. By regulating performance of controller based on pole placement
approach, the small error, faster time to steady state and small overshoot is achieved.

Acknowledgments
The Authors thank The Ministry of Technology Research and Higher Education (Kemenristek-Dikti),
Indonesia and Universitas Sriwijaya for their financial support in Excellence University Research
Project or Penelitian Unggulan Perguruan Tinggi (PUPT)

References
[1] Fiorini, P., & Shiller, Z. (1998). Motion planning in dynamic environments using velocity
obstacles. The International Journal of Robotics Research, 17(7), 760-772.
[2] Elfes, A. (1989). Using occupancy grids for mobile robot perception and navigation. Computer,
22(6), 46-57.
[3] Fierro, R., & Lewis, F. L. (1998). Control of a nonholonomic mobile robot using neural
networks. IEEE transactions on neural networks, 9(4), 589-600.
[4] Hu, H., & Brady, M. (1996). A parallel processing architecture for sensor-based control of
intelligent mobile robots. Robotics and autonomous systems, 17(4), 235-257.
[5] Ge, Shuzhi S., and Yun J. Cui. "Dynamic motion planning for mobile robots using potential
field method." Autonomous Robots 13.3 (2002): 207-222.
[6] Salomon, B., Garber, M., Lin, M. C., & Manocha, D. (2003, April). Interactive navigation in
complex environments using path planning. In Proceedings of the 2003 symposium on
Interactive 3D graphics (pp. 41-50). ACM.
[7] Yang, J. M., & Kim, J. H. (1999). Sliding mode control for trajectory tracking of nonholonomic
wheeled mobile robots. IEEE Transactions on Robotics and Automation, 15(3), 578-587.
[8] Hoffmann, G. M., Tomlin, C. J., Montemerlo, M., & Thrun, S. (2007, July). Autonomous
automobile trajectory tracking for off-road driving: Controller design, experimental
validation and racing. In American Control Conference (pp. 2296-2301).
[9] Lee, T. C., Song, K. T., Lee, C. H., & Teng, C. C. (2001). Tracking control of unicycle-modeled
mobile robots using a saturation feedback controller. IEEE transactions on control systems
technology, 9(2), 305-318.
[10] Fierro, R., & Lewis, F. L. (1995, December). Control of a nonholonomic mobile robot:
backstepping kinematics into dynamics. In Decision and Control, 1995., Proceedings of the
34th IEEE Conference on (Vol. 4, pp. 3805-3810). IEEE.
[11] Yang, H., Fan, X., Shi, P., & Hua, C. (2016). Nonlinear Control for Tracking and Obstacle
Avoidance of a Wheeled Mobile Robot With Nonholonomic Constraint. IEEE Transactions
on Control Systems Technology, 24(2), 741-746.
[12] Huang, J., Wen, C., Wang, W., & Jiang, Z. P. (2014). Adaptive output feedback tracking control
of a nonholonomic mobile robot. Automatica, 50(3), 821-831.
[13] Dhaouadi, Rached, and Ahmad Abu Hatab. "Dynamic modelling of differential-drive mobile
robots using lagrange and newton-euler methodologies: A unified framework." Advances in
Robotics & Automation 2013.