Modeling and Simulation of a Traction Control

Algorithm for an Electric Vehicle

with Four Separate Wheel Drives

* Remus Pusca, **Youcef Ait-Amirat **Alain Berthon ,**Jean Marie Kauffmann

*CREEBEL,**L2ES, Belfort, France Members IEEE
e-mail [email protected] **L2ES, Belfort, France

Abstract—The aim of this paper is to present a traction control II. Modeling

algorithm for an electric vehicle (EV) with four separate wheel A. Vehicle modeling.
drives. This algorithm is necessary to improve the handling and In order to improve vehicle comportment, a modeling of
stability of EV during cornering or under slippery road vehicle during cornering has been realized (Fig. 1). The
condition. It distributes the traction power among four drives
and especially an independent torque reference on each wheel.
following symbols are used : δ is vehicle steering angle, CG
The proposed algorithm is implemented in terms of a is center of gravity, ν the steering angle of the right front
hierarchical architecture, which incorporates all new known wheel, Ωi the rotational wheel speed, i∈[ 0-4], α the steering
vehicle systems ABS (Anti-lock Brake System), ASR (Anti Slip angle of the left front wheel, rw the radius of the wheel, r is
Regulation), ESP (Electronic Stability Program). To achieve vehicle yaw rate, e is wheel arm, lf, lr are distances of CG
good performances a correlation of the traction controller with location between rear and front axes. For the description of
motor performances has been implemented. Several simulation the model, a virtual front wheel is assumed centered between
results, which show the potential of such a algorithm, are
the two actual front wheels at rotational wheel speed Ωc_av
presented.
and a virtual wheel in CG of the vehicle at rotational speed
Keywords— EV; modeling; simulation; traction control. Ωcg.

I. Introduction α
f
δ
The principal constraints in vehicle design for personal Ff
w2

mobility are the development of a non-polluting high safety m1

δ
and comfortable vehicle. Taking into account these e
Ff

R2 VX
constraints, our interest has been focused on the 4 X 4 V
w1
R1 lf
electrical vehicle, with independent driving in-wheel-motor at x Rc

the front and with classical motors on the rear drive shaft. Mz L
y Vy
This configuration is a conceivable solution, the pollution of α
CG γ1

this vehicle is strongly decreased and electric traction gives

r
Rcg
lr

the possibility to achieve accurate and quick control of the e d2 m2 w3

δ
γ2
Fr
distribution torque. Torque control can be ensured by the Fr O
w4 d
inverter, so this vehicle does not require a mechanical d1

differential gear or gearbox. One of the main issues in the

Fig 1. Vehicle modeling during right stering.
design of this vehicle (without mechanical differential) is to
assume the car stability. During normal driving condition, all Then, speed of each wheel is given by :
drive wheel system requires a symmetrical distribution of
torque in the both sides. This symmetrical distribution is not  L + sign( δ )e ⋅ sin δ ⋅ sin δ
 Ω1 = L sin γ
⋅ Ω c_av
sufficient when the adherence coefficient of tires is changing 
: the wheels have different speeds. So, a traction control  L − sign (δ ) ⋅ e ⋅ sin δ ⋅ sin δ
Ω 2 = L sin α
⋅ Ω c _ av
system is needed. The proposed controller is implemented in  (1)
 L ⋅ cos δ + sign (δ ) ⋅ d ⋅ sin δ
terms of a hierarchical structure, which includes all new Ω 3 = ⋅ Ω c _ av
L
known vehicle systems ABS (Anti-lock Brake System), ASR  L ⋅ cos δ − sign (δ ) ⋅ d ⋅ sin δ
Ω 4 = ⋅ Ω c _ av
(Anti Slip Regulation), ESP (Electronic Stability Program).  L
The lateral dynamic of the vehicle can also been achieved by
such control. There are many studies done in this direction and steering angles are given by :
[2,3,4].

0-7803-7467-3/02/$17.00 ©2002 IEEE. 1671

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on April 21,2022 at 01:52:43 UTC from IEEE Xplore. Restrictions apply.
 ( L − sign (δ ) ⋅ e ⋅ sin δ ) ⋅ sin δ with Fa aerodynamic drag, Fc climbing forces and Fr rolling
α = sign (δ ) ⋅ arctg ( L ⋅ cos δ − sign (δ ) ⋅ ( e ⋅ cos δ + d ) ⋅ sin δ )
 2 force. During cornering (δ<0 for right steering) the different
 (2) load repartition is expressed in function as proposed in [5].
 ( L + sign (δ ) ⋅ e ⋅ sin δ ) ⋅ sin δ 
γ = sign (δ ) ⋅ arctg ( ) T r ⋅ R1 r2 ⋅d
L ⋅ cos δ + sign (δ ) ⋅ (e ⋅ cos δ + d 2 ) ⋅ sin δ T r1 = − sign ( δ ) ⋅ w
 4 ⋅ Rc ke ⋅ Rc

 Tr ⋅R 2 r2 ⋅d
The speed of front virtual wheel is used given by : T r 2 = + sign ( δ ) ⋅ w
 4 ⋅ Rc k e⋅R c
 (6)
cos( arctg
lr
⋅ tg ( δ )))
( T Tr ⋅ (d1 − d ) r w2 ⋅ d
L = − sign ( δ ) ⋅
Ω c _ av =
cos( δ )
⋅ Ω cg (3 )  r3 4 ⋅ d1 k e⋅d1

T Tr ⋅ (d1 + d ) r w2 ⋅ d
= + sign ( δ ) ⋅
From (1), it can be seen the wheels have different speeds  r 4 4 ⋅ d1 k e⋅d1
value during cornering. This is a consequence of different where ke [m/N] is tangential elasticity coefficient of the tire.
resistant (load) torque variation Tri for each wheel. The drive
torque Tdi given by wheel “wi” i∈[1-4] is : B. Motor modeling.
dΩ i A DC motor model is used [6,7]. It describes satisfactory the
Tdi = Tmi − Tri = ( J wheel + J v ) ⋅ + f r ⋅ Ωi ( 4)
dt dynamical behavior of the wheel-motor.
with Jv ,Jwheel vehicle and wheel inertia, Tmi is motor torque,
fr is friction coefficient, Ωi is angular speed of wheel “i”. In III. Traction Control
normal case, an uniform distribution of Tri, is considered The traction control algorithm is implemented in hierarchical
Tr architecture structure as shown in Fig. 2 :
Tri = and Tr = rw ( Fa + Fc + Fr ) (5)
4
S1
1
C1 *T1 *T1
0
*Tf Ω1 rw V1 αf
Tr1 s1
µ1 F1
Traction V1_calc
controller 1 charac1
in-wheel motor 1
slip_estimation1
Tr / 4 rw

ESP
Tr / 4 rw
1
load torque 1=on
Tr3 slip_estimation3 0=off
*Tr Ω3 rw V3 αr
C3 *T3 *T3 s3
µ3 F3 C1
F1 ESP
S3 V3_calc C2
in-wheel motor 3 charac3
speed estimation F3 C3
TC 3 C4
tau
V3
Vc
tau δ Fc Fr
V1
F4
Vc Fa
V2
F2
Vc aerodynamic
V4 drag Vehicle kinematics
pedal

in-wheel motor 4
S4
V4_calc αr
C4 *T4 *T4 µ F4
s4 4
*Tr
Ω4 rw V4
charac4
Tr4
TC 4 slip_estimation4

Tr / 4 rw

Tr / 4 rw
in-wheel motor 2
TC 2 V2_calc αf
Tr2 µ 2
F2
s2
*Tf Ω2 rw V2
C2
charac2
*T2 *T2
S2 slip_estimation2

Fig 2. General configuration of traction control algorithm.

0-7803-7467-3/02/$17.00 ©2002 IEEE. 1672

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on April 21,2022 at 01:52:43 UTC from IEEE Xplore. Restrictions apply.
This configuration can be divided in two parts, the first The best values of the adherence coefficient is obtained for
integrates the control algorithm and the second, vehicle s∈[0.05..0.25],[3]. This can be done only if the slip
kinematics and ESP system. In order to achieve a suggestive coefficient "s" is correctly estimated. In case of a straight
figure, just a few internal signals are presented in the line, a limitation of a slip coefficient in the stable area
explicit implementation of the blocks. (s<0.25) is sufficient. During a turn, a lateral force is
A. Hierarhical control program exerted by the tire and the adhesion coefficient “µi” is also
For implementation of control algorithm a modular structure influenced by the wheel sideslip angle αf, αr. By reference to
has been realized. The control development is explicated in Fig. 3, it is possible to see that the peak value of longitudinal
following paragraphs. characteristic decreases with increasing of sideslip angle.
1) Pedal reference The measurement of sideslip angles is difficult, so they are
In order to ensure the stability of the vehicle the four driving estimated as :
independent wheels systems requires a symmetrical
distribution of torque between the wheels of the same shaft V y + r ⋅l f V y − r ⋅ lr
(right and left). This symmetrical distribution can be assured α f = (7 )
− δ , α r =
Vx Vx
only if the reference torque applied to the motors is
correlated with there maximal speeds ( Τmi =f(Ωi ) ). So to Τhe lateral velocity Vy of the vehicle is obtained by
achieve good results, the pedal reference must take into integration of the lateral acceleration ay.
account the motor performance [8]. A restriction of motor t
reference inside of the torque speed curve is necessary. In
the case of different electric motor (EM) used in the front
V y =
∫
0
a y dt − r ⋅ V x (8 )

and rear vehicle drive shaft the motor characteristic that has To modify the slip wheel reference during turning, following
the worst torque must be used as reference by pedal block. approach is used:
This model gives the same reference torque for the front and
s min
rear drive shaft ( ∗Τr =∗Τf =∗Τmi ). s = s max + ⋅ min (α f ,α max ) (9)
An advanced development can be obtained by integration of λ ⋅ α max
energy availability in reference torque system [9]. where smax =5%, smin=4%, and λ is the characteristics
2) Slip control gradient during cornering defined by : λ= ∆µ/∆s.
During normal moving condition, the vehicle stability can During the turn, the correct estimation of wheel slip
be assured if the wheels system require a symmetrical coefficient “s” is done by equations (1--3) in speed
distribution of torque (right and left). This symmetrical estimation block.
distribution is not sufficient when the adherence coefficient 3) Traction controller
of tires vary. So, a traction controller is integrated in order The torque reference given by acceleration or brake pedal is
to adjust the slip coefficient for each wheel. A limitation of limited by pedal block in order to maintain the reference
slip coefficient in the both sides (negative and positive) is between the performance system boundaries. The output
the strategy used for implementation of ABS and ASR reference is the torque reference applied to traction
systems. As it is shown in [10], the value of the adherence controller block. This block corrects the torque reference in
coefficient is directly related to the wheel slip “s”and order to stabilize the vehicle if a critical case is detected
defined by : s=(Vw-V)/Max(Vw,V), with Vw is the wheel (skating of the wheel). This block can be divided in two
velocity and V, the vehicle velocity. A typical adhesion parts. The first, with priority in the modification of torque
characteristic is shown in Fig. 3. reference, prevents the skating of the wheels (maintaining
the wheel slip coefficient in the left side of longitudinal
characteristic peak, Fig.3). The second modify the torque
reference in order to correct vehicle yaw rate “r”. The same
block is used for each motor. With reference to (9) the
maximal value of sideslip angle has been limited to 12°.
B. Vehicle kinematics
For vehicle moving in cruise mode on a straight and flat
road with equal traction torque on the both sides, the
cornering force given by (10) is zero.

∑F = 2 F f + 2 Fr = - 2C fα α − 2 C rα ⋅α r = 0
S
y f (10 )
Fig. 3. Longitudinal and lateral adhesion coefficient during cornering for given
values of sideslip angle α.
with Cfα ,Crα front and rear cornering stiffness and Ff, Fr,
lateral forces applied of front and rear wheels. For simplicity

0-7803-7467-3/02/$17.00 ©2002 IEEE. 1673

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on April 21,2022 at 01:52:43 UTC from IEEE Xplore. Restrictions apply.
reasons a two-degree-of-freedom model with front wheels
V [p.u.] V [p.u.]
steering is used. In turn moving (Fig. 1), the cornering force 0.6
0.36
is not nil and to maintain a good trajectory a control vehicle
moment is necessary. The vehicle moment can be given as : 0.4
T + Td 4 − Td1 − Td 3 0.32

∑ M Z = 2l f F f − 2l r Fr + d ( d 2
rw
) (11 )
0.2 V3
0.28 Vcg

ΣMz = Jz.dr/dt with r vehicle yaw rate and Jz yaw inertia 0 V1

0.24
moment. -0.1
0 20 40 60 t [s] 4 5 6 t [s]
In this case, the cornering force, which equilibrates a) vehicle cycle b) wheel slip

centrifugal force, can be expressed as

Fig. 4 - Vehicle and wheel speed during slip of the wheel w1

•
∑F y = 2 F f + 2 Fr = ( m1 + m 2 )(V y + V x r ) (12 ) The intervention of ESP system (Fig 5.b) improves the
lateral dynamic stability of the vehicle. Low difference
( |r-rr| <ξ ) can be controlled changing vehicle steering
with m1+m2 = m vehicle mass, and Vx, Vy, longitudinal and
angle.
lateral velocities.
So during cornering, to assure a correct yaw moment, an
ESP system is implemented. This system uses a measured r [deg] r [deg]
0.1 0.02
yaw rate “r” and an estimated yaw rate of vehicle. In the 0.05
reference
yaw rate “r” 0
presented model, it is obtained from differential system : 0
-0.04 reference
 . -0.05
yaw rate “r”
• F1 + F2 + F3 + F4
Vx = V y r + −α fCfδ f -0.1 -0.08
vehicle yaw
Mv
• -0.15
vehicle yaw
rate “rr”

 Cr + C f Cr lr − C f l f Cf
δf -0.2 rate “rr” -0.12
V y = − Vy + ( − Vx ) r + (13)
 M v M v M v -0.25
0 4 6 10 14 t [s] -0.14
0 2 4 6 8 10 12 t [s]
 • Cr lr − C f l f Cr lr 2 + C f l f 2 Cflf d a) without ESP control b) with ESP control
r = Vy − r+ δ f + ( F1 + F3 − F2 − F4 )
 JV x JV x J J
 Fig. 5 - Vehicle yaw rate during a wheel sliping
To compute the yaw rate, we use the following expression
r = signe(δ). Vcg/Rcg with Vcg, Rcg vehicle speed and radius Consequently, the ESP system intervenes only if a high
in CG. difference (|r-rr| >ξ) is detected. The correction of vehicle
The driver, by changing vehicle steering angle, can correct a yaw moment is not possible during skating of the wheel, the
low yaw rate difference. For high difference the correction is slip regulation must have priority. Fig.6 shows the
realized by the implemented ESP system, which authorizes regulation of wheel slip in order to maintain the slip
different torque references to the both sides of vehicle. The coefficient “s1” in the stable area. We can see that the slip
output signals Ci are logic (0 or 1) integrated in Traction value is maintained inferior to 5%.
controller blocks. See [11] for more details.
s

IV. Simulation Results 0.05

0.04

To evaluate the performances of the proposed control a 0.03

s1
combined moving cycle has been considered. It starts with 0.02
s2
an acceleration interval (Fig 4.a). In this interval, a variable 0.01

adherence coefficient is assumed (Fig 4.b). So, the traction 0

system gives different torque references to wheels, in order 0 5 10 15 t [s]

to improve vehicle lateral stability.
Fig. 6 - Slip regulation during variable adherence for w1.
Fig. 5 shows the variation of vehicle yaw rate during
supposed variable adherence between tire w1 (front, right) The cycle continues with vehicle moving in cruise mode and
and road. The case without ESP (Fig.5.a) and with ESP beginning a right turn ( Fig 7.a, t ∈ [20- 45]s). In this case
intervention (Fig.5.b) has been considered. With reference to the wheels of vehicle do not follow the same trajectory and
Fig 5.a, we can see that difference between desired yaw rate so the wheels have different speeds.
and vehicle yaw rate. The variation adherence of wheel w1
gives in this case an oversteer behavior.

0-7803-7467-3/02/$17.00 ©2002 IEEE. 1674

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on April 21,2022 at 01:52:43 UTC from IEEE Xplore. Restrictions apply.
 δ [deg] V [p.u.]
0
i [p.u.]
-1 V2
0.52
-2 0.4
0.48
-3
0
-4 0.44
-5 Vcg i1
0.4 -0.4
-6 V3 V1
-7 t [s] 0.36
0 10 20 30 40 50 60 20 30 40 50 t [s]
-0.8
a) b) 50 52 54 56 58 60 t [s]

Fig. 7 - Steering angle and wheel speed during cornering.

Fig. 10 - Pedal reference and motor current during regenerative braking
Let's consider a different mass repartition between front and
rear shaft ( m1>m2). Owing to superior mass in the front V. Conclusion
shaft, during a turn the vehicle has understeer behaviour ( In this paper a modeling and simulation of a new traction
Fig.8.a). In this case the ESP system brakes the rear inside control algorithm for an electrical vehicle with four separate
wheel of curve, to stabilize vehicle trajectory (Fig.8.b). wheel drives has been proposed. Using the most advantage
of electric traction (quick and precise torque control) this
r [deg/s] r [deg/s]
algorithm which incorporates all new known vehicle
0
0
vehicle yaw rate
systems ABS, ASR, ESP is our solution of new constraints
-2

-4
“r” vehicle yaw rate
“r” imposed in vehicle design : high safety and non polluting
-4
-6 personal vehicle. The proposed control is necessary however
-8 if the road conditions vary to guarantee the lateral dynamic
-10
reference yaw rate
-8
reference yaw rate
stability. During traction and regenerative braking, a
-12
“r r” -10 “rr” correlation of traction control with motor performances has
20 30 40 50 t [s] 10 20 30 40 50 t [s]
been realized. In the turn for providing a good traction
a) b)
control, this algorithm must take into account different
Fig. 8 -Vehicle yaw rate during right stering and critical case.
wheel speeds. In this paper a modular and hierarchical
implementation is presented. The results obtained by
simulation with this model are satisfactory but in fact,
The variation of motor current imposed by ESP in order to control performances depend on the quality of the reference
correct vehicle understeer behavior is shown in Fig.9. vehicle velocity. As a solution, a fuzzy estimation speed is
proposed [10].
i [p.u.]
References
0.2
[1] C. Espanet, “Modélisation et Conception Optimale de Moteurs Sans
0 Balais a Structure Inversée, Application au Moteur –roue”. Thèse de
-0.2 doctorat 1999 Université de Franche-Comté.
-0.4 [2] S. Sakai, H. Sado, Y. Hori, “Motion Control in an Electric Vehicle with
i1
-0.6 Four Independently Driven In-Wheel Motors”, IEEE /ASME
-0.8 Transactions on Mechatronics, Vol.4, No. 1, March, 1999.
-1
i3 [3] U.Kiencke, “Observation of Lateral Vehicle Dynamics ”, 13th Triennial
20 20.5 21 21.5 22 22.5 t [s] World Congress, San Francisco, USA, 1996.
Fig. 9 - Motors current of wheels w3 and w1 during ESP control. [4] B.Arnet, M.Jufer, “Torque control on electric vehicles with separate wheel
drives”, Trondheim EPE, 1997.
[5] M.Untaru, N.Seitz, Gh. Peres, “Calculul si constructia automobilelor”, E
D P -Bucarest, 1987.
It is possible to remark that the wheel w3 is braked but as [6] T.Kenjo, S.Nagamori, “Permanent-Magnet and Brushless DC Motors”,
consequence of the low adhesion coefficient (a snowing road Clarendon Press Oxford 1985.
[7] P.Waltermann, “Modeling and Control of the Longitudinal and Lateral
surface for example) this torque is not sufficient to correct Dynamics of a Series Hybrid Vehicle”. Conference on Control
vehicle yaw motion. In this case a complementary braking Applications, Dearborn, Sept. 15-18,1996
torque is given by wheel w1 (in the same shaft). [8] A. Jackson, M.Brown, D.Crolla “Co ordinated Mobility Control of 6x6
off Road Vehicle with Individual Wheel Control”. European Automotive
After the return of the vehicle in cruise mode the simulation Congress , Bratislava, june. 18th –20th , 2001.
cycle is finished by a regenerative braking (Fig 10.a). [9] R. Pusca, Y. Ait-Amirat, A.Berthon, J.M. Kauffmann “Advanced Control
Applied to Hybrid Electrical Vehicle”. EVS19 , Busan, oct'02.
We can see that the regenerative torque (current) is limited [10] G. Spinnler, “Conception des machines, Principes et applications”, CH-
1015 Lausanne, 1997.
by motor performances. If the energy management is [11] R. Pusca, Y. Ait-Amirat, A.Berthon, J.M. Kauffmann “Fuzzy Logic
integrated, this regenerative torque must be also correlated Based Control Applied to Hybrid Electrical Vehicle with Four Separate
with the maximal recharge current accepted by the battery. wheels ”. VTC'01, Atlantic City, oct'01.

0-7803-7467-3/02/$17.00 ©2002 IEEE. 1675

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on April 21,2022 at 01:52:43 UTC from IEEE Xplore. Restrictions apply.