Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

2019 IEEE International Conference on Automatic Control and Intelligent Systems (I2CACIS 2019), 29 June 2019, Selangor, Malaysia

Fuzzy Logic Control of Goal-Seeking 2-Wheel


Differential Mobile Robot Using Unicycle Approach
Lukuman Abolore Yekinni Ado Dan-Isa
Sheik Bashir School, Tal-Udu Dept. of Electrical Engineering
Kano City, Nigeria Bayero University, Kano, Nigeria
Email: mallamluku@yahoo.com Email: danisa@buk.edu.ng

Abstract — In this work, fuzzy logic-based navigation control area, to improve wheeled mobile robot capabilities. For
for a 2-wheel mobile robot is proposed. An autonomous mobile example, in reference [16] the authors designed a controller
robot must be able to move safely in an environment, to reach to control the robot motion along a predefined path using
its target (or goal). The work began with the mathematical fuzzy logic. A total of 18 fuzzy rules were formulated, nine
model of the robot that involved the kinematic model. After
for each of the two (right and left) wheels of the robot. The
which the Simulink model was developed and then fuzzy logic
controller was designed for goal seeking. The problem with the researchers in [4] designed a new MIMO fuzzy logic
differential drive model is that separate fuzzy rules need to be controller for tracking trajectory and navigation behavior of
developed for both left and right wheels of the robot and the a mobile robot. This seems to be the general pattern of
steering angle cannot be controlled directly from the fuzzy fuzzy control of 2-wheel mobile robots. Rules are generated
rules. However, these limitations have been taken care of in on the basis of sensory inputs. These are then used to
this work. The linear velocity and steering angle of the mobile control each of the two wheels separately [5-7, 17, 18, 19,
robot were directly controlled from a single set of fuzzy rules 20, 21]. Separate fuzzy rules are made for each wheel. No
and this has helped to reduce the time of navigation of the doubt other variants of this scheme exist whereby other
robot. The performance was satisfactory and the results of the
mechanisms are added to effect the fuzzy control. For
simulation showed that the robot could reach its goal during
navigation in shorter time than if differential drive model alone example Akka and Khaber [22] designed an optimal
was used. tracking controller to make a wheeled mobile robot to track
a planned trajectory. The design is based on Dynamic
Keywords: Fuzzy Logic Control, Goal Seeking Controller, Window Approach (DWA), where the weights of the
Unicycle Model, Differential Wheel model, Robot Navigation objective function are dynamically adjusted using fuzzy
logic controller. For the tracking control, a discrete linear
quadratic tracker (LQT) was used by the authors. The work
I. INTRODUCTION in [23] was based on designing type-2 fuzzy self-tuning
PID controller to improve the autonomous driving action of
Mobile robots may be categorized as Automated a mobile robot. Although all cited works have been
Guided Vehicles (AGVs) or Autonomous Mobile Robots successful, the two wheels of the robot are controlled
(AMRs). The AGVs move around with the map of the separately, sometimes even with different set of rules. It
environment in their memories, thus have limited flexibility would seem more intuitive to use a single set of rules to
and applications. The AMRs on the other hand, navigate control the robot's motion in space.
without any prior description of the environment in their This paper is organized as follows: In the second
memories but rely solely on information from sensors and section of this paper, the mobile robot is described.
control algorithm to achieve their tasks. These types are Equations of the kinematics of the robot are given and the
more flexible than AGVs and could adapt to different design of the goal-seeking fuzzy logic controller is
environments. discussed. The performance of the proposed approach is
However, a mobile robot that carries out navigation demonstrated and discussed in section three. The
tasks in an uncertain environment needs some level of conclusion is given in section four.
intelligence to figure out its relationship with the
environment in order to compute the best path between its
current position and the intended target or goal position [1]. II. METHODOLOGY
Therefore, it is very important to incorporate intelligence
A. Mobile Robot Model
into an autonomous mobile robot so as to have a good
control and ability to adapt to changing environment and The mobile robot considered in this paper is a 2-wheel
demands. This intelligence is usually incorporated via non-holonomic robot with a third caster wheel to provide
artificial intelligence methodologies such as neural network stability. The sketch of the robot is shown in Fig.1. It has
[2], fuzzy logic [3, 4, 5, 6, 7] or several other metaheuristic two independent driving wheels arranged parallel to each
methods [8, 9, 10, 11]. Kashyep and Pandey [12] gave a other. Each of the two wheels is driven independently by a
good review of these methodologies as applied in wheeled DC motor. Thus this mechanism is able to drive the robot
robot motion planning. Some researchers even consider the forward and backward as well as steer its heading angle by
mobile robot navigation problem using hybrid methods [3, differentiating the linear velocity of the right wheel (Vr) and
13, 14]. In their review, Radzak et al [15] categorize the linear velocity of the left wheel (Vl). Even though this
control methodologies used for mobile robot control. configuration requires complex control strategy than the
Researchers continue to work, especially in fuzzy logic steer drive configuration, it is still often the first choice for
The authors wish to thank Bayero University, Kano for sponsoring the
most researchers in this area. This could be due to two
trip to present this work during the I2CACIS 2019 (IEEE) Conference at
Shah Alam, Malaysia, 29th June 2019.

978-1-7281-0784-4/19/$31.00 ©2019 IEEE 300

Authorized licensed use limited to: Universidad Militar Nueva Granada. Downloaded on October 31,2020 at 20:59:34 UTC from IEEE Xplore. Restrictions apply.
2019 IEEE International Conference on Automatic Control and Intelligent Systems (I2CACIS 2019), 29 June 2019, Selangor, Malaysia

reasons. Firstly, the two-wheel drive strategy has the However, for intuitive control, the robot’s translational
capability of making a small radius turning, even making a velocity (V) and angular speed (ω) should be the design
turning on a spot. Secondly, most experimental wheeled outputs. With these we can actually feel what the robot is
mobile robots for laboratory use are built with this doing. As shown by LaValle [24], the kinematics of a
configuration. The sketch of the two-wheeled robot is unicycle robot can simulate that of a differential one. This
shown in Fig. 1. fact can be exploited to design the fuzzy logic controller for
a differential wheel robot on the basis of a simple unicycle
model given below.
x = V cos(φ )
y = V sin(φ ) (2)
φ = ω

The robot’s angular velocity ω is simply the time rate of


change of its orientation.
Note that the robot’s velocity is the average of the
velocities Vr and Vl. Also the angular speed of the robot is
related to Vr and Vl and the distance between the two wheels
[24] as follows:
Fig. 1. Two-Wheeled Robot 1
V = (Vr + Vl )
2 (3)
The quantities in Fig.1 are described in Table I below. 1
ω = (Vr − Vl )
L
TABLE I. PARAMETERS OF THE ROBOT MODEL

Quantity Description Solving these two equations simultaneously gives the


Vr Linear velocity of the right wheel required differential wheel velocities as
Vl Linear velocity of the left wheel 2V + ωL
L Distance between the two wheels Vr =
2 (4)
ω Angular velocity of the whole robot 2V − ωL
V Linear velocity of the whole robot Vl =
φ Robot’s orientation with respect to x-axis 2
r Radius of the wheel
The fuzzy controller in this work is designed based on the
unicycle model in (2) because of its simplicity and intuitive
B. Equations Defining the Robot’s Kinematics appeal even though the robot itself is a differential drive
The first stage in the kinematic modeling of a robot is system. The controller produces V and φ as outputs. The
to express constraints on the motion of the individual wheel. required control inputs to the robot, namely Vr and Vl, are
The constraints enforce the concept of no lateral movement, computed from (4), noting that ω = dφ/dt, as given in (2).
that the wheels have pure rolling and non-slippage condition
during the motion in the wheel plane. The control inputs are C. Goal-Seeking Controller Design
the rate at which the right wheel is moving and the rate at
which the left wheel is moving, that means Vr and Vl The task of the robot is to reach a desired destination
respectively. But actually we care about the robot’s position in any environment. The goal seeking behavior is expected
(x, y) on a plane and its orientation/steering angle (φ). These to align the robot’s heading with the direction of the goal
parameters indicate where the robot is and in what direction coordinates. In order to achieve this, linguistic variables and
it is moving. Therefore, a robot model is that which relates values are designed for both the inputs and the outputs of
the controller. The linguistic variables for the goal seeking
Vr and Vl to the state of the robot (x, y, φ). Under the stated
constraints the kinematics equations of the differential drive behavior are the error angle (φe) (the angle between the
robot in the world frame are as follows [16, 24]. robot and the goal) and goal distance, Dg (the distance from
the robot to the goal). These variables serve as inputs to the
r controller. For the outputs of the controller, the linguistic
x = (Vr + Vl ) cos(φ )
2 variables are the robot's steering angle (φ) and velocity (V).
r The linguistic values for the input “Error Angle” are LB,
y = (Vr + Vl ) sin(φ ) (1) LS, N, RS, RB while the linguistic values for the input
2
r “Goal Distance” are VF, F, M, C, VC. Similarly, the
φ = (Vr − Vl ) linguistic values for the output “Steering Angle” are LB, LS,
L NT, RS, RB while for the output “Velocity” are AF, A, M,
D, DF. These are summarized in Table II.
This model gives what is required to map the control
inputs (Vr and Vl) into the state of the robot (x, y, φ).

978-1-7281-0784-4/19/$31.00 ©2019 IEEE 301

Authorized licensed use limited to: Universidad Militar Nueva Granada. Downloaded on October 31,2020 at 20:59:34 UTC from IEEE Xplore. Restrictions apply.
2019 IEEE International Conference on Automatic Control and Intelligent Systems (I2CACIS 2019), 29 June 2019, Selangor, Malaysia

TABLE II. SUMMARY OF LINGUISTIC VALUES fuzzify the variables, see Fig. 2. The parameters defining the
Input Variables triangular MFs for both the inputs and outputs are listed in
Error Angle, φe (rad.) Goal Distance, Dg (m) Table III. Mamdani fuzzy inference method was used to
Left Big (LB) Very Far (VF) infer fuzzy outputs from the rules listed in Table IV. The
Left Small (LS) Far (F) crisp controller outputs were produced from the inferred
Normal (N) Medium (M)
Right Small (RS) Close (C)
fuzzy outputs using centre of gravity (also called centroid or
Right Big (RB) Very Close (VC) centre of mass) defuzzification method.
Output Variables
Steering Angle, φ (˚) Robot’s Velocity, V (m/s)
Left Big (LB) Accelerate Fast (AF)
Left Small (LS) Accelerate (A)
No Turn (NT) Maintain (M)
Right Small (RS) Decelerate (D)
Right Big (RB) Decelerate Fast (DF)

Fuzzy logic control system design consists of three


main stages: fuzzification, inference, defuzzification. In this Fig. 2. Triangular Membership Function
work triangular membership functions (MFs) were used to

TABLE III. TRIANGULAR MEMBERSHIP FUNCTIONS PARAMETERS


Error Angle (φe) LB LS N RS RB
(radians) [-3 -2 -1] [-2 -1 0] [-1 0 1] [0 1 2] [1 2 3]

Goal Distance Dg VC C M F VF
(meters) [0 0 2] [0 2 4] [2 4 6] [4 6 8] [6 8 10]

Steering Angle (φ) LB LS NT RS RB


(degrees) [-3 -2 -1] [-2 -1 0] [-1 0 1] [0 1 2] [1 2 3]

Robot’s Velocity(V) AF A M D DF
(m/s) [2 2.5 3] [1.5 2 2.5] [1 1.5 2] [0.5 1 1.5] [0 0.5 1]

Fig. 3. Simulink Block Model for Goal Seeking Controller

TABLE IV. GOAL-SEEKING CONTROLLER RULES III. RESULTS AND DISCUSSION


LB LS N RS RB The Simulink block model of the differential drive
mobile robot used in this work for goal seeking behavior is
VF φ LB LS NT RS RB shown in Fig. 3 and the simulation is carried out in Matlab
environment. When the simulation was run, the mobile
V AF AF AF AF AF robot moved along straight line and was able to reach the
F φ LB LS NT RS RB goal point.
Many researchers use fuzzy logic rules to control
V AF AF AF AF AF differential drive robot, where the left and right wheels are
controlled separately. This is called the differential wheel
M φ LB LS NT RS RB approach. In this work a new approach was proposed,
V M M M A A whereby a unicycle wheel dynamics is used to simulate
differential wheel dynamics. As a result, the linear velocity
C φ LB LS NT RS RB and the orientation of the robot can be controlled directly
V D D D D D using a single set of fuzzy rules. For the purpose of
comparison simulations were carried out using the
VC φ LB LS NT RS RB differential wheel approach as proposed by Mac et al [4]
V DF DF DF DF DF and the unicycle-based approach proposed in this work. The

978-1-7281-0784-4/19/$31.00 ©2019 IEEE 302

Authorized licensed use limited to: Universidad Militar Nueva Granada. Downloaded on October 31,2020 at 20:59:34 UTC from IEEE Xplore. Restrictions apply.
2019 IEEE International Conference on Automatic Control and Intelligent Systems (I2CACIS 2019), 29 June 2019, Selangor, Malaysia

paths taken by the robot to reach its target are shown in Fig. Velocity for Goal Point (8,6)
4 through to Fig. 6. 2
Unicycle Model
1.8 Differential Model
Goal Point (9,7)
8
Unicycle Model 1.6
Differential Model
7 1.4

Velocity (m/s)
1.2
6
1

5 0.8
Y (m)

0.6
4
0.4
3
0.2

2 0
0 2 4 6 8 10 12 14 16
Time (s)
1
Fig. 5b. Velocities of the two approaches when goal point = (8, 6)
0
0 2 4 6 8 10
X (m)
Fig. 4a. Outputs of the two approaches when goal point = (9,7) Goal Point (7,6)
6
Unicycle Model
Differential Model
Velocity for Goal Point (9,7) 5
2
Unicycle Model
1.8 Differential Model
4
1.6
Y (m )

1.4 3
Velocity (m/s)

1.2

2
1

0.8
1
0.6

0.4
0
0 1 2 3 4 5 6 7
0.2
X (m)
0 Fig. 6a. Outputs of the two approaches when goal point = (7,6)
0 5 10 15 20
Time (s)
Velocity for Goal Point (7,6)
Fig. 4b. Velocities of the two approaches when goal point = (9,7) 2
Unicycle Model
1.8 Differential Model

Goal Point (8,6) 1.6


7
Unicycle Model
Differential Model 1.4
6
V e lo c ity (m /s )

1.2

5 1

4
0.8
Y (m )

0.6
3
0.4

2 0.2

1 0
0 2 4 6 8 10 12 14
Time (s)
0
0 1 2 3 4 5 6 7 8 9 Fig. 6b. Velocities of the two approaches when goal point = (7,6)
X (m)
Fig. 5a. Outputs of the two approaches when goal point = (8, 6)

978-1-7281-0784-4/19/$31.00 ©2019 IEEE 303

Authorized licensed use limited to: Universidad Militar Nueva Granada. Downloaded on October 31,2020 at 20:59:34 UTC from IEEE Xplore. Restrictions apply.
2019 IEEE International Conference on Automatic Control and Intelligent Systems (I2CACIS 2019), 29 June 2019, Selangor, Malaysia

Fig. 4a shows the path followed by the robot using the [3] N. Paykan, S. H. Abbasi, and F. Shabaninia, “Design of MIMO
Mamdani Fuzzy Logic Controller for Wall Following Mobile
differential wheel and the unicycle approaches. In each case,
Robot,” Advance in Intelligent Systems and Computing, vol. 195, pp
the robot was able to reach the goal point, though at 155-164, 2013.
different times. From the graphs, it can be seen that, for both [4] Thoa T. Mac, Cosmin Copot, Robin De Keyser, Trung D. Trana and
the differential wheel and unicycle approaches, the robot did Thich Vu, “MIMO Fuzzy Control for Autonomous Mobile Robot,”
Journal of Automation and Control Engineering, Vol.4 (2016).
not move toward the goal located at coordinate (9,7) directly
[5] Xi Li, Byung-Jae Choi, “Obstacle Avoidance of Mobile Robot by Fuzzy
from the beginning but first moved along the x-axis for few Logic System,” Proceeding of the 7th Int. Conference on Information
seconds before moving toward the goal point. It took the Security and Assurance, ASTL vol. 21, pp. 244-246, 2013.
differential wheel a little more time before it could align its [6] Khaoula Maatoug, Malek Njah, Mohamed Jallouli, “Free Navigation
and Obstacle Avoidance Based on Fuzzy Controller,” Proc. of the
heading toward this goal which resulted in delay in its
2nd World Congress on Computer Application and International
navigation. On the other hand, from the velocity-time graph Systems, 2015.
in Fig. 4b, the unicycle model started with a high velocity. [7] Hajer Omrane, Mohamed Slim Masmoudi and Mohamed Masmoudi,
This speed was maintained for a short time after which the “Fuzzy Logic Based Controlled for Autonomous Mobile Robot
Navigation,” Computational Intelligence and Neuroscience, 2016.
acceleration decreases till the robot reached its goal in 9s. [8] Andi Adriansyah, Yudhi Gunardi, Badaruddin, Eko Ihsanto, “Goal-
But for the differential wheel model, the robot started with Seeking Behavior–Based Mobile Robot Using Particle Swam Fuzzy
low velocity of about 0.5m/s, maintained the speed for about Controller,” TELKOMNIKA, vol.13, No.2, 2015.
3s before accelerating uniformly, building up its velocity [9] Zimit A.Y., Yap H.J., Hamza M.F., Siradjuddin I., Hendrik B.,
Herawan T. (2018) “Modelling and Experimental Analysis Two-
slowly to its maximum value of almost 1m/s. After that the
Wheeled Self Balance Robot Using PID Controller,” In: Gervasi O.
speed gradually decreased as the robot reached the goal et al. (eds) Computational Science and Its Applications – ICCSA
point in 18s. As shown in Fig. 5 and Fig. 6, the 2018. Lecture Notes in Computer Science, vol 10961. Springer,
performances were similar even when the goal coordinate is pp.683 – 698.
[10] B.K. Patle, Anish Pandey, Anish Pandey, Jagadeesh Anne and Dayal
changed. The results are summarized in Table V below.
R. Parhi, “Path planning in uncertain environment by using firefly
algorithm,” Defence Technology (2018), DOI:
10.1016/j.dt.2018.06.004
TABLE V. COMPARISON OF THE TWO APPROACHES [11] B.K. Patle, Dayal R. Parhi, Anish Pandey, Jagadeesh Anne and Sunil
Kumar Kashyap, “Matrix-Binary Codes based Genetic Algorithm
Goal Point Co- Time to Reach Goal Point for path planning of mobile robot,” Computers and Electrical
ordinate (x, y) in Journal, December 2017.
meters Differential Unicycle [12] Abhishek K. Kashyep and Anish Pandey, (2018) “Different Nature-
Drive Model Drive Model Inspired Techniques Applied for Motion Planning of Wheeled
(9, 7) 18s 9s Robot,” Critical Review. Int J Adv Robot Automation, 3(2). pp1-10.
(8, 6) 16s 8s [13] Panus Natharith, “Fuzzy Logic based Control of Mobile Robot
Navigation: A Case study on Robot Roomba Platform,” Scientific
(7, 6) 14s 7.9s Research and Essays, vol.8(2), 2013.
[14] Maïssa Boujelben, Chokri Rekik and Nabil Derbel (2016), “A hybrid
fuzzy-sliding mode controller for a mobile robot,” Int. J. Modelling,
Identification and Control, Vol. 25, No. 3, 2016. pp155-164.
IV. CONCLUSION [15] M S A Radzak , M A H Ali, S Sha’amri and A R Azwan (2018), “An
We have demonstrated in this work that unicycle overview on real-time control schemes for wheeled mobile robot,”
kinematic model can be used to represent the kinematic IOP Conf. Series: Materials Science and Engineering 342 (2018)
012059 DOI: 10.1088/1757-899X/342/1/012059
model of a 2-wheel differential drive robot. Fuzzy logic- [16] Vamsi Mohan Peri and Dan Simon, “Fuzzy Logic Controller for an
based controller was designed on the basis of the unicycle Autonomous Robot,” North American Fuzzy Information Processing
model and was made to successfully control the differential Society Conf., 2005.
drive robot. We were able to control the linear velocity of [17] Razif Rashid, I. Elamvazuthi, Mumtaj Begam, M. Arrofiq, “Fuzzy-
based Navigation and Control of a Non-Holonomic Mobile Robot,”
the robot directly using a single set of fuzzy rules without Journal of Computing Vol. 2, Issue 3, March 2010.
designing separate rules for the right and left wheels of the [18] C. G. Rusu, I. T. Birou, “Obstacle Avoidance Fuzzy System for
robot as was done in majority of the cited literature. This has Mobile Robot with IR Sensors,” 10th Int. Conf. on Development and
the result of reducing the time of navigation of the mobile Application Systems, Suceava, Romania, May 27-29, 2010.
[19] Gyula Mester, “Sensor-based Control of Autonomous Wheeled Mobile
robot. In addition to the speed control, the robot is also Robot,” Ipsi Journal, TIR, volume 6, Number 2, pp.29-34, 2010.
made to look as if it were steer-driven. This is achieved by [20] Birol Kocaturk, “Motion Control of Wheeled Mobile Robots,”
directly controlling the steering angle of the robot from the Interdisciplinary Description of Complex Systems 13(1), 41-47,
fuzzy rules rather than by controlling the angular velocities 2015.
[21] Karim Benbouadallah, Zhu Qi-dan, “Design of a Fuzzy Logic
of the wheels as is the case in differential wheel drive. Controller for a Mobile Robot Tracking a Moving Target,” 2nd
Comparing the two approaches, it was observed that the International Conference on Computer Science and Network
unicycle approach is more efficient than the differential Technology, 2012.
wheel approach in terms of reaching the goal quicker. [22] Khalid Akka, Farid Khaber, “Optimal Tracking Control of a
Trajectory planned via Fuzzy Reactive Approach for an
Autonomous Mobile Robot,” International Journal of Advance
Robotic System, 2018
REFERENCES [23] Park SH, Kim KW, Choi WH, Jie MS, Kim YI, “The Autonomous
[1] Chih-Hui Chiu and Ya-Fu Peng (2017), “Position and Angle Control for Performance Improvement of Mobile Robot using Type-2 Fuzzy
a Two-wheel Robot,” International Journal of Control, Automation Self-Tuning PID Controller,” Advanced Science and Technology
and Systems, October 2017, Volume 15, Issue 5, pp 2343–2354. Letters, Vol.138 (ISI 2016), pp.182-187.
[2] Bing Z, Meschede C, Röhrbein F, Huang K and Knoll AC (2018), “A [24] LaValle, SM. (2006), Planning Algorithms, Cambridge University
Survey of Robotics Control Based on Learning-Inspired Spiking Press, 2006. Chap.13.
Neural Networks,” Front. Neurorobot. 12:35. DOI:
10.3389/fnbot.2018.00035

978-1-7281-0784-4/19/$31.00 ©2019 IEEE 304

Authorized licensed use limited to: Universidad Militar Nueva Granada. Downloaded on October 31,2020 at 20:59:34 UTC from IEEE Xplore. Restrictions apply.

You might also like