Kinematics, dynamics and control design of 4WIS4WID mobile robots

Ming-Han Lee, Tzuu-Hseng S. Li

aiRobots Laboratory, Department of Electrical Engineering, National Cheng Kung University, 1 University Road, Tainan,
Taiwan 70101
E-mail: [email protected]

Published in The Journal of Engineering; Received on 2nd September 2014; Accepted on 30th October 2014

Abstract: Kinematic and dynamic modelling and corresponding control design of a four-wheel-independent steering and four-wheel-
independent driving (4WIS4WID) mobile robot are presented in this study. Different from the differential or car-like mobile robot, the
4WIS4WID mobile robot is controlled by four steering and four driving motors, so the control scheme should possess the ability to integrate
and manipulate the four independent wheels. A trajectory tracking control scheme is developed for the 4WIS4WID mobile robot, where both
non-linear kinematic control and dynamic sliding-mode control are designed. All of the stabilities of the kinematic and dynamic control laws
are proved by Lyapunov stability analysis. Finally, the feasibility and validity of the proposed trajectory tracking control scheme are confirmed
through computer simulations.

1 Introduction diversified so it can implement some motions such as lateral

Many structures of vehicles or mobile robots have been presented, moving or zero-turn radius. The 4WIS4WID mobile robot can
such as two-wheel-steering (2WS), four-wheel-steering (4WS), dif- achieve the same actions as car-like, 4WS and skid-steer mobile
ferential wheel, skid-steer and omni-directional drive. In recent robots. Moreover, the 4WIS4WID mobile robot can also implement
years, because of the rapid expansion of electric motors and their the same motions as the omni-directional mobile robot. However,
high manoeuvrability, vehicles or mobile robots based on the the 4WIS4WID mobile robot has to turn the orientation of wheels
four-wheel-independent steering and four-wheel-independent so it cannot instantly move in any direction like the omni-
driving (4WIS4WID) structure have been presented and researched directional mobile robot.
in many areas such as transportation, agriculture and service industry. The first purpose of this paper is to present the kinematic and
The 2WS or car-like structure [1–4] is like that of a traditional car, dynamic models, which can be utilised to design controllers to
but in this structure only the front wheels can steer and they are deal with control issues for 4WIS4WID mobile robot. The
coupled by mechanical linkages. The 4WS [5, 6] can steer the 4WIS4WID mobile robot has high manoeuvrability; the control
front wheels and rear wheels, but the wheels are coupled with mech- problem is more complex because it has to manipulate four-wheel
anical linkages so that whether they are 2WS or 4WS, the steering motors and four steering motors to move smoothly. Reference
angles are limited and are restricted by their mechanical structures. [25] presents an omni-directional steer-by-wire system to control
This limitation means that the vehicles or robots of 2WS or 4WS the steering angle of four wheels, where the instantaneous centre
cannot implement some motions, for example, lateral moving. of rotation (ICR) is considered. When robot moves straight, the
The differential wheeled mobile robot [7–11] has two parallel ICR is undefined or infinite. This disadvantage means that the
driving wheels on the right side and left side of the robot and one or control law has to be distinguished into three conditions, turning
two caster wheels to balance the robot. The two driving wheels right, moving straight and turning left; then it has to switch them.
drive independently, and this type of robot uses the velocity difference The kinematic control method proposed in [26] is also based on
between the two wheels to turn the direction and move. Another the ICR, where one extra control mode has to be induced when
similar type is skid-steer, which uses the same way of driving, but dif- the desired rotational velocity of mobile robot is zero.
ferent from the differential wheeled mobile robot, it may have four or Furthermore, the velocity control scheme developed in [26] also
more driving wheels [12, 13] or be equipped with tracks [14] without needs to be divided into three control modes based on three condi-
castor wheels such as skid-steer loaders or backhoes. tions. In [27], Selekwa and Nistler proposed a kinematic control law
The omni-directional mobile robots [15–18] are equipped with to control the steering angle and wheel velocity of four wheels,
omni wheels so that they can move in any direction instantly where the virtual rear and front steering angles should be given
without turning the direction of the wheels or changing the robot first. This phenomenon limits its application. Moreover, if the
orientation. Although this type has high manoeuvrability, it still virtual rear or front steering angle is zero, the wheel steering
has some disadvantages. The first disadvantage is that all the angle cannot be calculated. If the steering angle of any wheel is
wheels do not roll in the direction of the robot movement, so the zero, the wheel velocity cannot be calculated or is infinite.
efficiency is low. The second disadvantage is omni-wheel itself To design a better controller for the 4WIS4WID mobile robot, we
contains many wheels/rollers which lead greater resistance to present a kinematic model that includes eleven state variables, and a
rotate and imply greater loss of energy. dynamic model, which is developed by the Lagrangian equation. On
There are some vehicles and wheeled mobile robots, Nissan Pivo the basis of the proposed models, we design a controller that
2, Nissan Pivo 3, Toyota Fine-T, Toyota Fine-X, humanoid robot combines kinematic control with dynamic control to implement the
Justin [19], OK-1 [20], Care-O-bot 3 [21] and others [22–24] trajectory tracking control problem. Through the non-linear kine-
which adopt the 4WIS4WID structure. The 4WIS4WID mobile matic control law, the velocity and steering angle of every wheel
robot is equipped with four wheels, but the four wheels are not can be determined appropriately. In dynamic control law, this
coupled by mechanical linkages. Each wheel is steered independ- paper proposes a trajectory tracking control scheme based on sliding-
ently using a motor, which is equipped between the body and mode control (SMC) to reduce the tracking error with system uncer-
each wheel of the mobile robot. On the basis of the structure of tainties and external disturbances. The SMC and its extensions have
the 4WIS4WID mobile robot, the manoeuvrability is more been applied to many kinds of mobile robots [5, 28–30].

J Eng 2014 This is an open access article published by the IET under the Creative Commons
doi: 10.1049/joe.2014.0241 Attribution License (http://creativecommons.org/licenses/by/3.0/)
1
 The main contributions of this paper are as follows: (i) both kine- rigid body constraint for the robot which are represented as
matic and dynamic models of the 4WIS4WID mobile robot are
developed. They can help researchers to design and develop other vxi = vi cos (di ) = vx − yrwi v (1)
controllers and/or applications for the 4WIS4WID mobile robot.
(ii) both non-linear kinematic control and SMC-based dynamic vyi = vi sin (di ) = vy + xrwi v (2)
control schemes for the trajectory tracking problem of
4WIS4WID mobile robot are presented and the stabilities are con- By substituting the parameters of the four wheels into (1) and (2),
firmed by the Lyapunov stability theory. Moreover, the comparison the relationship of the four wheels and body of the robot can be
with the proportional–integral (PI) dynamic control is given. (iii) represented as
both the model and control system are not examined based on the
⎡ ⎤
concept of ICR, so the control algorithm is not necessarily ⎡ ⎤ v1
divided into some conditions addressed in the literature [25, 26]. vx ⎢ v2 ⎥
P vy = X ⎣ ⎥
⎣ ⎦ ⎢ (3)
The organisation of this paper is as follows: in Section 2, both the v3 ⎦
kinematic and dynamic models of the 4WIS4WID mobile robot are v
v4
addressed. Section 3 describes the trajectory tracking problem for
the 4WIS4WID mobile robot. Furthermore, the kinematic and ⎡ ⎤ ⎡ ⎤
1 0 −b c(d1 ) 0 0 0
dynamic controllers are presented. The trajectory tracking simula- ⎢0 1 ⎥
a ⎥ ⎢ s(d1 ) 0 0 0 ⎥
⎢ ⎢ ⎥
tion results based on the proposed control scheme proposed are ⎢1 0 −b ⎥ ⎢ 0 c(d2 ) 0 0 ⎥
⎢ ⎥ ⎢ ⎥
illustrated in Section 4. Finally, Section 5 gives conclusions. ⎢0 1 −a ⎥ ⎢ s(d2 ) 0 ⎥
where P = ⎢ ⎥, X = ⎢ 0 0 ⎥,
⎢1 0 b ⎥ ⎢ 0 0 c(d3 ) 0 ⎥
⎢ ⎥ ⎢ ⎥
2 System description ⎢0 1 ⎥
−a ⎥ ⎢ 0 0 s(d3 ) 0 ⎥
⎢ ⎢ ⎥
This section presents the kinematic and dynamic models of the ⎣1 0 b ⎦ ⎣ 0 0 0 c(d4 ) ⎦
4WIS4WID mobile robot. One can find that the concept of ICR 0 1 a 0 0 0 s(d4 )
is not necessary to setup these models.
and where c(·) and s(·) denote the trigonometric functions cos(·) and
2.1 Kinematic model sin(·), respectively. Moreover, the pseudo-inverse matrix of P is as
A simplified 4WIS4WID mobile robot system with four wheels, follows
without considering the motions of pitch, roll and heave, is ⎡ ⎤
shown in Fig. 1 and the configuration is depicted in Fig. 2. The 1/4 0 1/4 0 1/4 0 1/4 0
+ ⎢ 0 1/4 0 1/4 0 1/4 0 1/4 ⎥
robot is driven by four-wheel motors and through another four P =⎣ ⎦
motors to turn these four-wheel motors. x, y and θ are the postures b a b a b a b a
− − − −
of the robot in the world coordinate frame; wi denotes the wheel i; δi K K K K K K K K
denotes the steering angle of wheel i; a and b are, respectively, the
length between the centroid of the robot and each wheel. Before where K = 4a 2 + 4b 2. Premultiply (3) by P +, one can obtain the
developing the model of 4WIS4WID mobile robot, three configura- relationship of the four wheels and robot’s body as follows
tions for the robot are assumed as below. ⎡ ⎤
⎡ ⎤ c(d1 ) c(d2 ) c(d3 ) c(d4 ) ⎡ ⎤
v1
Assumption 1: The position vx ⎢ 4 4 4 4 ⎥ ⎢ v2 ⎥

of thefour wheels
in the robot  coordinate ⎣ vy ⎦ = ⎢
⎢ s(d1 ) s(d2 ) s(d3 ) ⎥
s(d4 ) ⎥⎢ ⎥ (4)

frame is predefined
 as xrw1 , yrw1 =
(a, b), xrw2 , yrw2 = (−a, b), ⎣ ⎦⎣ v3 ⎦
r
, yrw3 = ( − a, − b) xrw4 , yrw4 = (a, −b). The v 4 4 4 4

xw3 and
W1 W2 W3 W4 v4
xrwi , yrwi denotes the coordinates of wheel i in the robot coordinate
frame. v and vi are, respectively, the linearity velocity of the robot
 
and wheel i, where v = v2x + v2y and vi = v2xi + v2yi . ω is the
angular velocity of the robot’s body. ωi is the steering velocity of
wheel i.

Assumption 2: The positions of centre of mass and centroid of the

4WIS4WID mobile robot are the same.

Assumption 3: Both the radius and mass of all wheels are assumed
to be the same.

Under no slipping condition, the relationships of the velocity

between each wheel and the body of robot are obtained by the

Fig. 1 4WIS4WID robot system Fig. 2 Configuration of 4WIS4WID robot system in world coordinate frame

This is an open access article published by the IET under the Creative Commons J Eng 2014
Attribution License (http://creativecommons.org/licenses/by/3.0/) doi: 10.1049/joe.2014.0241
2



where Wi = −yrwi c(di ) + xrwi s(di ) / 4(xrwi )2 + 4(yrwi )2 . Note that Owing to v2x + v2y = ẋ2 + ẏ2 , (7) can be rewritten as
P +P = I3, where I 3 [ <3×3 is the identity matrix.
Combining the three states of the robot body, wheel rotating
angle and wheel steering angle, there are 11 state variables to 1 1 1 2 Iw  4
I  4
2
E = mẋ2 + mẏ2 + I u̇ + ẇ2i + d ḋ (8)
represent the position and posture of the mobile robot 2 2 2 2 i=1 2 i=1

T
q = x y u w1 w2 w3 w4 d1 d2 d3 d4 (5) Without considering the disturbances, the Lagrange formula is used
to derive the dynamic equation
After some calculations, one can obtain the kinematic model of
4WIS4WID mobile robot as follows  
d ∂E ∂E .
⎡c ⎤ − + F(q ) + td = N(q)t, (j = 1, 2, 3, . . . , 11)
⎡ ⎤ 1 c2 c3 c4 dt ∂q. j ∂qj
ẋ 0 0 0 0
⎢4 4 4 4 ⎥
⎢ ẏ ⎥ ⎢ s1 s2 s3 s4 ⎥ (9)
⎢ ⎥ ⎢ 0 0 0 0⎥ ⎡ ⎤
⎢ ⎥ ⎢4 ⎥ v1 .
⎢ u̇ ⎥ ⎢ 4 4 4 ⎥ where t is the torque input vector, F(q ) denotes the surface friction
⎢ ⎥ ⎢ W1 W2 W3 W4 0 0 0 0⎥ ⎢ ⎥
⎢ ẇ ⎥ ⎢ ⎥⎢ v2 ⎥ forces and td is the bounded disturbances. After doing some calcu-
⎢ 1 ⎥ ⎢ r−1 ⎥
0 0 0 0 ⎥⎢ ⎥
⎢ ⎥ ⎢ 0 0 0 ⎢v ⎥ lations, the dynamic model is obtained and represented as follows
⎢ ẇ2 ⎥ ⎢ ⎥⎢ 3 ⎥
⎢ ⎥ ⎢0 r−1 0 0 0 0 0 0⎥ ⎢ ⎥
. ⎢ ⎥ ⎢ ⎥⎢ v4 ⎥
q = ⎢ ẇ3 ⎥ = ⎢
⎢ ⎥ ⎢0 0 r−1 0 ⎥
0 0 0 0 ⎥⎢⎢v ⎥
⎥ . .
Mq + V m q + F + td = N t (10)
⎢ ẇ4 ⎥ ⎢ ⎥⎢ 1 ⎥
⎢ ⎥ ⎢0 0 0 r−1 0 0 0 0⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥⎢ v2 ⎥
⎢ ḋ1 ⎥ ⎢ ⎥⎢ ⎥ The metrics in (10) are represented as
⎢ ⎥ ⎢0 0 0 0 1 0 0 0 ⎥⎣ v3 ⎦
⎢ ḋ ⎥ ⎢ ⎥
⎢ 2 ⎥ ⎢0 0 0 0 0 1 0 0⎥
⎢ ⎥ ⎢ ⎥ v4
⎣ ḋ3 ⎦ ⎢ ⎥ M = diag{m, m, I, Iw , Iw , Iw , Iw , Id , Id , Id , Id }
⎣0 0 0 0 0 0 1 0⎦
ḋ4
0 0 0 0 0 0 0 1
Vm = 0
= Jv
(6) ⎡c c2 c3 c4 ⎤
1
. 0 0 0 0
where q is defined as the velocity vector in the world coordinate ⎢r r r r ⎥
⎢ s1 s2 s3 s4 ⎥
frame; wi denotes the angular of wheel i; ci and si are represented ⎢ 0 0 0 0⎥
⎢r ⎥
as cos(δi + θ) and sin(δi + θ), respectively. ⎢ r r r ⎥
⎢ T1 T2 T3 T4 0 0 0 0⎥
⎢ ⎥
⎢1 0 0 0 0 0 0 0⎥
⎢ ⎥
2.2 Dynamic model ⎢ 0 0 0 0⎥
N = ⎢0 1 0 0 ⎥
⎢0 0 1 0 0 0 0 0⎥
To derive the dynamic model, the kinetic energy of the mobile robot ⎢ ⎥
⎢0 0 0 1 0 0 0 0⎥
is described as ⎢ ⎥
⎢0 0 0 0 1 0 0 0⎥
⎢ ⎥
⎢0 0 0 0 0 1 0 0⎥
E = Ex + Ey + Eu + Ew + Ed ⎢ ⎥
⎣0 0 0 0 0 0 1 0⎦
1 1 1 2 0 0 0 0 0 0 0 1
= mb v2x + mb v2x + Iu u̇
2 2 2
4   1 4
1 2
F= fdw1 fdw2 fdw3 fdw4 fdd1 fdd2 fdd3 fdd4 T
+ mw v2i + Iw ẇ2i + I ḋ
2 i=1 2 i=1 d
1 1 1 2
= mb v2x + mb v2x + Iu u̇ td = tdw1 tdw2 tdw3 tdw4 tdd1 tdd2 tdd3 tdd4 T
2 2 2
4     1 4
1 2
+ mw v2ix + v2iy + Iw ẇ2i + I ḋ and
2 i=1 2 i=1 d
(7)
1 1 1 2 T
= mb v2x + mb v2x + Iu u̇ t = tw1 tw2 tw3 tw 4 td1 td 2 td3 td 4
2 2 2
m  4 
 2
+ w 4v2x + 4v2y + 4 a2i + b2i u̇ where diag is the abbreviation of diagonal matrix; Ti = r−1[xis(δi) −
2 i=1 yic(δi)]; fdwi and fddi are the friction forces affected on the rolling and
steering wheel i, respectively; tdwi and tddi are the bounded distur-
Iw 
4
I  4
2
+ ẇ2i + d ḋ bances affected on the rolling and steering wheel i, respectively; and
2 i=1 2 i=1 twi and tdi are the torques to roll and steer wheel i, respectively.
Differentiating (6) and substituting into (10), then multiplying the
1 1 1 2 Iw  4
I  4
2
result by J T, (10) is transferred into another form as follows
= mv2x + mv2y + I u̇ + ẇ2i + d ḋ
2 2 2 2 i=1 2 i=1
 .  m (q)v + F
 + td = N(q)
 t
where mb and mw are, respectively, the weight of the robot’s body M(q)v +V (11)
and each wheel; Iθ, Iw and Iδ are the moment of inertia about the .
rotating of the robot’s body, the rolling of the wheel and the steering where M = J T MJ [ <8×8 ,  m = J T MJ [ <8×8 ,
V
of the wheel, respectively; m = mb + 4mw and I = Iθ + 4mw(a 2 + b 2). 
F = J F [ < , td = J td [ <
T 8×1 T 8×1
and N  = J T N [ <8×8 .

J Eng 2014 This is an open access article published by the IET under the Creative Commons
doi: 10.1049/joe.2014.0241 Attribution License (http://creativecommons.org/licenses/by/3.0/)
3
  V
The variables of M,  m and N
 are 3 Trajectory tracking control design
⎡ ⎤ In this section, a trajectory tracking problem of the 4WIS4WID
 11
M  21 M
M  31 M  41 0 0 0 0 mobile robot is discussed and both non-linear kinematic and
⎢M  
M 22 M 32 M  42 0 0 0 0⎥
⎢ 21 ⎥ dynamic SMC controllers for the 4WIS4WID mobile robot are
⎢M  32 M
M  33 M  43 0 0 0 0⎥
⎢ 31 ⎥ designed.
⎢M  42 M
M  43 M  44 0 0 0 0⎥
 = ⎢ 41
M ⎥
⎢ 0 0 0 0 Id 0 0 0⎥
⎢ ⎥
⎢ 0 0 0 0 0 Id 0 0⎥
⎢ ⎥
⎣ 0 0 0 0 0 0 Id 0⎦ 3.1 Problem statement
0 0 0 0 0 0 0 Id
Suppose that a reference robot qd = [ xd (t) yd (t) ud (t) ]T moves
m
 I
 = IWi2 +
M
w
c2 + s2i + 2 , i = 1, 2, 3, 4 along a time-varying trajectory. The control objective is to let the
.
ii 16 i r robot track q d and the tracking error between the reference robot
  and the actual robot approach zero. On the basis of this objective,
 ij = IWj Wi + m cj ci + sj si
M the main task is to design a control law for the 4WIS4WID mobile
16 robot to ensure the actual robot is able to track the reference one.
i = 2, 3, 4; j = 1, 2, 3, 4; i = j
⎡ ⎤
V 11 V 21 V 31 V 41 0 0 0 0
⎢ V 21 3.2 Kinematic controller design
⎢ V 22 V 32 V 42 0 0 0 0⎥⎥
⎢ V 31 V 23 V 33 V 43 0 0 0 0⎥ Before designing the controllers, the vector of the tracking error is
⎢ ⎥
⎢ V 32 V 24 V 34 V 44 0 0 0 0⎥ defined as follows

Vm = ⎢ ⎥
⎢0 0 0 0 0 0 0 0⎥
⎢ ⎥
⎢0 0 0 0 0 0 0 0⎥ ⎡ ⎤
⎢ ⎥
⎣0 0 0 0 0 0 0 0⎦
xe
⎢ ⎥
0 0 0 0 0 0 0 0 qe = ⎣ ye ⎦ = R(u) · q̃e
  ue
V ij = IWj Ẇ i + m cj ċi + sj ṡi /16 ⎡ ⎤⎡ ⎤ (13)
cos u sin u 0 xd − x
⎢ ⎥⎢ ⎥
= ⎣ − sin u cos u 0 ⎦⎣ yd − y ⎦
i = 1, 2, 3, 4; j = 1, 2, 3, 4
0 0 1 ud − u
⎡ ⎤
 11
N  21
N  31
N  41
N 0 0 0 0
⎢N 21  22
N  32
N  42
N 0 0 0 0⎥
⎢ ⎥ and the tracking error dynamic is denoted as
⎢N  23
N  33
N  43
N 0 0 0 0⎥
⎢ 31 ⎥
⎢N  24
N  34
N  44
N 0 0 0 0⎥
 = ⎢ 32
N ⎥
⎢0 0 0 0 1 0 0 0⎥
⎢ ⎥ . . .
q e = q d − q = ẋe
T
⎢0 0 0 0 0 1 0 0⎥ ẏe u̇e (14)
⎢ ⎥
⎣0 0 0 0 0 0 1 0⎦
.
0 0 0 0 0 0 0 1 where R(θ) is the rotation matrix. After some computing, q e can be
derived as follows

 ii = Wi (yi c(di ) + xi s(di )) + si + ci + 1 ,

2 2
⎡ ⎤
N i = 1, 2, 3, 4 ye u̇ + vxd − vx
r 4r r ⎢ ⎥
C q̇e = ⎣ −xe u̇ + vyd − vy ⎦
and vd − v
⎡ ⎤
W (y c(d ) + xj s(dj )) si sj + ci cj 1 4
1 4
 ij = i j j
N + ⎢ e
y u̇ + (v cos d ) − (v cos d i ⎥
)
⎢ 4 i=1 id id
4 i=1 i ⎥ (15)
r 4r ⎢ ⎥
i = 1, 2, 3, 4; j = 1, 2, 3, 4; i = j ⎢   ⎥
⎢ 1 4
1 4
⎥
= ⎢ −xe u̇ + (v sin d ) − (v sin d ) ⎥
⎢ 4 id id
4 i i ⎥
⎢ i=1 i=1 ⎥
⎢ ⎥
 is a symmetric positive-definite matrix. ⎣ 4  4 ⎦
Property 1: M (vid Wid ) − (vi Wi )
i=1 i=1
˙ − 2V
Property 2: M  m is the skew symmetric.

Premultiply (11) by M −1 and regarding the friction forces F
 and where vid = v2xid + v2yid , δid = arctan2(vyid, vxid), Wid =
disturbances torques td as the system uncertainties and distur-
r  
r 2
r 2 
bances, then the dynamic equation of the 4WIS4WID can be −ywi c(did ) + xwi s(did ) / 4 xwi + 4 ywi , vxid = vxd − yrwi vd
r

described as and vyid = vyd + xrwi vd .

The kinematic tracking control law for the 4WIS4WID mobile robot
. 
v (t) = − M
−1  
V m v(t) + M
−1  −1 F
 t (t) − M
N  −1 td
 −M is derived as follows. Define the Lyapunov function candidate as
; −Av(t) + Bt(t) + d(t) (12)
1
2 
 −1 V  −1 N
 m, B = M  −1 F
 and d = −M  −1 td .
 −M Vk = x + y2e + u2e ≥ 0 (16)
where A = M 2 e

This is an open access article published by the IET under the Creative Commons J Eng 2014
Attribution License (http://creativecommons.org/licenses/by/3.0/) doi: 10.1049/joe.2014.0241
4
Taking the time derivative of Vk implies Although (z1, z2) = (−1, π), then

V̇ k = xe ẋe + ye ẏe + ue u̇e 

4
(vid cos did − vi cos di )
 
1 1
4 4 i=1
= xe (v cos did ) − (v cos di )    
4 i=1 id 4 i=1 i 
4
ai
= vid cos did − − a2i + b2i · −  (24)
  a2i + b2i
1 1
i=1
4 4 (17)
+ ye (v sin did ) − (v sin di ) 
4
4 i=1 id 4 i=1 i
= ( − kx xe + ku ywi ue ) = −4kx xe
 
4 
4 i=1
+ ue (vid Wid ) − (vi Wi )
i=1 i=1 
4
(vid sin did − vi sin di )
Let the control laws be defined as i=1
   

4
bi
 = vid sin did − − ai + bi · − 
2 2 (25)
vic = z1 a2i + b2i (18) i=1 a2i + b2i

4
dic = z2 + arctan 2(bi , ai ) (19) = (−ky yd − ku xwi ue ) = −4ky ye
i=1

where ai = vid cos δid + kxxe − kθywiθe and bi = vid sin δid + kyye + and
kθxwiθe; kx, ky and kθ are all positive constants; both z1 and z2 are
defined as

4 
4
(vid Wid ) − (vi Wi )

(+1, 0), |did − di | ≤ p/2 i=1 i=1
 ⎞
(z1 , z2 ) = (20) ⎛

(−1, p), otherwise xrwi −ky ye − ku xrwi ue
1 4
⎝−
yrwi −kx xe + ku yrwi ue
⎠
= +
By substituting both (18) and (19) into (17), one can obtain the 4 i=1 (xrwi )2 + (yrwi )2 (xrwi )2 + (yrwi )2
results as follows.
Although (z1, z2) = ( + 1, 0), then = −ku ue
(26)

4
(vid cos did − vi cos di ) According to the above results, one can find (17) becomes
i=1
  

4
ai V̇ k = −4kx x2e − 4ky y2e − ku u2e (27)
= vid cos did − a2i + b2i  (21)
i=1 a2i + b2i
Note that (19) is the control signal of steering angle. To control the

4
= (−kx xe + ku ywi ue ) = −4kx xe steering angle, the control input of the steering angular velocity of
i=1 wheel i for the 4WIS4WID mobile robot is defined as


4
ḋi = ḋic + kdi die (28)
(vid sin did − vi sin di )
i=1
   where kdi is a positive constant of wheel i.

4
bi
= vid sin did − ai + bi · 
2 2 (22) Define the Lyapunov function candidate for the four steering
i=1 a2i + b2i angles of 4WIS4WID mobile robot as

4
= (−ky ye − ku xwi ue ) = −4ky ye 1 4
i=1 Vd = d2 ≥ 0 (29)
2 i=1 ie


4 
4
where δie = δic − δi. Taking the time derivative of Vδ and consider-
(vid Wid ) − (vi Wi ) ing (28) imply
i=1 i=1
⎛  

 xr −k y − k xr u ⎞
1 4
⎝−
yrwi −kx xe + ku yrwi ue wi y e u wi e
⎠

4 
4

= + V̇ d = die ḋie = die ḋic − ḋi (30)
4 i=1 (xrwi )2 + (yrwi )2 (xrwi )2 + (yrwi )2 i=1 i=1

4  
1 k u (yr )2 + ku ue (xrwi )2
= − u e wi By choosing appropriate gains of kx, ky, kθ and kdi , V̇ k and V̇ d can be
4 i=1 (xrwi )2 + (yrwi )2 guaranteed less than zero for all t ≥ 0 while qe ≠ 0 and δie ≠ 0.
Therefore, using the kinematic control laws, (18), (19) and (28)
= −ku ue
to control the 4WIS4WID mobile robot, the tracking error (13) is
(23) asymptomatically stable.

J Eng 2014 This is an open access article published by the IET under the Creative Commons
doi: 10.1049/joe.2014.0241 Attribution License (http://creativecommons.org/licenses/by/3.0/)
5
Fig. 3 Trajectory tracking control scheme of the 4WIS4WID mobile robot

3.3 Dynamic controller design bounded uncertain parameters and the bounded disturbances d(t)
are existing, (12) is rewritten as
In this section, a dynamic tracking controller, SMC, will be
designed for the 4WIS4WID mobile robot. Define the velocity
. ˆ
error as v (t) ; −Av(t) + B̂t(t) + d(t) (35)
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
ew1 v1c v1
⎢ ew ⎥ ⎢ v2c ⎥ ⎢ v2 ⎥ where  = A + DA and B̂ = B + DB. DA [ <8×8 and
⎢ 2 ⎥ ⎢ ⎥ ⎢ ⎥ DB [ <8×8 are the bounded system uncertainties which are gener-
⎢ ew ⎥ ⎢ v3c ⎥ ⎢ v3 ⎥
⎢ 3⎥ ⎢ ⎥ ⎢ ⎥ ated by perturbations of system parameters. By considering the un-
⎢ ew ⎥ ⎢ v4c ⎥ ⎢ v4 ⎥
e=⎢ 4 ⎥ = ⎢ ⎥ ⎢ ⎥
⎢ ed ⎥ ⎢ ḋ1c ⎥ − ⎢ ḋ1 ⎥ = vc (t) − v(t) (31) certainties and disturbances, (33) is rewritten as
⎢ 1⎥ ⎢ ⎥ ⎢ ⎥
⎢ ed ⎥ ⎢ ḋ2c ⎥ ⎢ ḋ2 ⎥
⎢ 2⎥ ⎢ ⎥ ⎢ ⎥ . . ˆ
⎣ ed ⎦ ⎣ ḋ ⎦ ⎣ ḋ ⎦ S (t) = v c (t) + Av(t) − B̂t(t) + d(t) + le(t)
3 3c 3
.
ed4 ḋ4c ḋ4 = v c (t) + (A + DA)v(t) − (B + DB)t(t)
+ d(t) + le(t)
and by choosing a PI-like sliding surface as (36)
. ˆ
= v c (t) + Av(t) − B̂t(t) + le(t) + (DAU(t)
T
S(t) = s1 s2 s3 s4 s5 s6 s7 s8 − DBt(t) + d(t))
t .
(32) = v c (t) + Av(t) − Bt(t) + le(t) + D(t)
= e + l e(t) dt
0
where D(t) = DAU(t) − DBt(t) + d(t) [ <8×1 , and the element
where l [ < is a diagonal positive-definite matrix and l is
8×8
Di of D(t) is defined as |Di | ≤ D i , where D
 i is a bound positive
selected to stabilise the sliding surface. Although the sliding constant of Di.
surface S(t) approaches zero, the tracking errors will also Therefore, in order to ensure the system is stable, an auxiliary
converge to zero. Therefore, the objective is to design a controller discontinuous control law is added and defined as
to let the system errors reach on the sliding surface. To derivate
the equivalent control laws teq(t), both the uncertainties and distur-
bances are assumed to be zero. The time derivative of (32) is
tr (t) = B−1 KSat(S) (37)

. . . .
S (t) = e + le = v c (t) + Av (t) − Bt(t) + le(t) (33) where K [ <8×8 is the switch gain matrix concerned with upper
bounds of uncertainties and K is positive-definite and diagonal
.
Given S(t) = 0, then the equivalent control law teq(t) can be
obtained as follows ⎡ ⎤
sat(s1 , 1)
⎢ sat(s2 , 1) ⎥
. ⎢ ⎥
teq (t) = B−1 (v c (t) + Av(t) + le(t)) (34) ⎢ sat(s3 , 1) ⎥
⎢ ⎥
⎢ sat(s4 , 1) ⎥
Sat(S, 1) = ⎢
⎢ sat(s5 ,
⎥
Once the teq(t) has been determined, one should derive the auxiliary
⎢ 1) ⎥
⎥
discontinuous control law tr(t) to overcome the uncertainties and ⎢ sat(s6 , 1) ⎥
⎢ ⎥
disturbances of the mobile robot. Suppose that A and B have the ⎣ sat(s , 1) ⎦
7
sat(s8 , 1)
Table 1 Parameters used in simulations
ε is a small positive constant and sat is the saturation function which
Description Symbol Value Unit is adopted to suppress the chattering behaviour. Sat is represented as

coordinate of wheel 1 (xrw1 , yrw1 ) (0.2, 0.25) m 

(xrw2 , yrw2 ) sign(si ), |si | . 1
coordinate of wheel 2 (−0.2, 0.25) m sat(si , 1) =
coordinate of wheel 3 (xrw3 , yrw3 ) (−0.2, −0.25) m si /1, |si | ≤ 0
coordinate of wheel 4 (xrw4 , yrw4 ) (0.2, −0.25) m
wheel radius r 0.05 m
robot’s body mass mb 26 kg Table 2 Control parameters
vehicle inertia moment Iz 1.5 kgm2
wheel mass mw 1 kg Description Symbol Value
wheel rolling torque limitation twi 5 Nm
wheel rolling inertia moment Iw 0.003 kgm2 gains of (18) and (19) kx, ky, kθ 4, 4, 3
wheel steering torque limitation tdi 1 Nm gains of (28) kdi 5
wheel steering inertia moment Iδ 0.005 kgm2 integral gains of (32) λi 5
limit of steering angle δi (−π, π) rad gains of (37) ki 10

This is an open access article published by the IET under the Creative Commons J Eng 2014
Attribution License (http://creativecommons.org/licenses/by/3.0/) doi: 10.1049/joe.2014.0241
6
Fig. 4 Simulation results of the straight trajectory of the four initial positions
a x–y plot of the 4WIS4WID mobile robot
b–d Tracking errors of xe, ye and θe

The element kii in K is designed as By taking the time derivative of VSMC

.
 i + hi ,
ki = D i = 1, 2, 3, . . . , 8 (38) V̇ SMC = S(t)T S(t)
.
= S(t)T (v c (t) + Av(t)
where ηi > 0. − Bt(t) + le(t) + D(t))
Combining teq(t) and tr(t), the consequential SMC control law (41)
for the mobile robot is = S(t)T (−K · Sat(S)

8

t(t) = teq (t) + tr (t) + D(t)) ≤ (−hi |si (t)|)

.  (39) i=1
= B−1 v c (t) + Av(t) + le(t) + K · Sat(S)
Thus, VSMC is bounded for all time despite the uncertainties existed.
To show that the dynamic control law is stable for the tracking The influence created by the uncertainties and disturbances could be
control, the Lyapunov function candidate is defined as reduced. According to the results of (27), (30) and (41), using the
control laws (18), (19), (28) and (39), one can find that the
4WIS4WID mobile robot will converge exponentially to the refer-
1
VSMC = S T (t)S(t) ≥ 0 (40) ence trajectory. All of the dynamic errors satisfy qe → 0 and e → 0
2 as t → ∞.

Fig. 5 Steering angle and rolling torque of the four wheels of the initial position d
a Steering angle
b Rolling torque

J Eng 2014 This is an open access article published by the IET under the Creative Commons
doi: 10.1049/joe.2014.0241 Attribution License (http://creativecommons.org/licenses/by/3.0/)
7
Fig. 6 Simulation results of Case (I)
a x–y plot of the 4WIS4WID mobile robot
b–d Tracking errors of xe, ye and θe
e Steering angle of four wheels
f Rolling torques of four wheels

The whole control scheme is shown in Fig. 3, where the inner and 4.1 Test by a straight trajectory
outer loops are used to control the dynamics and kinematics of
The straight trajectory utilised to test the control scheme is gener-
4WIS4WID mobile robot, respectively. The kinematic control law
ated as follows
is implemented by equations (18), (19) and (28) and the dynamic
control law is realised by (39).
Xd (t) = t
4 Simulations Yd (t) = t


In this section, three computer simulations are performed to dem- ud (t) = arctan 2 Ẏ d (t), Ẋ d (t)
onstrate the effectiveness and feasibility of the proposed control
schemes for the 4WIS4WID mobile robot. The following simula-
tions will define the trajectory and the initial position (X, Y, θ) of Four initial positions are selected to track the trajectory, which
the robot. All the parameters of the mobile robot are listed in are a(X, Y, θ) = (0, 0, 0), b(X, Y, θ) = (−1, 0, 180°), c(X, Y, θ) =
Table 1. Note that the limit of the steering angle in Table 1 (−1, −1, −90°) and d(X, Y, θ) = (1, 2, 0). Fig. 4a shows the
means the steering motors cannot turn the wheel infinitely in x–y plot of the trajectory tracking result of the four initial posi-
the same direction because the wires used to transmit signals tions. The tracking errors of the four initial positions of the
and power to the wheels might be twisted to breaking. They straight trajectory are shown in Figs. 4b–d. Figs. 5a and b rep-
can only turn clockwise or counterclockwise in a half circle. resent the four-wheel steering angles and torques of the initial
Table 2 lists all the control parameters. In Section 4.1, a straight position (X, Y, θ) = (10, 3, 90°), respectively. One can find that
trajectory is used to test the feasibility of the control scheme. as time increases, the tracking errors of this system are conver-
After the test in Section 4.2, two curvilinear trajectories are ging to zero. The simulation results demonstrate that the
selected to demonstrate the performances using the presented 4WIS4WID mobile robot can be stably driven using the
control scheme. control scheme.

This is an open access article published by the IET under the Creative Commons J Eng 2014
Attribution License (http://creativecommons.org/licenses/by/3.0/) doi: 10.1049/joe.2014.0241
8
Fig. 7 Simulation results of Case (II)
a x–y plot of the 4WIS4WID mobile robot
b–d Tracking errors of xe, ye and θe
e Steering angle of four wheels
f Rolling torques of four wheels

4.2 Curvilinear trajectory simulations For Case (II), a trajectory is generated by

In this section, two trajectories are utilised to demonstrate the results
and performances of the proposed control schemes.
Xd (t) = 2 sin (1.2t) + 2 cos (0.6t)
For Case (I), a circle trajectory is generated by
Yd (t) = 2 sin (0.6t)

+ 2 cos (1.2t) 
ud (t) = arctan 2 Ẏ d (t), Ẋ d (t)
Xd (t) = 3 cos (0.5t)
Yd (t) = 3 sin (0.5t) The initial position of the robot is at (X, Y, θ) = (2, 1, 90°). The

 random uncertainties and disturbances given by the following equa-
ud (t) = arctan 2 Ẏ d (t), Ẋ d (t)
tion are added

If the initial position of the robot is at (X, Y, θ) = (2, 2, 0) and the T

D(t) = 5 sin (10t) 1 2 2 1 1 2 2 1
uncertainties and disturbances are given and bounded by

T
The simulation results are shown in Fig. 7, which includes the track-
D(t) = 5 sin (10t) 1 1 1 1 1 1 1 1 ing trajectory, tracking errors, four-wheel steering angles and
torques.
Both simulation results demonstrate that the proposed trajectory
Simulation results of tracking trajectory control are shown in Fig. 6, tracking control scheme can make the 4WIS4WID mobile robot
where tracking errors, four-wheel steering angles and torques are track the desired trajectories successfully and stably although the
given. uncertainties and disturbances exist.

J Eng 2014 This is an open access article published by the IET under the Creative Commons
doi: 10.1049/joe.2014.0241 Attribution License (http://creativecommons.org/licenses/by/3.0/)
9
Fig. 8 Simulation results of Case (I) by the PI dynamic controller
a x–y plot of the 4WIS4WID mobile robot
b–d Tracking errors of xe, ye and θe
e Steering angle of four wheels
f Rolling torques of four wheels

4.3 Comparative study the parameters of the ICR, so the discontinuous condition when
the 4WIS4WID mobile robot moves straight does not encounter.
The purpose of this section is to demonstrate the superiority of
To obtain a better performance for the trajectory tracking control,
the proposed controller, where the kinematic controller is un-
the proposed kinematic control scheme is derived by the non-linear
changed and the SMC dynamic controller is replaced by the PI
control and the dynamic control scheme is established by the SMC
controller. The control law (39) is rewritten as
t control technique. Once all of the controller parameters are assigned
t(t) = K P e + K I 0 e(t) dt, where KP = diag{5, 5, 5, 5, 2, 2, 2, 2}
correctly, the control law can reduce the trajectory tracking errors
and KI = diag{0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5}. For the
and suppress the affection of the system uncertainties and external
Case (I), the simulation results of the PI dynamic controller are
disturbances efficiently. Using the Lyapunov stability theory, the
depicted in Fig. 8. By comparing Fig. 6 with Fig. 8, one can
stabilities of the kinematic and dynamic controllers have been
find that the convergent speed of the tracking errors in Fig. 8
proved. The simulation results demonstrate that whether the
is slower than that in Fig. 6. Furthermore, the tracking errors of
desired path is a straight line or curvilinear path, the 4WIS4WID
ye and θe have some tremors in Fig. 8 and these phenomena
mobile robot can track the path successfully. In comparison with
are influenced by D(t). It shows that using the SMC is better
the PI dynamic control, the SMC control is indeed effective in sup-
than PI control to suppress the uncertainties and disturbances in
pressing the influences of uncertainties and disturbances for the tra-
this research.
jectory tracking control of 4WIS4WID mobile robot.

5 Conclusions
6 Acknowledgments
This paper has proposed new kinematic and the dynamic models for
4WIS4WID mobile robot. These two models can be used to This work supported in part by the Ministry of Science and
examine some new control schemes and/or other applications for Technology of the Republic of China under grant
the 4WIS4WID mobile robot. On the basis of these models, the NSC101-2221-E-006-193-MY3, in part by the Ministry of
dynamic trajectory tracking control for the 4WIS4WID mobile Education, and the aim for the Top University Project to the
robot has been presented. The control scheme does not include National Cheng Kung University (NCKU) are greatly appreciated.

This is an open access article published by the IET under the Creative Commons J Eng 2014
Attribution License (http://creativecommons.org/licenses/by/3.0/) doi: 10.1049/joe.2014.0241
10
7 References [17] Lins Barreto S.J.C., Scolari Conceicao A.G., Dorea C.E.T., Martinez
L., de Pieri E.R.: ‘Design and implementation of model-predictive
[1] Yeh Y.C., Li T.H.S., Chen C.Y.: ‘Adaptive fuzzy sliding-mode control with friction compensation on an omnidirectional
control of dynamic model based car-like mobile robot’, mobile robot’, IEEE/ASME Trans. Mechatronics, 2014, 19, (2),
Int. J. Fuzzy Syst., 2009, 11, (4), pp. 272–286 pp. 467–476
[2] Egerstedt M., Hu X., Stotsky A.: ‘Control of mobile platforms using a [18] Huang H.C.: ‘SoPC-based parallel ACO algorithm and its application
virtual vehicle approach’, IEEE Trans. Autom. Control, 2001, 46, to optimal motion controller design for intelligent omnidirectional
(11), pp. 1777–1782 mobile robots’, IEEE Trans. Ind. Inform., 2013, 9, (4),
[3] Hwang C.L., Chang N.W.: ‘Fuzzy decentralized sliding-mode control pp. 1828–1835
of a car-like mobile robot in distributed sensor-network spaces’, IEEE [19] Dietrich A., Wimbock T., Albu-Schaffer A., Hirzinger G.: ‘Reactive
Trans. Fuzzy Syst., 2008, 16, (1), pp. 97–109 whole-body control: dynamic mobile manipulation using a large
[4] Kumar U., Sukavanam N.: ‘Backstepping based trajectory tracking number of actuated degrees of freedom’, IEEE Robot. Autom.
control of a four wheeled mobile robot’, Int. J. Adv. Robot. Syst., Mag., 2012, 19, (2), pp. 20–33
2008, 5, (4), pp. 403–410 [20] Qian H., Lam T.L., Li W., Xia C., Xu Y.: ‘System and design of an
[5] Nishihara T.O., Kumamoto H.: ‘Automatic path-tracking controller omni-directional vehicle’. Proc. IEEE Int. Conf. on Robotics and
of a four-wheel steering vehicle’, Veh. Syst. Dyn., 2009, 47, (10), Biomimetics, Bangkok, Thailand, 2009, pp. 389–394
pp. 1205–1227 [21] Connette C.P., Pott A., Hägele M., Verl A.: ‘Control of an
[6] Song Y.D., Chen H.N., Li D.Y.: ‘Virtual-point-based fault-tolerant pseudo-omnidirectional, non-holonomic, mobile robot based on an
lateral and longitudinal control of 4 W-steering vehicles’, IEEE ICM representation in spherical coordinates’. Proc. IEEE Int. Conf.
Trans. Intell. Transp. Syst., 2011, 12, (4), pp. 1343–1351 on Decision and Control, Cancun, Mexico, 2008, pp. 4976–4983
[7] Ruiz U., Murrieta-Cid R., Marroquin J.L.: ‘Time-optimal motion [22] Clavien L., Lauria M., Michaud F.: ‘Instantaneous centre of rotation
strategies for capturing an omnidirectional evader using a differential estimation of an omnidirectional mobile robot’. Proc. IEEE Int.
drive robot’, IEEE Trans. Robot., 2013, 29, (5), pp. 1180–1196 Conf. on Robotics and Automation, Anchorage, AK, USA, 2010,
[8] Ailon A., Zohar I.: ‘Control strategies for driving a group of pp. 5435–5440
nonholonomic kinematic mobile robots in formation along a time- [23] Ploeg J., van der Knaap A.C.M., Verburg D.J.: ‘ATS/AGV-design,
parameterized path’, IEEE/ASME Trans. Mechatronics, 2012, 17, implementation and evaluation of a high performance AGV’. Proc.
(2), pp. 326–336 IEEE Intelligent Vehicles Symp., 2002, 1, pp. 127–134
[9] Zhang Q., Lapierre L., Xiang X.: ‘Distributed control of coordinated [24] Lin C.J., Hsiao S.M., Wang Y.H., Yeh C.H., Huang C.F., Li T.H.S.:
path tracking for networked nonholonomic mobile vehicles’, IEEE ‘Design and implementation of a 4WS4WD mobile robot and its
Trans. Ind. Inform., 2013, 9, (1), pp. 472–484 control applications’. Proc. IEEE Int. Conf. on Systems Science
[10] Park B.S., Yoo S.J., Park J.B., Choi Y.H.: ‘A bioinspired and Engineering, Budapest, Hungary, 2013, pp. 235–240
neurodynamics-based approach to tracking control of mobile [25] Lam T.L., Qian H., Xu Y.: ‘Omnidirectional steering interface and
robots’, IEEE Trans. Ind. Electron., 2012, 59, (8), pp. 3211–3220
control for a four-wheel independent steering vehicle’, IEEE/ASME
[11] Censi A., Franchi A., Marchionni L., Oriolo G.: ‘Simultaneous cali-
Trans. Mechatronics, 2010, 15, (3), pp. 329–338
bration of odometry and sensor parameters for mobile robots’, IEEE
[26] Jiang S.Y., Song K.T.: ‘Differential flatness-based motion control of a
Trans. Robot., 2013, 29, (2), pp. 475–492
steer-and-drive omnidirectional mobile robot’. Proc. IEEE Int. Conf.
[12] Yi J., Wang H., Zhang J., Song D., Jayasuriya S., Liu J.: ‘Kinematic
modeling and analysis of skid-steered mobile robots with applications on Mechatronics and Automation 2013, Takamatsu, Japan, August
to low-cost inertial-measurement-unit-based motion estimation’, 2013, pp. 1167–1172
IEEE Trans. Robot., 2009, 25, (5), pp. 1087–1097 [27] Selekwa M.F., Nistler J.R.: ‘Path tracking control of four wheel inde-
[13] Yi J., Song D., Zhang J., Goodwin Z.: ‘Adaptive trajectory tracking pendently steered ground robotic vehicles’. Proc. IEEE Conf. on
control of skid-steered mobile robots’. Proc. IEEE Int. Conf. on Decision and Control and European Control Conf. (CDC-ECC),
Robotics and Automation, 2007, pp. 2605–2610 Orlando, FL, USA, December 2011, pp. 6355–6360
[14] Morales J., Martinez J.L., Mandow A., Garcia-Cerezo A., Pedraza S.: [28] Hwang C.L., Wu H.M.: ‘Trajectory tracking of a mobile robot
‘Power consumption modeling of skid-steer tracked mobile robots on with frictions and uncertainties using hierarchical sliding-mode
rigid terrain’, IEEE Trans. Robot., 2009, 25, (5), pp. 1098–1108 under-actuated control’, IET Control Theory Appl., 2013, 7, (7),
[15] Kim K.B., Kim B.K.: ‘Minimum-time trajectory for three-wheeled pp. 952–965
omnidirectional mobile robots following a bounded-curvature path [29] Jean J.H., Lian F.L.: ‘Robust visual servo control of a mobile robot
with a referenced heading profile’, IEEE Trans. Robot., 2011, 27, for object tracking using shape parameters’, IEEE Trans. Control
(4), pp. 800–808 Syst. Technol., 2012, 20, (6), pp. 1461–1472
[16] Indiveri G.: ‘Swedish wheeled omnidirectional mobile robots: [30] Rubagotti M., Della Vedova M.L., Ferrara A.: ‘Time-optimal sliding-
kinematics analysis and control’, IEEE Trans. Robot., 2009, 25, mode control of a mobile robot in a dynamic environment’, IET
(1), pp. 164–171 Control Theory Appl., 2011, 5, (16), pp. 1916–1924

J Eng 2014 This is an open access article published by the IET under the Creative Commons
doi: 10.1049/joe.2014.0241 Attribution License (http://creativecommons.org/licenses/by/3.0/)
11