This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2021.3050722, IEEE
Access

Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.Doi Number

Integrated Chassis Control with Four-Wheel

Independent Steering under Constraint on
Front Slip Angles
Seongjin Yim1, Member, IEEE
1
Research Center for Electrical and Information Technology, Seoul National University of Science and Technology, Nowon-gu, Seoul, 01811,
Republic of Korea

Corresponding author: Seongjin Yim (e-mail: [email protected]).

This work supported by Research Program supported by Seoul National University of Science and Technology.

ABSTRACT This paper presents a method to design an integrated chassis controller with four-wheel
independent steering (4WIS) under the constraint on front slip angles for electric vehicles (EVs) adopting
in-wheel motor (IWM) driving system. To improve lateral stability and maneuverability of a vehicle, direct
yaw moment control strategy is adopted. A control allocation method is adopted to distribute control yaw
moment into tire forces, generated by 4WIS. If corrective steering angles of 4WIS are added to front
steering angles generated by a driver, it can deteriorate control performance because the lateral tire force of
front wheels easily saturated and it causes loss of required yaw moment needed to stabilize a vehicle. To
cope with the problem, it is necessary to impose constraints on front slip angles. To compensate the loss of
control yaw moment caused by the constraint on front slip angles, a constrained control allocation method
is presented. Simulation on driving simulation tool, CarSim®, shows that the proposed integrated chassis
controller is capable of maintaining lateral stability and maneuverability without performance deterioration
under the constraint on the front slip angles.

INDEX TERMS Integrated chassis control (ICC), Four-wheel independent steering (4WIS), In-wheel
motor system, Independent steering system, Control allocation, Constraint on front slip angles

I. INTRODUCTION TVD has implemented with center or active differentials [7].

Over the last decade, in-wheel motors (IWMs) have been On the contrary, 4WID is quite natural for IWMs. In this
developed by researchers and automotive industry [1]. paper, 4WID and TVD are synonyms. Besides 4WIB and
Typical example of it is the skateboard platform or HyWire 4WID in IWM, 4-wheel independent steering (4WIS)
developed by GM [2]. Recently, this was revived by Canoo becomes available if each IWM has a steering actuator.
[3]. In view of vehicle stability control, IWM has a function 4WIS is the most general architecture of steering actuators.
of 4-wheel independent drive (4WID) and braking (4WIB), For example, 4WIS can be shrunk into active front steering
which consist of traction motor with reduction gear and (AFS), rear wheel steering (RWS), front wheel independent
electro-mechanical brake (EMB) or electronic wedge brake steering (FWIS) and rear wheel independent steering
(EWB), respectively [4]. This function can enhance control (RWIS) [8]. 4-wheel steering (4WS) is the combination of
performance better. AFS and RWS, and 4WIS is just the combination of FWIS
4WIB can be regarded as an electronic stability control and RWIS.
(ESC), which is based on hydraulic brake system, and has a The combination of 4WIB, 4WID and 4WIS in EVs with
function of independent or differential braking [5]. The IWMs is the ideal actuator combination in view of vehicle
difference between 4WIB and ESC is that 4WIB makes use stability control. To make fully use of actuator
of electronic brakes such as EMB and EWB. In this paper, combinations for vehicle stability control, it is necessary to
4WIB and ESC are synonyms. 4WID can be regarded as coordinate the tire forces generated by these actuators. This
torque vectoring device (TVD), which can generate is called control allocation problem or integrated chassis
independent traction torque at each wheel [6]. Generally,

VOLUME XX, 2017 1

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2021.3050722, IEEE
Access

control (ICC) [6,9-25]. There have been several researches uncertainties [27]. In this paper, a time-delay control
for control allocation with multiple actuators in ICC. These methodology is adopted to design a robust vehicle stability
researches formulated the control allocation problem as an controller.
optimization one, and applied several algorithms to solve it This paper investigates the integrated chassis control
such as weighted pseudo-inverse based control allocation with 4WIB, 4WID and 4WIS under the constraint on the
(WPCA) [9,11-16] and a fixed-point control allocation front slip angles for EVs with IWMs. The control
method [10], etc. Among these, WPCA can solve the performance measures considered in this paper are
problem in real time because only algebraic computation is maneuverability and lateral stability. For control allocation
needed to find an optimum solution. Moreover, it can easily with 4WIB, 4WID and 4WIS, an integrated chassis
handle constraints imposed on actuators. So, WPCA is controller has been proposed in the previous works [19].
adopted for control allocation in this paper. The controller has been composed of the upper-level and
The actuators that have been used for control allocation lower-level controllers in this research. The upper one
are ESC, AFS/ARS/4WS and TVD. However, there have computes a control yaw moment for vehicle stability based
been little researches for control allocation with 4WIB, on the time-delay control methodology. The control yaw
4WID and 4WIS in ICC. In the previous work, a sequential moment, calculated by the upper one, is realized by tire
quadratic programming (SQP) and a MPC have been forces, generated by several single actuators or actuator
applied for control allocation with these actuators [17-19]. combinations such as 4WIB, 4WID and 4WIS. WPCA is
The typical research for ICC with 4WIB, 4WID and 4WIS selected as a method for distributing the control yaw
are the work of Yim. In the research, WPCA was applied to moment into tire forces of 4WIB, 4WID and 4WIS. To
solve the control allocation problem with several cope with the saturation of front lateral tire forces or
combinations of 4WIB, 4WID and 4WIS [20]. physical constraint on front slip angles, CWPCA is applied
Among 4WIS, 4WIB and 4WID, 4WIS is preferred [16,22]. This is the contribution of this paper, which is to
because it consumes a smaller amount of energy compared propose the control allocation method with the multiple
to 4WIB and 4WID. Moreover, 4WIS does slightly reduce actuators, 4WIS, 4WIB and 4WID, under the constraint on
the vehicle speed. However, the front lateral tire forces can front slip angles. To investigate and compare the control
be easily saturated by 4WIS, especially FWIS, because the performance in terms of 4WIS under the constraint on front
steering angle of FWIS is added to that of driver. Moreover, slip angles, simulation has been conducted on driving
there is physical constraint on the magnitude of lateral tire simulation tool, CarSim.
force caused by a friction circle. Under the constraints, This paper consists of four sections. In Section II, the
FWIS cannot create the lateral tire forces, needed to configuration of the proposed ICC, i.e., the upper-level and
generate the desired yaw moment [16,20,21]. So, the slip lower-level controllers, is explained. The control allocation
angles generated by FWIS should be restricted to a certain method, needed to determine the steering angles, traction
value which gives the maximum lateral tire forces. This torque and brake pressure of 4WIS, 4WID and 4WIB, is
causes the loss of the desired yaw moment, which results in introduced in Section II. In the section, a method needed to
the deterioration of control performance. So, the loss of the handle the constraint on front slip angles is also proposed.
desired yaw moment is to be compensated. For the purpose, Simulation is conducted and simulation results are analyzed
a constrained WPCA (CWPCA) and WPCA based in Section III. In the last section, Section IV, the conclusion
constrained unified chassis control (CUCC) were proposed of this research is given.
in the previous works [16,21,22]. These researches have
restricted on the magnitude of the lateral tire forces of front II. DESIGN OF INTEGRATGED CHASSIS CONTROLLER
wheels. Besides it, the longitudinal tire force was
considered at control allocation stage in the previous work A. DESIGN OF UPPER-LEVEL CONTROLLER
[23-25]. In this paper, CWPCA is adopted for control In the upper-level controller, a desired yaw moment, i.e.,
allocation with RWIS, 4WIB and 4WID in order to control yaw moment is calculated by a yaw moment
compensate the loss of the control yaw moment. controller. For the purpose, a vehicle model and controller
Generally, vehicle systems have uncertain and time- design methodology are needed. In this paper, 2-DOF
varying parameters. These make performance of vehicle bicycle model and time-delay control are adopted to design
control system be deteriorated. So, it is necessary to design the yaw moment controller.
a controller which is robust to parameter variations. In the Fig. 1 is a representing picture of 2-DOF model. In this
area of vehicle stability control, several approaches have model, ‘2 DOF’ stands for yaw and lateral motions [28]. It
been done for the purpose. Typical approach for designing is assumed that the longitudinal velocity vx is constant. So,
robust vehicle stability control is to apply the robust state variables in the model are yaw rate, , and side-slip
controller design methodologies such as H loop-shaping angle, . vy is the lateral velocity, which is included in  .
and -synthesis [18,19,26]. Another approach is to use the With these variables, the differential equation of motion for
Takagi-Sugeno fuzzy model in order to capture parameter the model is given in (1). In the equation, m and Iz are the

VOLUME XX, 2017 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2021.3050722, IEEE
Access

vehicle total mass and the yaw moment of inertia, estimation [29]. Hereafter, the time indices (t) and () are
respectively. In (1), the control yaw moment, MB, is the omitted.
control input to be calculated to improve vehicle stability.
In (1), the term d represents unknown disturbances and s  t     t    d  t       t 
(3)
parameter uncertainties. Tire slip angles of the front wheel
f and of the rear wheel r are defined as (2) using , , and s  t    Ks  t   K  0 (4)
vx. In (1), the lateral tire forces of the front and rear wheels,
Fyf and Fyr, are functions of the tire slip angles, f and r, M B  t   I z  d  t 
respectively. The desired yaw rate, d, is derived using the  Fyf  t  cos  f  t   Fyr  t  cos  r  t  
steering angle of front wheels, f, and the longitudinal  I z      t  
 mvx  t  
velocity, vx [15].   (5)
 l f Fyf  t  cos  f  t   lr Fyr  t  cos  r  t 
 I z  K    t    d  t       t   d  t 

d  t   I z  d  
 Fyf   cos  f    Fyr   cos  r   
 I z        
FIGURE 1. 2-DOF bicycle model.  mvx   
  (6)
 l f Fyf   cos  f    lr Fyr   cos  r  

mvx   t     t    I z  K       d          M B  
 Fyf  t  cos  f  t   Fyr  t  cos  r  t 
I z   t   l f Fyf  t  cos  f  t   lr Fyr  t  cos  r  t  (1) In (5), Fyf, Fyr, and  are hard to measure. So, these
variables should be estimated by observers or estimators.
 M B  t   d  t  There have been several approaches for the purpose [30].
Since the most of vehicle parameters are uncertain and time-
 l f  t  varying, the parameter adaptation schemes have been
 f  t     t     f t  incorporated to a state observer [31]. In this paper, a sliding
 vx
 (2) mode observer, as given in [32], is adopted to estimate Fyf
  t     t   lr   t     t  and Fyr because it is quite simple and can give good
 r
vx
r
estimation results.  or vy is estimated by a signal-based
The aim of the proposed controller is to enhance both extended Kalman filter (EKF), which does not need any
maneuverability and lateral stability. The maneuverability models for state estimation, as proposed in the previous
in this research means that the yaw motion of a vehicle research [33].
accurately follows driver’s intention. The driver’s intention
has been represented as d. Thus, in order to enhance B. DESIGN OF LOWER-LEVEL CONTROLLER:
maneuverability, the controller should make  follow d. CONTROL ALLOCATION
Lateral stability means that  is maintained as small as Subsequent to computation of MB from the upper-level
possible. In the previous research,  must be smaller than controller, tire forces needed to generate MB are to be
3deg for lateral stability [15]. determined by the lower-level controller. Tire forces are
In short, a vehicle stability controller should make  generated by a single actuator or several actuator
follow d and  be a smaller value than 3deg. For these combinations with 4WIB, 4WID and 4WIS. In this research,
reasons, the sliding surface consists of two error terms as WPCA is used to determine the tire forces generated by
shown in (3). The tuning parameter  is used to 4WIS or 4WIB or 4WID or its combinations.
compromise the yaw rate error, d, with  . The stability Fig. 2 shows MB and the tire forces on four wheels. As
condition, as given in (4), should be satisfied to make both shown in Fig. 2, the front left, front right, rear left and rear
error terms be zero [15,16]. From (3), (4) and (1), MB is right wheels are indexed as 1, 2, 3, and 4, respectively. In
obtained as (5). In (5), d(t) represents unknown this figure, the driver’s intention is heading for positive yaw
disturbances and parameter uncertainties. It should be direction, i.e., clock wise. In this figure, Fx1, Fx2, Fx3 and Fx4
estimated in order to calculate the control yaw moment. If are the longitudinal braking or driving forces, generated by
the sampling time L is sufficiently small, d(t), at current 4WIB or 4WID, respectively. Fy1, Fy2, Fy3, and Fy4 are the
time step t, can be calculated with the state variables at the lateral tire forces, generated by 4WIS. These tire forces are
previous time step,  = t-L. In (6), d(t) is calculated with the to be determined in order to generate MB. For the purpose,
values of the previous time step, . This is called time-delay a WPCA is adopted [9,13-16,20,22].

VOLUME XX, 2017 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2021.3050722, IEEE
Access

 1Fy21  5 Fx21 2 Fy22  6 Fx22

J  
  12  22

 3 Fy23  7 Fx23  4 Fy24  8 Fx24
  2
 2
 qT Mq
 3 4
 (8)
  1 1 1 1 1 1 1 1 
M  diag   2 ,  2 ,  2 ,  2 ,  2 ,  2 ,  2 ,  2  σ
  1 2 3 4 1 2 3 4 
σ  diag         
 1 2 3 4 5 6 7 8
FIGURE 2. Coordinate system corresponding to tire forces and control
yaw moment.
Eqs. (7) and (8) are the general form of control allocation
with 4WIB, 4WID and 4WIS for vehicle stability control.
In Fig. 2, the steering angles, 1, 2, 3, and 4, of each
With these equations, actuator combinations with 4WIB
wheel can be set for each actuator in 4WIS. This is the most
and 4WID can be represented with the variable weights, 5,
general case. The steering angles, 1, 2, 3, and 4,
6. 7, and 8. The detailed description on how to represent
correspond to Fy1, Fy2, Fy3 and Fy4, respectively. For ICC,
the actuator combinations with 4WIB and 4WID can be
4WIS, 4WIB and 4WID can be combined into several found in the previous researches [13,15].
actuator combinations. The optimization problem, as given in (7) and (8), is a
Eq. (7) shows the geometric relation of the tire forces and quadratic programming problem with an equality constraint.
MB. Eq. (9) shows the definition of the objective function Adopting the method of Lagrange multiplier with the
for WPCA. In (8), the variable i is defined as i Fzi i.e., problem, the solution of the optimization problem can be
the radius of friction circle at each wheel. In (8), Fzi stands algebraically solved as (9). After obtaining the optimum
for the vertical force of wheel i, and  is the tire-road solution, qopt, the longitudinal tire forces Fx1~4 are converted
friction coefficient. In (8), i cannot be measured. So, it into braking pressure PB and traction torque TD, as given in
should be estimated. Fzi can be estimated from the (10). The braking torque is generated by EMB or EWB.
longitudinal and lateral acceleration signals. In the previous The traction torque is generated by an IWM. In (10), rw and
work [34], this approach has been used for unified chassis KB are the radius of a tire and the torque-pressure constant,
control.  can be also estimated from dual EKFs [31,35]. respectively. The lateral tire forces Fy1~4 from the optimum
In (8),  is a vector which consists of fictitious variable solution, qopt, are converted into the corrective angles of
weights i. In this research,  is used to represent several 4WIS, as given in (11). In (11), Ci stands for the cornering
actuator combinations such as 4WIS and 4WIS+4WIB, stiffness of each wheel.  is a tuning parameter modulating
4WIB+ 4WID, and 4WIB+4WID+4WIS. The explanation the magnitude of Ci. In general, reducing  can enhance
on the roles of  can be found in the previous work cornering performance because it generates larger i. This
[13,15,22]. is the most important parameters used to tune the control
performance [8,18].
 Fy1 
 
1
F  qopt  M 1GT GM 1GT M B (9)
 y2 
 Fy 3 
  TD,i  rw Fxi if Fxi  0  i  1, 2,3, 4 
c1 c2 c3 c4 c5 c6 c7 c8   Fy 4   M B
 (7) 
 rw Fxi (10)
 Fx1  if Fxi  0  i  1, 2,3, 4 
G  Fx 2   PB ,i  K
F   B ,i

 x3  Fyi
Fx 4   i   i  1, 2,3, 4  (11)
 Ci
q
c1  l f cos 1  t f sin 1 , c2  l f cos  2  t f sin  2 ,
C. CONSTRAINT ON FRONT SLIP ANGLES
c3  lr cos  3  tr sin  3 , c4  lr cos  4  tr sin  4 , As shown in (11), smaller  generates larger i, which can
c5  l f sin 1  t f cos 1 , c6  l f sin  2  t f cos  2 , enhance the control performance. However, larger i of
c7  lr sin  3  tr cos  3 , c8  lr sin  4  tr cos  4 front wheels can make the lateral tire forces of the wheels
easily saturated. Moreover, there is a constraint on the
magnitude of lateral tire force by a friction circle. If the
front lateral tire forces are saturated, steering of FWIS or
4WIS cannot generate MB, needed to stabilize a vehicle.
This means the deterioration of the control performance. So,

VOLUME XX, 2017 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2021.3050722, IEEE
Access

it is necessary to restrict the front slip angles and the front

0 if  i   p
lateral tire forces in order to prevent the saturation of lateral i   i  1, 2 (13)
tire forces. Moreover, it is also necessary to compensate the  Fyi ,bnd if  i   p
loss of the control yaw moment caused by the saturation of M B ,bnd  M B  c11  c2 2
the front lateral tire force or the physical constraint on the (14)
front slip angles [16,19-23]. To omit the variables, Fy1 and Fy2, new variables i are
Fig. 3 shows the relationship between the tire slip angle defined as (15). If RWIS is available, then the diagonal
and the lateral tire force. As shown in Fig. 3, the lateral tire matrix of the variable weights is set as (16). In (16), is a
force Fy has the maximum value Fy,max at the tire slip angle very small value, i.e., 1e-4. With the variable weights of
of p. However, the relationship in Fig. 3 does not take the (16), only Fy3 and Fy4 can be generated from WPCA and the
friction circle into account. So, the maximum of lateral tire other tire forces become zero. If RWIS, 4WIB and 4WID
force is to be calculated from the friction circle, as given in are available for the purpose, then the diagonal matrix of
(12). As shown in (12), the maximum value or bound of the variable weights is given as (17). With the variable weights
lateral tire force, Fyi,bnd, of the wheel i is dynamically of (17), Fy3, Fy4 and Fx1~4 can be generated from WPCA. So,
calculated from the definition of friction circle [19,20]. with the variable weights, it is not necessary to modify the
force-moment equilibrium and objective function of WPCA
Fy when considering the saturation of the front lateral tire
forces. This is called constrained WPCA (CWPCA).
5000
Fy,max   0.9 1e  4 if  i   p
4500 i   i  1, 2 (15)
1 if  i   p
4000
3500   0.6 σ  diag 1  2   1 1 1 1
3000 (16)
2500 σ  diag 1  2      
2000   0.35 (17)
1500   0.2
1000 There are several advantages in using CWPCA. The first
500 is that CWPCA requires a smaller amount of computation
00 0.2p 0.4 0.6 0.8 1 time because only an algebraic calculation is needed to
FIGURE 3. Tire slip angle and lateral tire force for several tire-road
obtain the optimum solution. Generally, an iterative
friction coefficients. procedure is needed if constrains are added to an
optimization problem. This fact is hold for the control
  Fzi 
2
Fyi ,bnd   Fxi2 , i  1, 2 (12) allocation problem [10]. For CWPCA, it is not needed. The
second is that the structure of the optimization problem is
The lateral tire forces of front wheels, Fy1 and Fy2, not needed to be modified when handling the constraint on
generated by FWIS or 4WIS, should not exceed its front slip angles by using the variable weights as given in
maximum Fy1,bnd and Fy2,bnd, respectively. In other words, (16) and (17). This is the typical feature of WPCA, as
the front slip angles should not exceed p. If Fy1 and Fy2 pointed in the previous work [13].
exceed its peak values, i.e., Fy1,bnd and Fy2,bnd, or the front
slip angles, 1 and 2, exceed its peak values, i.e., p1 and III. VALIDATION WITH SIMULATION
p2, Fy1 and Fy2 are set to the constants, Fy1,bnd and Fy2,bnd, Simulation study is presented to verify the performance of
respectively. In that case, Fy1 and Fy2 should be omitted IC with 4WIS and the actuator combinations of 4WIS,
from the optimization variable, x. So, the WPCA should be 4WIB and 4WIB under the constraint on the front slip
reformulated by omitting Fy1 and Fy2. angles.
The front lateral tire forces, Fy1 and Fy2, are set to The simulation was done via CarSim [36]. The controller
constants, Fy1,bnd and Fy2,bnd, respectively, and omitted from was implemented on MATLAB/Simulink environment. A
the optimization variable if those are saturated. As a result, closed-loop maneuver with a driver model is adopted as a
the force-moment equilibrium equation (7) and the simulation scenario. Typical closed-loop maneuver is
objective function of WPCA (8) should be modified. For double lane change on moose test track. As described in the
the purpose, new variables i representing the maximum previous works, this maneuver at high speed is so severe
lateral tire force are defined as (13). With the new variables, that any other maneuvers can be covered by it [13]. The
the control yaw moment MB in the objective function is driver model used to generate a steering wheel angle is
modified into MB,bnd, as given in (14). As shown in (13) given in CarSim. This is the implementation of the previous
and (14), MB is not changed if the front slip angles do not research [37]. The driver model replicates an unskilled
exceed p. driver, whose preview time was modelled as 0.75 sec [20-
22]. The initial vehicle speed was set to 80 km/h.

VOLUME XX, 2017 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2021.3050722, IEEE
Access

For this simulation, the D-segment SUV model was 3

Yaw Rate Error [deg/s]

selected, which is also a built-in model in CarSim. The 2
Unconstrained 4WIS
Constrained 4WIS
parameters of D-segment SUV, needed to design the upper- 1
level controller with 2-DOF bicycle model, are shown in
0
Table 1. The actuators, 4WIB, 4WID and 4WIS, were
modelled as the 1st-order system. Time constants of these -1

systems were set to 0.05, 0.05 and 0.05, respectively. In -2

4WID, the maximum power of the IWM for each wheel is -3
0 1 2 3 4 5 6 7 8
37kW [20]. Time [sec]
Table 1. Parameters of D-segment SUV in CarSim. (a) Yaw rate error (-d)
2
ms 1429kg lf 1.05m Unconstrained 4WIS

Side-slip angle [deg]

Iz 1765kg-m2 lr 1.57m Constrained 4WIS
1
Cf 36,000N/rad tf 0.75m
Cr 50,000N/rad tr 0.745m 0

-1
The first simulation is conducted to check the effect of
the proposed CWPCA on 4WIS. Tire-road friction -2
0 1 2 3 4 5 6 7 8
coefficient was set to 0.6. The slip angles of front wheels Time [sec]
are restricted to 12deg. To compensate the loss of the (b) Side-slip angle ()
control yaw moment due to the constraint on the front slip
FIGURE 4. Simulation results for each case
angles, RWIS in 4WIS is used.
20
Figs. 4, 5, 6 and 7 show the simulation results for each
FL
case. Fig. 5 shows the slip angles of each wheel for the
Slip Angle [deg]

FR
RL
unconstrained and constrained 4WIS. In Fig. 5, the legends, 10 RR
FL, FR, RL and RR represents the front left, front right,
rear left and rear right wheels, respectively. As shown in 0
this figure, the front slip angles of the unconstrained 4WIS
are quite large, and those of the constrained 4WIS were
-10
restricted to 12deg. Fig. 6 shows the lateral tire force with 0 1 2 3 4 5 6 7 8
respect to the slip angle at front left and front right wheels. Time [sec]

As shown in this figure, the lateral tire forces of the (a) For unconstrained 4WIS

unconstrained 4WIS are nearly identical to those of the 20

constrained 4WIS while the front slip angles of the FL
Slip Angle [deg]

FR
constrained 4WIS are restricted to 12deg. Generally, the 10
RL
RR
lateral tire forces of rear wheels were not saturated, as
shown in Fig. 6. Fig. 4 shows the yaw rate errors and the
0
side-slip angles of each case, which represent the
maneuverability and lateral stability, respectively. As
shown in this figure, the yaw rate error and the side-slip -10
0 1 2 3 4 5 6 7 8
angle of the constrained 4WIS is slightly worse than those Time [sec]
of the unconstrained 4WIS. The performance deterioration (b) For constrained 4WIS
of the constrained 4WIS can be negligible because the yaw FIGURE 5. Slip angles for 4WIS
rate error and the side-slip angle are within the criteria of
the maneuverability and the lateral stability, i.e., 0.08 rad/s
(=4.6deg/s) and 3deg, respectively [15]. Fig. 7 shows the
corrective steering angles of 4WIS for each case. As shown
in Fig. 7-(b), the steering angles of front wheels of the
constrained 4WIS become larger than those of the
unconstrained 4WIS. Especially, the steering angle of front
right wheel is significantly increased.

VOLUME XX, 2017 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2021.3050722, IEEE
Access

FIGURE 7. Corrective steering angles for 4WIS

Front Left Wheel Front Right Wheel
4000 4000

2000 2000 The second simulation is conducted to check the effect of

the proposed CWPCA on 4WIS on low friction road. The
0 0
simulation condition is identical to the first one except that
-2000 -2000 the tire-road friction coefficient was set to 0.3, which means
-4000 -4000 a low friction road.
-10 0 10 20 -10 0 10 20
Figs. 8, 9, 10 and 11 show the simulation results for each
Fy [N]

Rear Left Wheel Rear Right Wheel case on low friction road. Compared to the results of the
4000 4000
first simulation, the yaw rate error became larger, as shown
2000 2000 in Fig. 8. On the other hand, the side-slip angle was
0 0 maintained as small as the first simulation. As shown in
-2000 -2000
Figs. 9 and 10, the front slip angles of the unconstrained
4WIS became larger than those of the first simulation
-4000 -4000
-10 0 10 20 -10 0 10 20 because the tire-road friction coefficient is low. Moreover,
Slip Angle [deg] Slip Angle [deg] the front slip angles of the constrained 4WIS did not exceed
(a) For unconstrained 4WIS the predefined maximum one, i.e., 12 deg. This means that
Front Left Wheel Front Left Wheel the proposed method is effective in restricting the front slip
4000 4000 angles. As shown in Fig. 10, the rear slip angles were
2000 2000 maintained as small as the rear lateral tire forces were not
0 0
saturated. As shown in Fig. 11, the corrective steering
angles of 4WIS got also larger than those of the
-2000 -2000
unconstrained 4WIS. These results show that the proposed
-4000
-10 0 10 20
-4000
-10 0 10 20
method is effective for a vehicle driving on low friction
Fy [N]

road.
Front Left Wheel Front Left Wheel
4000 4000
10
2000 2000 Unconstrained 4WIS
Yaw Rate Error [deg/s]

Constrained 4WIS
0 0
5
-2000 -2000

-4000 -4000
-10 0 10 20 -10 0 10 20 0
Slip Angle [deg] Slip Angle [deg]

(b) For constrained 4WIS

0 2 4 6 8 10
FIGURE 6. Lateral tire forces with respect to slip angles for 4WIS Time [sec]
8 (a) Yaw rate error (-d)
6 FL
2
4WIS Angle [deg]

FR
RL Unconstrained 4WIS
Side-slip angle [deg]

4 RR
Constrained 4WIS
2 1
0
-2 0
-4
0 1 2 3 4 5 6 7 8
Time [sec] -1
0 2 4 6 8 10
(a) For unconstrained 4WIS Time [sec]
8 (b) Side-slip angle ()
6 FL
4WIS Angle [deg]

FR FIGURE 8. Simulation results for each case

RL
4 RR

2
0
-2
-4
0 1 2 3 4 5 6 7 8
Time [sec]
(b) For constrained 4WIS

VOLUME XX, 2017 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2021.3050722, IEEE
Access

30 (b) For constrained 4WIS

FL
FR FIGURE 10. Lateral tire forces with respect to slip angles for 4WIS
Slip Angle [deg]

20 RL
RR
10 12
FL
FR
9

4WIS Angle [deg]

RL
0 RR
6

-10 3
0 2 4 6 8 10
Time [sec] 0

(a) For unconstrained 4WIS -3

30 -6
0 2 4 6 8 10
FL Time [sec]
FR
Slip Angle [deg]

20 RL (a) For unconstrained 4WIS

RR
12
10 FL
FR
9

4WIS Angle [deg]

RL
RR
0 6
3
-10
0 2 4 6 8 10 0
Time [sec]
-3
(b) For constrained 4WIS
-6
FIGURE 9. Slip angles for 4WIS 0 2 4 6 8 10
Time [sec]
(b) For constrained 4WIS
Front Left Wheel Front Right Wheel
2000 2000 FIGURE 11. Corrective steering angles for 4WIS

1000 1000
The third simulation was conducted to check the effect of
0 0
actuator combinations on the control performances when
-1000 -1000 compensating the loss of control yaw moment. Tire-road
-2000 -2000 friction coefficient was set to 0.6. To compensate the loss
-10 0 10 20 30 -10 0 10 20 30
Fy [N]

of the control yaw moment due to the constraint on the

Rear Left Wheel Rear Right Wheel front slip angles, RWIS, 4WIB and 4WID are used. Under
2000 2000
the constraint on front wheels, four constrained cases,
1000 1000 CASE1, CASE2, CASE3, and CASE4 representing 4WIS,
0 0 4WIS+4WID, 4WIS+4WIB, and 4WIS+4WID+4WIB, are
-1000 -1000 considered in simulation, respectively. The legend 4WIS
means the unconstrained case. Figs. 12, 13 and 14 show the
-2000 -2000
-10 0 10 20 30 -10 0 10 20 30 simulation results and control inputs for each case. As
Slip Angle [deg] Slip Angle [deg] shown in these figures, four cases show nearly identical
(a) For unconstrained 4WIS
performances. This means that 4WIS, exactly RWIS, can
Front Left Wheel Front Left Wheel effectively compensate the loss of the control yaw moment.
2000 2000
This can be expected from the previous research [20]. As
1000 1000 expected in the previous simulation, the front slip angles for
0 0 all cases did not exceed the predefined limit of slip angle,
-1000 -1000 i.e., 12deg, as shown in Fig. 14.
In the previous research, 4WIS can improve the
-2000 -2000
-10 0 10 20 30 -10 0 10 20 30 maneuverability and the lateral stability without any other
Fy [N]

Front Left Wheel Front Left Wheel actuators, i.e., 4WIB and 4WID [20]. If only FWIS is
2000 2000 available for control allocation and there is the restriction
1000 1000 on the front slip angles, then 4WIB or 4WID is needed to
compensate the loss of the control yaw moment because
0 0
there is no RWIS or RWS. Generally, if there are multiple
-1000 -1000 actuators used for control allocation, then magnitude of
-2000 -2000 control input on each actuator decreases, compared to the
-10 0 10 20 30 -10 0 10 20 30
Slip Angle [deg] Slip Angle [deg] sole use of a particular actuator. However, this fact is not

VOLUME XX, 2017 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2021.3050722, IEEE
Access

hold for these cases. In other words, the use of 4WIB or CASE3 CASE4

Applied Brake Pressure [MPa]

2 2
4WID did not reduce the magnitude of the steering angle of FL
FR
FL
FR
4WIS, as shown in Fig. 13. This means that 4WIS is quite 1.5
RL
RR 1.5
RL
RR
effective for control allocation with or without the
constraint on front slip angles. 1 1

3 0.5 0.5
Yaw Rate Error [deg/s]

4WIS
2 CASE1
CASE2
CASE3 0 0
1 CASE4 0 2 4 6 8 0 2 4 6 8
Time [sec] Time [sec]
0
(b) Braking pressures of constrained 4WIS+4WIB and 4WIS+4WIB+4WID
-1
CASE1 CASE2
-2 8 8
FL FL
-3 6 FR 6 FR
0 1 2 3 4 5 6 7 8 RL RL
4 RR 4 RR
Time [sec]
2 2
(a) Yaw rate error (-d) 0 0

4WIS Angle [deg]

2 -2 -2
4WIS
-4 -4
Side-slip angle [deg]

CASE1
CASE2
0 2 4 6 8 0 2 4 6 8
1
CASE3
CASE4 CASE3 CASE4
0 8 8
FL FL
6 FR 6 FR
RL RL
-1 4 RR 4 RR
2 2
0 0
-2
0 1 2 3 4 5 6 7 8 -2 -2
Time [sec] -4 -4
0 2 4 6 8 0 2 4 6 8
(b) Side-slip angle () Time [sec] Time [sec]
80
c) Corrective steering angles of constrained 4WIS, 4WIS+4WID,
4WIS+4WIB and 4WIS+4WIB+4WID
70
Vx [km/h]

FIGURE 13. Control inputs for each constrained case

4WIS Front Left Wheel Front Right Wheel

60 CASE1
CASE2 4000 4000
CASE1 CASE1
CASE3 CASE2 CASE2
CASE4 2000 CASE3 2000 CASE3
50 CASE4 CASE4
0 1 2 3 4 5 6 7 8
0 0
Time [sec]
(c) Longitudinal velocity (vx) -2000 -2000

FIGURE 12. Simulation results for each case -4000 -4000

-10 0 10 20 -10 0 10 20
Fy [N]

CASE2 CASE4 Rear Left Wheel Rear Right Wheel

100 100 4000 4000
FL FL
CASE1 CASE1
Traction Torque [Nm]

FR FR CASE2 CASE2
80 RL 80 RL 2000 CASE3 2000 CASE3
RR RR CASE4 CASE4
60 60 0 0

40 40 -2000 -2000

20 20 -4000 -4000
-10 0 10 20 -10 0 10 20
Slip Angle [deg] Slip Angle [deg]
0 0
0 2 4 6 8 0 2 4 6 8
Time [sec] Time [sec] FIGURE 14. Lateral tire force with respect to slip angle of front wheels
for each case
(a) Tration torques of constrained 4WIS+4WID and 4WIS+4WIB+4WID

IV. CONCLUSION
In this research, the integrated chassis control was designed
with 4WIS, 4WIB and 4WID under the constraint on front
slip angles on in-wheel motor driven electric vehicles. In the
upper-level controller, the control yaw moment was
calculated with time-delay control. In the lower-level

VOLUME XX, 2017 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2021.3050722, IEEE
Access

controller, the control yaw moment was converted into tire [15] S. Yim, “Coordinated control with electronic stability control and
active steering devices,” Journal of Mechanical Science and
forces, generated by a single 4WIS or a particular actuator Technology, Vol. 29, No. 12, pp.5409~5416, 2015.
combination of 4WIB, 4WID and 4WIS. Variable weights [16] J. Nah and S. Yim, “Optimization of control allocation with ESC,
were adopted to represent several actuator combinations in a AFS, ARS and TVD in integrated chassis control,” Journal of
Mechanical Science and Technology, Vol.33, No.6, pp.2941–2948,
single framework. To handle the constraint on the front slip 2019.
angles, CWPCA was formulated. To compensate the loss of [17] E. Ono, Y. Hattori, Y. Muragishi and K. Koibuchi, “Vehicle
control yaw moment, RWIS, 4WIB and 4WID were adopted. dynamics integrated control for four-wheel-distributed steering and
Simulation was conducted on CarSim. From the simulation four-wheel-distributed traction/braking systems,” Vehicle System
Dynamics, Vol. 44, No.2, pp.139-151, 2006.
results, it was known that CWPCA is effective for control [18] P. Hang, X. Chen, S. Fang and F. Luo, “Robust controller for four-
allocation under the constraint on front slip angles, and that wheel-independent-steering electric vehicle with steer-by-wire
RWIS is enough to compensate the loss of control yaw system,” International Journal of Automotive Technology, Vol.18,
No.5, pp.785−797, 2017.
moment. When only 4WIS is available, there was slight [19] P. Hang, X. Chen, F. Luo and S. Fang, “Robust control of a four-
difference in control performance between unconstrained and wheel-independent-steering electric vehicle for path tracking,” SAE
constrained cases. This tendency was still hold when 4WIB International Journal of Vehicle Dynamics, Stability, and NVH, Vol.1,
No.2, pp.307-316, 2017.
or 4WID was used as an actuator. This means that the sole [20] J. Nah and S. Yim, “Vehicle stability control with four-wheel
use of 4WIS is effective as an actuator for ICC regardless of independent braking, drive and steering on in-wheel motor driven
the constraint on front slip angles. electric vehicles,” Electronics, Vol.9, 2020.
[21] S. Yim, S. Kim and H. Yun, “Coordinated control with electronic
stability control and active front steering using the optimum yaw
REFERENCES moment distribution under a lateral force constraint on the active front
[1] S. Murata, “Innovation by in-wheel-motor drive unit,” Vehicle System steering,” Proceedings of Institution of Mechanical Engineers, Part D:
Dynamics, Vol. 50, No. 6, pp.807-830, 2012. Journal of Automobile Engineering, Vol. 203, No. 5, pp.581-592, 2016.
[2] U. Eberle, “GM’s research strategy: Towards a hydrogen-based [22] S. Yim and Y. H. Jo,” Integrated chassis control with AFS, ARS and
transportation system,” FuncHy Workshop, Hamburg, Sept. 2006. ESC under lateral force constraint on AFS,” JMST Advances, Vol. 1,
[3] T. Koo, S. Hong, D. Kim, H. Tae, J. Kim, J. Won and H. Y. Bhae, “A pp.13-21, 2019.
study on the controllability of the vehicle steer-by-wire systems,” SAE [23] H. Heo, E. Joa, K. Yi and K. Kim, “Integrated chassis control for
Technical Paper 2020-01-0646, 2020. optimized tyre force coordination to enhance the limit handling
[4] J. S. Cheon, “Brake by wire system configuration and functions using performance,” Proceedings of Institution of Mechanical Engineers,
front EWB (Electric Wedge Brake) and rear EMB (Electro-Mechanical Part D: Journal of Automobile Engineering, Vol. 203, No. 8, pp.1011–
Brake) Actuators,” SAE Technical Pape, 2010-01-1708, 2010. 1026, 2016.
[5] A.T. van Zanten, R. Erhardt, G. Pfaff, F. Kost, U. Hartmann and T. [24] E. Joa, K. Park, Y. Koh, K. Yi and K. Kim, “A tyre slip-based
Ehret, “Control aspects of the Bosch-VDC,” Proceedings of 3th integrated chassis control of front/rear traction distribution and four-
International Symposium on Advanced Vehicle Control, pp.573-608, wheel independent brake from moderate driving to limit handling,”
1996. Vehicle System Dynamics, Vol. 56, No. 4, pp. 579-603, 2018.
[6] L. Zhai, T. Sun and Jie Wang, “Electronic stability control based on [25] E. Joa, K. Yi, K. Sohn and H. Bae, “Four-wheel independent brake
motor driving and braking torque distribution for a four in-wheel motor control to limit tire slip under unknown road conditions,” Control
drive electric vehicle,” IEEE Transactions on Vehicular Technology, Engineering Practice, Vol. 76, pp.79–95, 2018.
Vol. 65, No. 6, pp.4726-4739, 2016. [26] M.A. Khan, M.F. Aftab, E. Ahmed and I. Youn, “Robust differential
[7] M. Croft-White and M. Harrison, “Study of torque vectoring for all- steering control system for an independent four wheel drive electric
wheel drive vehicles,” Vehicle System Dynamics, Vol. 44, suplement, vehicle,” International Journal of Automotive Technology, Vol. 20,
pp.313-320, 2006. No.1, pp.87-97, 2019.
[8] S. Yim, “Comparison among active front, front independent, 4-wheel [27] X. Jin, Z. Yu, G. Yin and J. Wang, “Improving vehicle handling
and 4-wheel independent steering systems for vehicle stability control,” stability based on combined AFS and DYC system via robust Takagi-
Electronics, Vol. 9, 798, 2020. Sugeno fuzzy control,” IEEE Transactions on Intelligent
[9] J. Wang and R.G. Longoria, “Coordinated vehicle dynamics control Transportation Systems, Vol. 19, No. 8, pp.2696-2702, 2018.
with control distribution,” Proceedings of the 2006 American Control [28] R. Rajamani, Vehicle Dynamics and Control, Springer: New York,
Conference, pp.5348-5353, 2006. 2006.
[10] J. Wang, J. M. Solis and R. G. Longoria, “On the control allocation [29] K. Youcef-Toumi and S-T. Wu, “Robustness and stability analysis of
for coordinated ground vehicle dynamics control systems,” time delay control,” Proceedings of American Control Conference,
Proceedings of the 2007 American Control Conference, New York pp.2691-2695, 1992.
City, USA, pp.5724-5729, 2007. [30] M. UgrasCuma and T. Koroglu, “A comprehensive review on
[11] O. Mokhiamar and M. Abe, “Simultaneous optimal distribution of estimation strategies used in hybrid and battery electric vehicles,”
lateral and longitudinal tire forces for the model following control,” Renewable and Sustainable Energy Reviews, Vol. 42, pp.517–531,
ASME Journal of Dynamic Systems, Measurement, and Control, Vol. 2015.
126, pp.753-763, 2004. [31] X. Jin, J. Yang, Y. Li, B. Zhu, J. Wang, and G. Yin, “Online
[12] O. Mokhiamar and M. Abe, “How the four wheels should share forces estimation of inertial parameter for lightweight electric vehicle using
in an optimum cooperative chassis control,” Control Engineering dual unscented Kalman filter approach,” IET Intelligent Transport
Practice, Vo.14, No.3, pp.295–304, 2006. Systems, Vol.14, No. 5, pp.412-422, 2020.
[13] S. Yim, J. Choi, and K. Yi, “Coordinated control of hybrid 4WD [32] G. Baffet, A. Charara and D. Lechner, “Estimation of vehicle side-slip,
vehicles for enhanced maneuverability and lateral stability,” IEEE tire force and wheel cornering stiffness,” Control Engineering Practice,
Transactions on Vehicular Technology, Vol. 61, No. 4, pp.1946-1950, Vol. 17, No. 11, pp.1255-1264, 2009.
2012. [33] H.H. Kim and J. Ryu, “Sideslip angle estimation considering short-
[14] H. Fujimoto and K. Maeda, “Optimal yaw-rate control for electric duration longitudinal velocity variation,” International Journal of
vehicles with active front-rear steering and four-wheel driving-braking Automotive Technology, Vol. 12, No. 4, pp.545-553, 2011.
force distribution,” IECON 2013 - 39th Annual Conference of the IEEE [34] W. Cho, J. Yoon, J. Kim, J. Hur and K. Yi, “An investigation into
Industrial Electronics Society, Vienna, Austria, pp.6514-6519, 2013. unified chassis control scheme for optimised vehicle stability and

VOLUME XX, 2017 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2021.3050722, IEEE
Access

maneuverability,” Vehicle System Dynamics, Vol. 46, Supplement,

pp.87-105, 2008.
[35] D. Hu, C. Zong, and X. Na, “Combined estimation of vehicle states
and road friction coefficients using dual extended Kalman filter,”
Proceedings of 10th International Symposium on Advanced Vehicle
Control, pp.309-314, 2010.
[36] Mechanical Simulation Corporation, VS Browser: Reference Manual,
The graphical user interfaces of BikeSim, CarSim, and TruckSim, Ann
Arbor, Michigan, 2009.
[37] C.C. MacAdam, “Application of an optimal preview control for
simulation of closed-loop automobile driving,” IEEE Transactions on
Systems, Man, and Cybernetics, Vol. 11, No. 6, pp.393-399, 1981.

Seongjin Yim received the B.S. degree in

mechanical engineering from Yonsei University,
Korea, in 1995, and the M.S. and Ph.D. degrees
in mechanical engineering from the Korea
Advanced Institute of Science and Technology
(KAIST) in 1997 and 2007, respectively.
From 2008 to 2010, he was a Post-Doctoral
Researcher with BK21 School for Creative
Engineering Design of Next Generation
Mechanical and Aerospace Systems in Seoul
National University. From 2011 to 2013, he was
a Research Professor in Advanced Institutes of Convergence Technology
in Seoul National University. Since 2019, he has been an Associate
Professor in Department of Mechanical and Automotive Engineering,
Seoul National University of Science and Technology, Korea. His research
interests include integrated chassis control systems with V2V
communication, active roll control with state and parameter estimation,
electric power steering and steer-by-wire system.

VOLUME XX, 2017 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/