Hindawi

Mathematical Problems in Engineering

Volume 2021, Article ID 7703165, 10 pages
[Link]

Research Article
Takagi–Sugeno Fuzzy Controller and Sliding Mode Controller for
a Nonholonomic Mobile Robot

Hafedh Abid
Laboratory of Sciences and Techniques of Automatic Control and Computer Engineering (Lab-STA) Sfax,
National School of Engineering of Sfax, University of Sfax, Sfax, Tunisia

Correspondence should be addressed to Hafedh Abid; abidhafedh@[Link]

Received 16 April 2021; Revised 18 May 2021; Accepted 8 June 2021; Published 18 June 2021

Academic Editor: Zain Anwar Ali

Copyright © 2021 Hafedh Abid. This is an open access article distributed under the Creative Commons Attribution License, which
permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
This paper focuses on the nonholonomic wheeled mobile robot. We have presented a scheme to develop controllers. Two
controllers have been developed. The first concerns the kinematic behavior, while the second relates to the dynamic behavior of the
mobile robot. For the kinematic controller, we have used a Takagi–Sugeno fuzzy system to overcome the nonlinearities present in
model, whereas for the second controller, we have used the sliding mode approach. The sliding surface has the identical structure
as the proportional integral controller. The stability of the system has been proved based on the Lyapunov approach. The
simulation results show the efficiency of the proposed control laws.

1. Introduction proposed a robust backstepping controller for the uncertain

kinematic model of the wheeled mobile robot based on a
In the last decades, the path of travel is considered as one of nonlinear disturbance observer in order to cope with model
the critical problems in the field of mobile robotics. The uncertainties and the external disturbances. Topalov [7]
trajectory tracking consists of guiding the robot through proposed an adaptive fuzzy approach for the kinematic
intermediate points to reach the final destination. This controller. This method was able to decrease the effect of
tracking is carried out under a constraint time, which means unmodeled disturbances. In [8], a dynamic Petri recurrent
that the robot must reach the goal within a predefined time. fuzzy neural network was proposed. In [9], the proposed
In the literature, the problem is treated as the tracking of a controller combines nonlinear time varying feedback with
reference robot that moves to the desired trajectory with a an integral sliding mode controller. The latter is obtained by
certain rhythm. The real robot must follow precisely the introducing an integral term in the switching manifold.
reference and reduce the distance error, by varying its linear In [10], a robust adaptive mobile robot controller is
and angular velocities [1, 2]. There are many works that have presented using backstepping for kinematics and dynamics
focused on tracking the trajectory of the mobile robots, and motions, and the adaptive process was based on the neural
they consider the mobile robot as a particle; in this case, the network. In [11], a classical parallel distributed compensation
inputs are velocities. Their aims are kinematic models. In [3], (PDC) control law, based on Takagi–Sugeno fuzzy modeling,
the kinematic control law approach supposes that the is proposed. The controller comprises sixteen rules in which
control signal generates the exact motion commanded. On the control gains have been calculated using LMI techniques.
the contrary, some works consider the kinematic aspect and In [12], the authors present an adaptive controller with
the dynamic aspect for the mobile robot. In this case, the consideration of unknown model parameters.
actuator inputs signals are torques instead of velocities [4]. In [13], the authors suggest a controller of a mobile robot
In [5], Lee et al. suggest a technique for designing the in Cartesian coordinates with an approach angle based on
tracking control of wheeled mobile robots based on a new the sliding mode. In [14], the authors combine hybrid
sliding surface with an approach angle. In [6], authors backstepping kinematic control with the adaptive integral
2 Mathematical Problems in Engineering

sliding mode kinetic control of the three-wheeled mobile

robot.
Most of the works deal with nonholonomic wheeled y
mobile robot, which is used for kinematic motion of a
classical controller arising from the backstepping method
[2, 10, 12, 14, 15]. θ
This paper includes two main contributions. First, a new
controller based on Takagi–Sugeno fuzzy systems for ki- yc
nematic motion. This lastly uses three fuzzy rules. The C
second contribution consists of developing for the dynamic
part a controller based on the sliding mode. The sliding 2L
surface, which is based on linear and angular velocities of the 2r
robot, has the similar structure as the proportional integral
controller. The switching control term of the latter controller
combines the two sliding surfaces.
The remainder of this paper is organized as follows. x
Section 2 is devoted to the description of the kinematic and O
xc
dynamic models of the two-wheeled mobile robot. Section 3
that is reserved to the controllers design includes two Figure 1: Representation of the navigation environment.
subsections, the first is reserved to the development of the
new T-S type fuzzy controller of the kinematic behavior,
where (x, y), v, and ω represent, respectively, the instanta-
whereas the second is consecrated to the design of the
neous position coordinates of point C of the mobile robot in
dynamic motion controller using the sliding mode ap-
the global Cartesian frame and the measurements at point C
proach. The stability analysis is checked in the both pre-
of the linear and angular speeds of the robot. The state
cedent subsections by the Lyapunov approach. Then, Section
variables of mobile robot are q � 􏼂 x y θ 􏼃T :
4 is sacred to the presentation of the simulation results.
√� 2
vd � x_ d + y_ 2d ,
2. Mobile Robot Modeling € d − y_ d x€d ,
x_ d y (3)
wd �
In this section, we are interested in the modeling of the x_ 2d + y_ 2d
robot, which is composed of two driving wheels and a drive
shaft in the center, as shown in Figure 1. Indeed, the Section where vd and wd represent, respectively, the desired linear
2.1 is reserved for kinematic modeling, while Section 2.2 and angular velocity.
concerns dynamic modeling. The state kinematic model of the mobile robot in Car-
We define the current position (xc , yc ) and the heading tesian frame coordinates is given by the following
angle θ, which constitute the coordinates of the middle point expression:
of the mobile robot and the angle between the heading x_ cos θ 0
direction and the x-axis to describe the current posture ⎡⎢⎢⎢ ⎤⎥⎥⎥ ⎡⎢⎢⎢ ⎥⎥⎤⎥ v(t)
position of the mobile robot. Figure 1 depicts the current q_ � ⎢⎢⎢⎣ y_ ⎥⎥⎥⎦ � ⎢⎢⎢⎣ sin θ 0 ⎥⎥⎥⎥⎦􏼢
⎢ 􏼣 � J(θ)Vm , (4)
w(t)
posture position of a two wheels mobile robot in Cartesian θ_ 0 1
frame coordinates.
The nonholonomic constraint of a wheeled mobile robot with
is given by the following equation: T
Vm � 􏼂 v ω 􏼃 ,
y_ cos θ − x_ sin θ � 0. (1) cos θ sin θ 0 T (5)
J(θ) � 􏼢 􏼣 .
0 0 1

2.1. Fuzzy Kinematic Model of Robot. Based on the New- In order to develop a T-S fuzzy controller, which sta-
ton–Euler equations [16] and the previous hypotheses, the bilizes the system and allows the robot to follow the desired
state equations of the mobile robot are represented by the path, we need a fuzzy model. In this context, we proceed to
following equations’ system [17]: determine a fuzzy model of the robot.
The posture vector error is not specified in the global
x_ � v cos θ, frame coordinate system, but quite as a vector error in the
⎪
⎧
⎪
⎨ local frame coordinate system of the robot: qe (t) �
⎪ y_ � v sin θ, T
􏼂 e1 e2 e3 􏼃 .
⎪
⎩_ (2)
θ � ω, The posture vector error qe is computed based on the
􏽱������ T
actual posture vector q(t) � 􏼂 x y θ 􏼃 and the reference
v � x_ 2 + y_ 2 , T
posture vector qd (t) � 􏼂 xd yd θd 􏼃 :
Mathematical Problems in Engineering 3

q_d (t) � J θd 􏼁Vm d , (6) e3

T
where Vm d � 􏼂 vd ωd 􏼃 . Y
So, θr
xd − x ex yr
⎢
⎡
⎢
⎢ ⎤⎥⎥⎥ ⎡⎢
⎢
⎢ ⎤⎥⎥⎥
q􏽥 � qd − q � ⎢
⎢ ⎥⎥ ⎢
⎢ ⎥⎥
⎣ yd − y ⎥⎥⎦ � ⎢
⎢
⎢ ⎣ ey ⎥⎥⎦.
⎢ (7)
θd − θ eθ 2L

The relation between the local frame and the global

frame, as shown in Figure 2, is given by the following
equation: θ
y
qe � Re q􏽥, (8) C

where 2r
X
cos θ sin θ 0 O xr x
⎢
⎡
⎢
⎢ ⎥⎥⎥⎤
Re � ⎢
⎢ −sin θ cos θ 0 ⎥⎥⎥. (9)
⎢
⎣ ⎥⎦ Figure 2: Trajectory tracking.
0 0 1
Equation (8) allows transforming the magnitudes de- The member ship function for the error e3 is given in
scribed in the global coordinate system to the local coor- Figure 3.
dinate system: The rules of the local models are given by the following
expression:
_ q􏽥 + R q􏽥_ .
q_e � R (10)
e e
if e3 is μ1 , then q_e � A1 Vm d + B1 Vm ,
However, by differentiating equation (10), which con- if e3 is μ2 , then q_e � A2 Vm d + B2 Vm , (13)
tains the linear speed and the angular speed terms, we obtain
the derivative of the error vector, which is expressed by the if e3 is μ3 , then q_e � A3 Vm d + B3 Vm .
following equation: The T-S fuzzy model of equation (12) is given by the
⎪ e_1 � ωe2 − ] + ]d cos e3 ,
⎧
⎪
following equation:
⎨
q_e � ⎪ e_2 � −ωe1 + ]d sin e3 , (11) 3
⎪
⎩
e_3 � ωd − ω. q_e � 􏽘 μi Ai Vm d + Bi Vm 􏼁, (14)
i�1
The posture error model can be rewritten as follows:
where μi and Ai and Bi represent, respectively, the weight
e_1 cos e3 0 −1 e2
⎢
⎡
⎢ ⎤⎥⎥⎥ ⎡ ⎢
⎢ ⎥⎤⎥⎥ vd ⎢
⎡
⎢ ⎥⎤⎥⎥ v assigned to each rule and the matrices associated to the local
⎢
⎢
⎢ _ ⎥
⎥ ⎢
⎢
⎢ ⎥
⎥ ⎢
⎢
⎢ ⎥⎥⎥􏼢 􏼣.
⎢
⎢ e
⎣ ⎦ ⎣
2 ⎥
⎥ � ⎢
⎢ sin e 3 0 ⎥
⎥ 􏼢
⎦ w 􏼣 + ⎢
⎢
⎣ 0 −e 1 ⎥⎦ w (12) model.
e_3 0 1
d
0 1 With,
1 0
We note that equation (12) contains trigonometric ⎡⎢⎢⎢ ⎥⎥⎤⎥
nonlinearities which are cos (e3) and sin (e3). However, the A1 � ⎢⎢⎢⎣ e3 0 ⎥⎥⎥⎥⎦,
⎢
nonlinearities depend on the error e3, whose range of var- 0 1
iation is from −pi/2 to pi/2.
The advantage of the T-S type fuzzy approach is that it 0 0
⎡⎢⎢⎢ ⎥⎥⎤⎥
allows describing the nonlinear model by linear submodels. A2 � ⎢⎢⎢⎣ 1 0 ⎥⎥⎥⎥⎦,
⎢
Indeed, each submodel represents a local linear relation
0 1
between the inputs and the outputs and all the nonlinearities (15)
are reported in the premises of the fuzzy rules [18]. 0 0
⎡⎢⎢⎢ ⎤⎥⎥⎥
Based on the theory of T-S fuzzy systems, the nonlinear A3 � ⎢⎢⎢⎣ −1 0 ⎥⎥⎥⎥⎦,
model (12) can be transformed into three local models,
which are inferred by fuzzy rules. The three local models are 0 1
described by the following systems of equations: −1 e2
From the weights assigned to each rule, the state vector ⎡⎢⎢⎢ ⎥⎥⎤⎥
B1 � B2 � B3 � ⎢⎢⎢⎢⎣ 0 −e1 ⎥⎥⎥⎥⎦.
of the fuzzy models is inferred as follows (which corresponds
to a barycentric aggregation). 0 1
4 Mathematical Problems in Engineering

μij The expressions of linear and angular velocities of the

mobile robot, (v, w), depend on the left and right linear
velocities of the motors. They are expressed by the following
μ2 1μ1 μ3 equations:
v + vl
v� r ,
2
(20)
e3 vr − vl
w� .
–Π/2 Π/2 2L
Figure 3: Membership function.

3. Design of Robot Controllers

2.2. Dynamic Model of Robot. The dynamic equation of the
wheeled mobile robot is given by the following equation: In this work, we consider the kinematic and dynamic be-
havior of the robot. The purpose of the control design is to
M(q)q€ + C(q, q) _ � B(q)τ − AT (q)λ,
_ q_ + F(q) (16) allow the robot to follow the virtual robot. The latter rep-
resents the reference robot and provides the desired path
_ is the centripetal and Coriolis matrix, F(q)
where C(q, q) _ is
defined by the following vector: qd (t) � 􏼂 xd yd θd 􏼃T .
the friction force, τ represents the torque vector, and
The architecture of the control scheme of the robot,
AT (q) � 0:
includes six blocks, as shown in Figure 4. The first block
m 0 0 generates the desired states, whereas the second block
⎢
⎡
⎢
⎢ ⎤⎥⎥⎥
⎢ ⎥⎥⎥ transforms the error from the local frame into the general
M(q) � ⎢
⎢
⎢
⎢
⎢ 0 m 0 ⎥⎥⎥, frame. The third and fourth blocks are reserved, respectively,
⎢
⎣ ⎥⎦
0 0 Jg for kinetic and dynamic controllers. The fifth and sixth
blocks, respectively, describe the behavior of the kinematic
cos θ cos θ and dynamic models of the robot.
⎢
⎡ ⎥⎤⎥⎥
1⎢
⎢
⎢
⎢ ⎥⎥⎥
B(q) � ⎢
⎢ sin θ sin θ ⎥⎥⎥, (17)
r⎢
⎢
⎣ ⎦
3.1. Fuzzy Kinematic Controller. In this section, we are in-
L −L
terested in the search for a T-S type fuzzy controller, which
0 0 0 guarantees the convergence of the kinematic errors towards
⎢
⎡
⎢
⎢ ⎤⎥⎥⎥ zero in the local coordinate system and allows the robot to
⎢ ⎥⎥⎥
_ �⎢
C(q, q) ⎢
⎢
⎢
⎢ 0 0 0 ⎥⎥⎥, follow the desired path.
⎣ ⎦
Based on the T-S fuzzy model (14), the rules for the local
0 0 0
controllers are given by the following expressions:
where m and Jg represent, respectively, the mass and the
T
moment inertia of the wheeled mobile robot. L and r ]d e2
if e3 is μ1 , then Vm � 􏼢 ] � K1 e1 + ]d ω � K3 e3 + + ωd 􏼣 ,
represent, respectively, the distance separating the two e3
driving wheels and the wheel radius. Without considering
T
disturbances and uncertainties, the latest equation becomes ]d e2
if e3 is μ2 , then Vm � 􏼢 ] � K1 e1 + ]d ω � K3 e3 + + ωd 􏼣 ,
as e3
M(q)V_ m � B(q)τ, (18) ]d e2 T
if e3 is μ3 , then Vm � 􏼢 ] � K1 e1 + ]d ω � K3 e3 + + ωd 􏼣 .
where e3

m 0 (21)
⎢
M(q) � ⎡
⎣ ⎤⎥⎦,
0 Jg The global T-S fuzzy controller is given by the following
equation:
1⎡ 1 1
⎢
B(q) � ⎣ ⎥⎦⎤,
r L −L 3 3

(19) Vm � ⎡⎣ ]c � 􏽘 μi ]i � K1 e1 + μ1 ]ω
d c � 􏽘μi ωi
v i�1 i�1
⎢
Vm � ⎡
⎣ ⎤⎥⎦, (22)
T
w ]d e 2
� ωd + K3 e3 + μ1 ]d e2 + μ2 − μ3 􏼁 􏼣 .
τr e3
⎢
τ �⎡
⎣ ⎤⎥⎦.
τl If e3 � 0, then μ1 � 1 and μ2 � μ3 � 0, so
Mathematical Problems in Engineering 5

T-S fuzzy Vc ec Sliding τ1, τr V

Reference Re mode Dynamic kinematic
+ kinematic +
trajectory qr – – dynamic model model
controller
q controller
V

Figure 4: Architecture of the robot controller.

3 towards zero for any arbitrary reference trajectory. However,

⎣ ]c � 􏽘 μi ]i � K1 e1 + μ1 ]ω
Vm � ⎡ dc we have developed a controller based on the sliding mode
i�1 approach because the latter is considered a robust approach
(23)
3 T [19, 20]. In this case, we define two sliding surfaces. The first
� 􏽘 μi ωi � ωd + K3 e3 + μ1 ]d e2 ⎤⎦ . surface depends on linear velocity, while the second uses
i�1 angular velocity, S � 􏼂 sv sw 􏼃T :
Sv t
S(t) � 􏼢 􏼣 � ee (t) + K 􏽚 ee (δ)dδ, (30)
3.1.1. Stability Analysis. To check the stability of the robot, Sw 0
we use Lyapunov’s theory. However, we choose the fol-
lowing Lyapunov candidate function: where sv and sw are given, respectively, by equations (31) and
T
(32), ee � (Vc − Vm ) � 􏼂 ev ew 􏼃 . With ev � vc − v and
1 1 1 ew � wc − w,
V � e21 + e22 + e23 . (24)
2 2 2
t
The derivative of Lyapunov function is as follows. Sv (t) � ev (t) + kv 􏽚 ev (δ)dδ, (31)
So, 0

V_ � e_1 e1 + e_2 e2 + e_3 e3 , t

Sw (t) � ewv (t) + kw 􏽚 ew (δ)dδ. (32)
V_ � μ1 ωe2 − ] + ]d 􏼁 + μ2 ωe2 − ]􏼁 + μ3 ωe2 − ]􏼁􏼁e1 0

+ μ1 −ωe1 + ]d e3 􏼁 + μ2 −ωe1 + ]d 􏼁 + μ3 −ωe1 − ]d 􏼁􏼁e2 However, the derivatives of the sliding surfaces sv (t) and
sw (t) are given by the following expressions:
+ μ1 ωd − ω􏼁 + μ2 ωd − ω􏼁 + μ3 ωd − ω􏼁􏼁e3 .
(25) S_v (t) � e_v (t) + kv ev (t),
(33)
So, S_w (t) � e_w (t) + kw ew (t).
] e
V_ � −] + μ1 ]d 􏼁􏼁e1 + 􏼠μ1 ]d e2 + ωd + μ2 − μ3 􏼁 d 2 − ω􏼡e3 . The dynamic motion of the robot is described by
e3 equation (7) which can be transformed as
(26)
V_ m � (M(q))− 1 B(q)τ. (34)
If we choose the following linear and angular velocities,
Equation (34) can be written as
]c � K1 e1 + μ1 ]d ,
􏽥
V_ m � Bτ, (35)
3 (27)
]d e2 where B􏽥 � (M)− 1 B
ωc � 􏽘 μi ωi � ωd + K3 e3 + μ1 ]d e2 + μ2 − μ3 􏼁 .
i�1
e3 Based on the sliding mode theory, the controller includes
two terms which are known as equivalent control law and
Equation (26) becomes switching control. The global control law is expressed as
V_ � −K1 e21 − K3 e23 ≤ 0, if e3 ≠ 0. (28) u � τ � ueq + us � τ eq + τ s . (36)
If e3 � 0, then μ1 � 1 and μ2 � μ3 � 0. So, ]c � K1 e1 + ]d The equivalent control law ueq is computed by recog-
and ωc � 􏽐3i�1 μi ωi � ωd + K3 e3 + ]d e2 . nizing that S_ � 0 which is a necessary condition for the state
Also, trajectory to stay in the sliding surface [19, 20]. The de-
V_ � −K1 e21 − K3 e23 ≤ 0. (29) rivative of the sliding surface is
_ � e_e (t) + Kee (t),
S(t) (37)
The derivative of the Lyapunov function is negative and
the stability of the system is guarantee. with e_e � (V_ c − V_ m ) and ee � (Vc − Vm ).
Thus, substituting (35) for (37), we obtain
3.2. Dynamic Controller Based on Sliding Mode. In this 0
section, we are interested in the development of a controller, _ � V_ c − Bτ
S(t) 􏽥 + Kee (t) � 􏼢 􏼣, (38)
which guarantees the convergence of the posture error qe 0
6 Mathematical Problems in Engineering

with However, the equivalent control law ueq � τ eq can be

computed as
1 ⎡ J 0
M
−1
� ⎣ g ⎦⎤,
mJg 0 m 􏽥 − 1 V_ c + B
ueq � τ eq � B 􏽥 − 1 Kee (t), (40)

1 ⎡ J J with
􏽥 � M− 1 B �
B ⎣ g g ⎦⎤,
rmJg mL mL
(39)
mL Jg
􏽥− 1 � r ⎢
B ⎡
⎣ ⎥⎦⎤,
2L mL −Jg

−1 1 ⎡ L 1
B � ⎣ ⎦⎤.
2Lr L −1

]_ c � K1 e_1 + μ1 ]_ d
⎢
⎡
⎢
⎢ ⎤⎥⎥⎥
⎢
⎢ ⎥⎥⎥
V_ c � ⎢
⎢
⎢
⎢
⎥⎥⎥. (41)
⎣ ω_ � ω_ + K e_ + μ ]_ e + ] e_ 􏼁 + μ − μ 􏼁 ]_ d e2 + ]d e_2 􏼁e3 − ]d e2 e_3 ⎥⎥⎦
⎢
⎢
c d 3 3 1 d 2 d 2 2 3
e23

So,

rm rJg K1 e_1 + μ1 ]_ d
⎢
⎡
⎢
⎢ ⎥⎥⎤⎥ ⎡ ⎢ ⎤⎥⎥⎥
ueqr 2 2L ⎜ ⎟
⎡⎢⎢⎢⎣
⎢
⎢
⎤⎥⎥⎥⎦ � ⎢
⎢
⎥
⎥
⎥⎥⎥⎜ ⎛
⎜
⎜
⎜
⎢
⎢
⎢
⎢ ⎥⎥⎥ ⎡⎢⎢ λev ⎤⎥⎥⎞⎟
⎟
⎟
⎟
⎢
⎢
⎢ ⎥
⎥⎥⎥⎜⎜
⎜ ⎢
⎢
⎢ ⎥
⎥ + ⎣ ⎦ ⎟
⎟
⎟ ,
⎢
⎢ ⎥ ⎜
⎝ ⎢
⎢
⎣ ]
_ e + ] _
e 􏼁e − ] e _
e ⎥
⎥
⎦ ⎟
⎠
ueql ⎢
⎣ rm rJg ⎥⎦
⎢ ω_ d + K3 e_3 + μ1 ]_ d e2 + ]d e_2 􏼁 + μ2 − μ3 􏼁 d 2 d 2 3 d 2 3 λe w
− e3
2 2L
rm rJg K1 e_1 + μ1 ]_ d + λev
⎢
⎡ ⎥⎤⎡
ueqr ⎢
⎢
⎢
⎢
⎢ 2 2L ⎥⎥⎥⎥⎥⎢ ⎢
⎢
⎢
⎤⎥⎥⎥
⎥⎥⎥
⎡⎢⎢⎢⎣ ⎤⎥⎥⎥⎦ � ⎢
⎢
⎢ ⎥⎥⎥⎢⎢
⎢ ⎥⎥⎥,
⎢
⎢ ⎥
⎥ ⎢
⎢
⎢ ⎥⎥
⎢ ⎥⎢ ]_ e ] e ] e e_
ueql ⎣ rm rJg ⎥⎥⎦⎣ ω_ d + K3 e_3 + μ1 ]_ d e2 + ]d e_2 􏼁 + μ2 − μ3 􏼁􏼠 d 2 + d 2 − d 22 3 􏼡 + λew ⎦
⎢
⎢
− e3 e3 e3
2 2L
rm rJg ]_ e + ]d e_2 􏼁e3 − ]d e2 e_3
⎢
⎡
⎢
⎢ K1 e_1 + μ1 ]_ d + λev 􏼁 + 􏼠ω_ d + K3 e_3 + μ1 ]_ d e2 + ]d e_2 􏼁 + μ2 − μ3 􏼁􏼠 d 2 􏼡 + λew 􏼡 ⎤⎥⎥⎥
ueqr ⎢
⎢
⎢
⎢
2 2L e23 ⎥⎥⎥
⎥⎥⎥
⎡⎢⎢⎢⎣ ⎢
⎤⎥⎥⎥⎦ � ⎢
⎢ ⎥⎥⎥.
⎢
⎢
⎢ ⎥⎥⎥
⎢
⎢ ⎥⎥
ueql ⎢
⎢
⎣ rm K e_ + μ ]_ + λe 􏼁􏼁 − rJg 􏼠ω_ + K e_ + μ ]_ e + ] e_ 􏼁 + μ − μ 􏼁􏼠 ]_ d e2 + ]d e_2 􏼁e3 − ]d e2 e_3 􏼡 + λe 􏼡 ⎥⎥⎦
⎢
1 1 1 d v d 3 3 1 d 2 d 2 2 3 w
2 2L e23
(42)

Finally, we obtain

rm rJg ]_ e + ]d e_2 􏼁e3 − ]d e2 e_3

ueqr � K1 e_1 + μ1 ]_ d + λev 􏼁 + 􏼠ω_ d + K3 e_3 + μ1 ]_ d e2 + ]d e_2 􏼁 + μ2 − μ3 􏼁􏼠 d 2 􏼡 + λew 􏼡,
2 2L e23
(43)
rm rJg ]_ e + ]d e_2 􏼁e3 − ]d e2 e_3
ueql � K1 e_1 + μ1 ]_ d + λev 􏼁􏼁 − 􏼠ω_ d + K3 e_3 + μ1 ]_ d e2 + ]d e_2 􏼁 + μ2 − μ3 􏼁􏼠 d 2 􏼡 + λew 􏼡.
2 2L e23
Mathematical Problems in Engineering 7

The switching control term is generally chooses as Evolution of mass

us � −ηsign(S), with η≻0. This term can be canceled when 30
the system reaches the sliding surface [20]. In this paper, the 25
switching control law is chosen as follows: 20
15
−ks1 sgn sv + sw 􏼁􏼁
τs � 􏼢 􏼣. (44) 10
−ks1 sgn sv − sw 􏼁􏼁 5
0
0 1000 2000 3000 4000 5000
3.2.1. Reaching Condition and Stability Analysis. To verify Figure 5: Evolution of mass of the robot.
the reaching condition, we need to just check the following
condition. The Lyapunov candidate function is chosen as
Evolution of inertia
1 1
V � ST S. (45) 0.9
2 0.8
0.7
0.6
The derivative can be expressed as 0.5
0.4
V_ � ST S,_ 0.3
0.2
0.1
V_ � ST 􏽨V_ c − Bτ
􏽥 + Kee (t)􏽩, 0
(46) 0 1000 2000 3000 4000 5000
V_ � ST 􏽨V_ c − B
􏽥 􏼐τ eq + τ s 􏼑 + Kee (t)􏽩, Times

V_ � ST S_ � ST 􏽨V_ c − B
􏽥 􏼐τ eq + τ s 􏼑 + Kee (t)􏽩. Figure 6: Evolution of inertia of the robot.

Based on equation (40), V_ c − Bτ

􏽥 eq + Kee (t) � 0
Evolution of radius
So, 0.16
rm rJg 0.14
⎡⎢⎢⎢ ⎥⎤ 0.12
⎢⎢⎢ 2 2L ⎥⎥⎥⎥⎥ −ks sgn sv + sw 􏼁􏼁 0.1
T T
V_ � S S_ � S ⎢⎢⎢⎢ ⎢ ⎥⎥⎥⎡⎢⎢⎣ ⎤⎥⎥⎦,
⎥⎥⎥ 0.08
⎢⎢⎢ ⎥
⎣ rm rJg ⎥⎦ −ks sgn sv − sw 􏼁􏼁 0.06
− 0.04
2 2L 0.02
rJg 0
rm 0 1000 2000 3000 4000 5000
⎡⎢⎢⎢ −ks sgn sv + sw 􏼁 − ks sgn sv − sw 􏼁 ⎤⎥⎥
⎥⎥⎥
⎢⎢⎢ 2 2L ⎥⎥⎥
T
V_ � S ⎢⎢⎢⎢ ⎢ ⎥⎥⎥, Figure 7: Evolution of radius of the robot.
⎢⎢⎢ ⎥⎥⎥
⎣ rm rJg ⎦
−ks sgn sv + sw 􏼁 + ks sgn sv − sw 􏼁
2 2L Evolution of x1 and x1ref
6.5
rm rJg
⎢⎢⎢⎡ −ks sgn sv + sw 􏼁 − ks sgn sv − sw 􏼁 ⎥⎥⎤ 6
⎢⎢⎢ 2 2L ⎥⎥⎥
_ ⎢
⎢ ⎥⎥⎥ 5.5
s
V � 􏼂 sv w 􏼃⎢⎢ ⎢ ⎥⎥⎥,
⎢⎢⎢⎣ rJg ⎥⎥⎥⎦ 5
rm
−ks sgn sv + sw 􏼁 + ks sgn sv − sw 􏼁
2 2L 4.5
4
k rm sv + sw 􏼁 ks rJg sv − sw 􏼁
V_ � s sgn sv + sw 􏼁 − sgn sv − sw 􏼁, 3.5
2 2L
3
k rm 􏼌􏼌 􏼌􏼌 ks rJg 􏼌􏼌 􏼌
􏼌􏼌 sv − sw 􏼁􏼌􏼌􏼌. 2.5
V_ � − s 􏼌􏼌 sv + sw 􏼁􏼌􏼌 −
2 2L
2
(47) 0 1000 2000 3000 4000 5000

We note k1s � (ks rm)/2 and k2s � (ks rJg )/2L. So, Figure 8: Evolution of the robot trajectory along the x-axis (blue
􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 denotes the trajectory and magenta denotes the reference
V_ � −ks 􏼌􏼌sv + sw 􏼌􏼌 − k2s 􏼌􏼌sv − sw 􏼌􏼌≺0. (48) trajectory).

4. Simulation Results (see Figure 1) are m � 17 kg, J � 0.537 kg · m2 , L � 0.24 m,

and r � 0.1 m and initial position 􏼂 xo yo θ 􏼃 � 􏼂 5 5 π/12 􏼃.
In this part, we present the simulation results of the non- The reference speed has been chosen as the variable as shown by
holonomic wheeled mobile robot. The parameters of the robot the following equation:
8 Mathematical Problems in Engineering

Evolution of x2 and x2ref Evolution of e3

8 0.8
7.5 0.7
7 0.6
0.5
6.5
0.4
6
0.3
5.5 0.2
5 0.1
4.5 0
4 –0.1
0 1000 2000 3000 4000 5000
3.5
3 Figure 12: Evolution the angular error e3.
0 1000 2000 3000 4000 5000
Figure 9: Evolution of the robot trajectory along the y-axis (blue
denotes the trajectory and magenta denotes the reference Evolution of X2r-X2
0.15
trajectory).
0.1

Evolution of speed 0.05

2
0
1.8
–0.05
1.6
1.4 –0.1

1.2 –0.15
0 1000 2000 3000 4000 5000
1
0.8 Figure 13: Evolution of the error xr2 − x2.
0.6
0.4 Evolution of error X1ref – X1
0.2 0.05
0 1000 2000 3000 4000 5000
Figure 10: Evolution of linear speed v. 0

–0.05
Evolution of error speed
0.6 –0.1

0.4
–0.15
0.2
–0.2
0

–0.2 –0.25
0 1000 2000 3000 4000 5000
–0.4 Figure 14: Evolution the error xr1 − x1.

–0.6
0 1000 2000 3000 4000 5000
Figures 5–7 show, respectively, the evolution in time of
Figure 11: Evolution of the error of the linear speed ev � vr − v. mass, inertia, and radius of the robot.
Figures 8 and 9 show, respectively, the evolution in time
of the robot along the x- and y-axis (the trajectories of robot
⎧
⎪ v � vro + 910− 7 t, for t ≤ 15 s, and its reference).
⎪
⎪
⎨ r Figures 10 and 11 show, respectively, the behavior
⎪ vr � vr , for 15 ≤ t ≤ 27 s, (49) evolution of linear speed v and the error speed ev � vr − v.
⎪
⎪
⎩ −7 Figures 8–10 show that all the signals follow quickly the
vr � vr − 910 t, for t≻27 s and vr ≥ 0.8.
reference. The curves of the reference signals and the re-
The initial speeds have been chosen as vro � 0.8, sponses overlap in a very short time after start-up.
vd � 0.78, and vg � 0.38. Figure 12 shows the evolution the angular error e3.
Mathematical Problems in Engineering 9

Evolution of e1 Conflicts of Interest

0.045
0.04 The author declares that there are no conflicts of interest.
0.035
0.03
0.025 References
0.02
0.015 [1] T. Fukao, H. Nakagawa, and N. Adachi, “Adaptive tracking
0.01 control of a nonholonomic mobile robot,” IEEE Transactions
0.005 on Robotics and Automation, vol. 16, no. 5, pp. 609–615, 2000.
0 [2] G. Zidani, D. Said, and L. Chrifi-Alaoui, A. Benmakhlouf and
–0.005 S. Chaouch, Back- stepping controller for a wheeled mobile
0 1000 2000 3000 4000 5000
robot,” in Proceedings of the 4th International Conference on
Figure 15: Evolution of e1. Systems and Control (ICSC’15), Sousse,Tunisia, April 2015.
[3] M. Egerstedt, X. Hu, and A. Stotsky, “Control of mobile
platforms using a virtual vehicle approach,” IEEE Transactions
Evolution of e2 on Automatic Control, vol. 46, no. 11, pp. 1777–1782, 2001.
0.25 [4] J. Yang and J. kim, “Sliding mode motion control of non-
0.2 holonomic mobile robots,” IEEE Control Systems, vol. 19,
no. 2, pp. 15–23, 1999.
0.15 [5] J.-K. Lee, J.-B. Park, and Y.-H. Choi, “Tracking control of
nonholonomic wheeled mobile robot based on new sliding
0.1
surface with approach angle,” in Proceedings of the 3rd IFAC
0.05 Symposium on Telematics Applications The International
Federation of Automatic Control, Seoul, Republic of Korea,
0 November 2013.
[6] M. A. Moqbel Obaid, A. R. Husain, and A. A. Mohammed Al-
–0.05
0 1000 2000 3000 4000 5000 kubati, “Robust backstepping tracking control of mobile robot
based on nonlinear disturbance observer,” International
Figure 16: Evolution of e2. Journal of Electrical and Computer Engineering (IJECE), vol. 6,
no. 2, pp. 901–908, 2016.
[7] A. V. Topalov, J.-H. Kim, and T. P. Proychev, “Fuzzy-net
Figures 13 and 14 show, respectively, the behavior control of non-holonomic mobile robot using evolutionary
evolution of errors x2ref − x2 and x1ref − x1. feedback-error-learning,” Robotics and Autonomous Systems,
Figures 15 and 16 show, respectively, the behavior vol. 23, no. 3, pp. 187–200, 1998.
evolution of errors e1 and e2. [8] R.-J. Chia-Ming Liu and C.-M. Liu, “Design of dynamic Petri
Figures 12–16 show that all the errors rapidly tend to- recurrent fuzzy neural network and its application to path-
wards zero. This clearly shows that the system converges tracking control of nonholonomic mobile robot,” IEEE
towards the desired trajectories in a very short time. This Transactions on Industrial Electronics, vol. 56, no. 7,
proves the effectiveness of the proposed command. pp. 2667–2683, 2009.
[9] A. Bessas, B. Atallah, and F. Boudjema, “Integral sliding mode
control for trajectory tracking of wheeled mobile robot in
presence of uncertainties,” Journal of Control Science and
5. Conclusion Engineering, vol. 2016, Article ID 7915375, 10 pages, 2016.
[10] R. Fierro and F. L. Lewis, “Control of a nonholomic mobile
In this paper, we have presented a scheme to control a robot: backstepping kinematics into dynamics,” Journal of
nonholonomic robot. We have proposed two controllers. Robotic Systems, vol. 14, no. 3, pp. 149–163, 1997.
The first concerns the kinematic behavior, while the second [11] E.-H. Guechi, J. Lauber, M. Dambrine, G. Klancar, and
relates to the dynamic behavior of the mobile robot. To S. Blažic, “PDC control design for non-holonomic wheeled
overcome the nonlinearities present in the model, we have mobile robots with delayed outputs,” Journal of Intelligent and
used a Takagi–Sugeno fuzzy system for the kinematic Robotic Systems, vol. 60, no. 3, pp. 395–414, 2010.
controller. For the second controller, we have used the [12] Y. Hung-Hsiu and H.-K. Huang, “Experimental verification
sliding mode technique, which is known as a robust tech- for adaptive tracking control of a nonholonomic mobile
nique. The proposed sliding surface has the same structure as robot,” in Proceedings of the 33rd Annual Conference of the
the proportional integral controller. The Lyapunov approach IEEE Industrial Electronics Society (IECON), Taipei, Taiwan,
November 2007.
has been used to prove stability of the system. The results
[13] J. K. Lee, Y. H. Choi, Y. H. Choi, and J. B. Park, Sliding mode
presented in Section 4 show the efficiency of the proposed tracking control of mobile robots with approach angle in
control laws. cartesian coordinates,” International Journal of Control,
Automation and Systems, vol. 13, no. 3, pp. 718–724, 2015.
[14] A. Nikranjbar, A. Masoud Haidari, and B. Ali Asghar Atai,
Data Availability “Adaptive sliding mode tracking control of mobile robot in
dynamic environment using artificial potential fields,” Journal
There is no specific data used to support the conclusions of
of Computer and Robotics, vol. 11, no. 1, pp. 1–14, 2018.
the study. All the necessary information exists in the paper.
10 Mathematical Problems in Engineering

[15] K. L. Besseghieur, R. Tre˛biński, W. Kaczmarek, and

J. Panasiuk, “Trajectory tracking control for a nonholonomic
mobile robot under ROS,” Journal of Physics Conference Se-
ries, vol. 1016, Article ID 012008, 2018.
[16] J. Craig, Introduction to Robotics: Mechanics and Control,
Addison-Wesley, Harlow, England, 1em plus 0.5em minus
0.4em, 1989.
[17] M. K. Bugeja, S. G. Fabri, and L. Camilleri, “Dual adaptive
dynamic control of mobile robots using neural networks,”
IEEE Transactions on Systems, Man, and Cybernetics, Part B
(Cybernetics), vol. 39, no. 1, pp. 129–141, 2009.
[18] Y. Morère, “Mise en oeuvre de lois de commande pour les
modèles flous de type TakagiSugeno,” Université de Valen-
ciennes et du Hainaut-Cambrésis, Valenciennes, France,
Thèse de doctorat, 2001.
[19] H. Abid, C. Mohamed, and T. Ahmed, “An indirect adaptive
fuzzy sliding mode controller for a class of SISO non-linear
systems,” International Journal of Modelling, Identification
and Control, vol. 4, no. 2, 2008.
[20] Utkin, “Variable structure systems with sliding mode,” IEEE
Transactions on Automatic Control, vol. 22, no. 2, pp. 212–222,
1997.