IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 34, NO.

5, SEPTEMBER/OCTOBER 1998 1131

Traction Control of Electric Vehicle: Basic

Experimental Results Using the Test
EV “UOT Electric March”
Yoichi Hori, Member, IEEE, Yasushi Toyoda, and Yoshimasa Tsuruoka

Abstract— The most distinct advantage of the electric vehi-

cle is its quick and precise torque generation. However, most
electric vehicles developed to date have not yet utilized it. In
this paper, two novel traction control techniques of an electric
vehicle using this advantage are proposed. One is the model-
following control and the other is the optimal slip ratio control.
The basic effectiveness of the proposed methods is demonstrated
by real experiments using the dc-motor-driven test vehicle “UOT
(University of Tokyo) Electric March.”
Index Terms— Antilock braking system, electric vehicle, esti-
mation, model following control, motion control, road condition
estimation, robust control, slip ratio control, traction control.

I. INTRODUCTION

R ECENTLY, many electric vehicles (EV’s) have been

developed [1], mainly to solve environmental and energy
problems caused by the use of internal combustion engine
Fig. 1. Characteristics of driving and lateral forces.

vehicles (ICV’s). Some of them already have enough per-

formance, even in practical use. However, they have not best control performance, the estimation method of the road
yet utilized the most remarkable advantage of the EV. The surface condition is proposed, and its basic realizability is
generated torque of electric motors can be controlled much shown by actual experiments.
more quickly and precisely than that of internal combustion
engines [13]. It is well known that the adhesion characteristics
II. STATE OF THE ART OF TRACTION CONTROL
between tire and road surface are greatly affected by the
control of the traction motor. This means that the vehicle Traction control is the control which suppresses tire slip
stability and safety can be greatly improved by controlling when the vehicle is accelerating on an icy road, for example.
the motor torque appropriately. If we can use special low-drag It is realized by controlling the traction force. As a result,
tires with smaller energy loss, the range of one battery charge driving and cornering performance are improved.
will be greatly expanded. We should consider two forces acting on the vehicle body.
In this paper, we will propose some novel traction control They are the driving (longitudinal) and side (lateral) forces
techniques, which can be realized only by utilizing the electric [2]. As depicted in Fig. 1, these force characteristics strongly
motor’s quick torque response [14]. They are the model- depend on the slip ratio . In acceleration, is defined by (1),
following control (MFC) and the optimal slip ratio control. where and are the wheel and vehicle speeds:
By using a newly developed dc-motor-driven test vehicle, the
“UOT (University of Tokyo) Electric March,” we will show (1)
some successful experimental results. In order to achieve the
The side force takes its maximum value when
Paper IPCSD 98–36, presented at the 1997 IEEE/IEEJ Joint IAS Power and becomes quickly smaller for bigger . If increases by
Conversion Conference, Nagaoka, Japan, August 3–6, and approved for
publication in the IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS by the a sudden decrease of road friction, the side force becomes
Industrial Drives Committee of the IEEE Industry Applications Society. drastically smaller. This causes serious problems, including
Manuscript released for publication May 26, 1998. drift-out in front-wheel-driven cars, spin in rear-wheel-driven
Y. Hori and Y. Tsuruoka are with the Department of Electrical Engi-
neering, University of Tokyo, Tokyo, 113-8656 Japan (e-mail: hori@hori.t.u- cars, and drift-out with rotation in four-wheel-driven cars.
[Link]). Such a loss of cornering force is extremely dangerous. The
Y. Toyoda was with the Department of Electrical Engineering, University average traction force is also decreased.
of Tokyo, Tokyo, 113 Japan. He is now with the Power Device Development
Center, Hitachi Works, Hitachi Corporation, Hitachi, Ibaraki, 319-1221 Japan. We think that traction control can be classified into the
Publisher Item Identifier S 0093-9994(98)07228-4. following two steps.
0093–9994/98$10.00  1998 IEEE
1132 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 34, NO. 5, SEPTEMBER/OCTOBER 1998

TABLE I
COMPARISON OF TRACTION CONTROLS FOR ICV

Fig. 2. Block diagram of MFC.

1) Longitudinal control—For example, this is the adhesion more. Traction control can be realized only by soft-
improvement control to prevent slip. This is achieved by ware. Even the lowest cost “basic car” can have high-
controlling the traction force. performance traction control.
2) Lateral control—For example, this is the yaw control to 2) Quick Response—In the ICV, more than 200 ms are
keep the yaw motion at zero. At present, this is achieved needed to open the throttle actuator. The actual response
mainly by controlling the steering angle. is much slower, because additional delay in the mechan-
For the lateral control, the steering angle of the front wheels ical system must be included. In contrast, the response
is the dominant control input [3]–[5]. Such a technique is time of the electric motor torque is less than 10 ms.
already well developed for the ICV. Most of these results can 3) Easy Controller Design—In the ICV, unknown strong
be applied to the EV in a much more sophisticated manner. nonlinearity lies in the transfer characteristics from the
For example, by introducing the independent control of four control input (for example, air valve angle to engine, oil
in-wheel motors, we can realize completely a new motion pressure of brake system, etc.) to the generated torque.
control of the EV [6] and [7]. However, in this paper, as our This makes it difficult to construct a mathematical
first attempt, we focus our discussion on longitudinal control. model for controller design. In the EV, by applying
To realize the effective traction control system, we need simple current control, the generated torque is exactly
a sophisticated mechanism to quickly reduce the excessive proportional to the torque command.
driving torque. In the ICV, this is realized mainly by the
following three techniques. IV. MFC
1) Engine Control—Engine torque itself is suppressed. To In this paper, we propose two control strategies: “MFC”
reduce the air supply is the basic technique, but for and “optimal slip ratio control.” MFC is the starting point of
quicker response, advanced techniques like fuel-cut and our research project of “the control of an EV,” and its basic
spark timing shift are used together. feasibility is demonstrated here by actual experiment.
2) Brake Control—Wheel rotation itself can be stopped
by braking. This method has quicker response than the A. Principle of MFC
engine control. Independent control of left and right tires Fig. 2 shows the block diagram of MFC. is the
is effective for -split braking. Brake control should be current command proportional to the acceleration pedal angle.
used together with the engine control, because brake is the rotational speed of the driving shaft. increases
parts often have thermal problems. drastically when the tire slips. Although the vehicle dynamics,
3) Mission Control—Driving torque of the slipping tire is including tire characteristics and road surface friction, are very
transferred to the nonslipping tire. This technique is complicated, if we introduce the slip ratio , the vehicle body
effective for -split road. As the total torque cannot be can be seen as one inertia system having the equivalent inertia
reduced, this mission control should be applied together moment of
with the engine control.
(2)
Table I summarizes the advantages and disadvantages of
these techniques. Here, , and are the shaft inertia moment, vehicle
weight, and tire radius, respectively. Equation (2) means that,
when slip occurs, the vehicle seems lighter. Therefore, we use
III. ADVANTAGE OF EV
the following inertia moment with in the reference
The electric vehicle has the following great advantages for model:
the realization of high-performance traction control.
(3)
1) Low Cost—In the case of an ICV, the above-mentioned
techniques need additional costly hardware, e.g., throttle When there is no slip, actual is almost equal to .
and brake actuators. The EV does not need anything A signal is not generated from the MFC controller. If the tire
HORI et al.: TRACTION CONTROL OF EV 1133

Fig. 3. Slip experiment.

slips, the actual wheel speed increases immediately. The (a) (b)
model wheel speed does not increase. By feeding the speed
difference back to the motor current command, the actual
motor torque is reduced quickly, and it induces readhesion.
As this control function is needed only in a relatively higher
frequency region, we used a high-pass filter on the feedback
pass. In actual implementation, in order to avoid the offset
problem of an integrator, two high-pass filters are inserted
before taking the difference between the actual and the model
speeds. When the feedback torque from the MFC blocks is
positive, it is forced to be zero.

B. Experimental Result of the MFC (c) (d)

Fig. 3 shows the slip experiment using the UOT Electric Fig. 4. Experimental results of MFC. (a) Motor current. (b) Slip ratio. (c)
Wheel speed. (d) Vehicle speed.
March. We used iron plates as a slippery road surface. Water
is scattered to reduce the friction coefficient. The vehicle is
accelerated by the constant current command of 300 A. The
feedback gain in Fig. 2 is 30. The front wheels are on the
slippery area between and 1.7 s.
Experimental results are given in Fig. 4. We can see that
the MFC can reduce the motor current effectively when the
vehicle goes onto the slippery area, and then the slip ratio is
kept much lower compared to the case of current control only.
Some vibration observed in the current waveform in Fig. 4(a)
can be suppressed in the future.

V. OPTIMAL SLIP RATIO CONTROL

The MFC is a very rough approach, although it has been
Fig. 5. Block diagram of the optimal slip ratio controller.
shown that the motor control is really effective for adhesion
improvement. If we want more exactly to regulate the slip
ratio within the desired range, a more precise approach is where
needed. Fig. 5 shows the idea of the optimal slip ratio control motor torque (force equivalent);
developed from this viewpoint. When the optimal slip ratio friction force;
is decided by the road condition estimator, the slip ratio wheel inertia (mass equivalent);
controller receives the command and tries to realize it. vehicle weight.
The friction force between the road and wheel is given by
A. Vehicle Model
We assume that the two motor torques and friction forces (6)
are the same in left and right tires and that the rolling and air
frictions are small enough. In Fig. 6, the kinematic equations where is the vertical force given by .
of the wheel and vehicle take the forms From (1), the following perturbation system is derived:

(4) (7)
and
where and are the wheel and vehicle speeds at the
(5)
operational point. The friction force is represented using ,
1134 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 34, NO. 5, SEPTEMBER/OCTOBER 1998

Fig. 6. Vehicle model.

Fig. 8. Nominal slip ratio is given by a = 1.

Fig. 7. Slip ratio controller.

the gradient of curve, as

(8)
By combining (7) and (8) with the perturbation forms of
(4) and (5), the transfer function from the motor torque to the
slip ratio is finally given by

(9)

where the time constant is given by (10), which is propor-

tional to the wheel speed :

(10) Fig. 9. Root locus against parameter variation.

The typical value of in our experimental vehicle is

150–200 ms when and the vehicle speed is around Fig. 8 shows the nominal slip ratio used in the slip ratio
10 km/h. Note that can be negative in the right-hand side of controller. We defined it by . The point of is
the peak point of the curve. located just in the left side of the peak and is stable. Both of
the longitudinal and lateral forces are kept still high.
B. Design of Slip Ratio Controller
We used a simple proportional integral (PI) controller with C. Robustness to Parameter Variation
a variable gain as the slip ratio controller given by (11), as Because the actual system parameters change widely, we
depicted in Fig. 7. Its nominator compensates for the pole of should investigate the robustness of the slip ratio controller.
(9). The integral gain is constant and the proportional gain is Fig. 9 draws the root locus to continuous change of and
proportional to the vehicle speed: (actual ). From the figures, we can see that the roots move
to the left-half plane when the controller gain increases.
(11) It is interesting that this controller stabilizes the system, even
when actual is negative, although the roots move toward the
Finally, the transfer function from the slip ratio command
unstable region.
to the actual slip ratio becomes

(12) D. Simulation of Slip Ratio Control

Fig. 10 shows the vehicle model we used in the simulation.
represents the motor torque and the total gear ratio of
If , this is a simple first-order delay characteristic the drive train. represents the summation of traction force
with a time constant which can be adjusted by . Here, we transferred to the contact point of tire and road surface. It is
put this response time from 50–100 ms. the product of the traction coefficient and , the
HORI et al.: TRACTION CONTROL OF EV 1135

Fig. 10. Vehicle model used in the simulation.

Fig. 12. Simulation of the slip ratio control.

We showed the relation between the slip ratio and the

friction coefficient in Figs. 1 and 11, but it varies very
widely according to road surface condition, as shown in Fig.
14. It is clear that the slip ratio where the friction force takes
its maximum value varies according to road condition. This
means that the road condition should be estimated relatively
quickly for giving the optimal slip ratio to the slip ratio control.
To know the road surface condition, we should estimate the
Fig. 11. – characteristics used in the simulation. friction coefficient [10]–[12]. If we can measure the vehicle
speed directly by using the nondriven wheel, the friction
coefficient can be obtained by (13) based on (3) and (4):
vertical load on the contact point. is defined as the function
of , which is given by the measured curve shown in Fig. 11. (13)
Fig. 12 is the simulation result. The response time of the
slip ratio controller is set to be 100 ms. We can see good When the vehicle speed cannot be measured directly, we
response characteristics. can estimate based on

E. Experimental Results of Slip Ratio Control (14)

Fig. 13 shows the experimental results of the slip ratio
control using the laboratory-made experimental EV “UOT In our case, we can use both of these methods. Fig. 15 shows
Electric March.” Here, the response time is 50 ms and the the estimation result of the curve of a dry asphalt road
target slip ratio is 0.1 in Fig. 13(a) and is changed stepwise when no slip control is active. At the point around ,
from 0.3 to 0.1 in Fig. 13(b). the gradient of the – curve is about 1.
Basically, we can see fairly good performance, but there Fig. 16 shows the estimation results on a wet iron surface
are some problems. First, the actual value of was much under the slip ratio control proposed in the previous section.
smaller than the nominal value 1. This made the response time Here, the optimal slip ratio is smaller than 0.05. It is also
longer than the designed value. Next, in Fig. 13(b), we see an noticed that, in our experiment shown in Fig. 13(a), the actual
undershoot to the slip ratio command of 0.1. This is because gradient of the curve at was almost . We
the motor controller we used is only a 1-quadrant chopper, can see that the slip ratio controller still works effectively,
which cannot absorb the motor current. even when the operation point is unstable, but, in this case,
we should have commanded a lower slip ratio.
For effective traction control, it is enough to know the
VI. ESTIMATION OF ROAD CONDITION gradient of the curve. We are going to introduce an
In the previous section, we showed the effective slip ratio adaptive identification method for realization of “the optimal
control. The next problem is how to give the optimal slip ratio slip ratio control based on the estimation of the road surface
to the slip ratio controller. condition,” which is our next target.
1136 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 34, NO. 5, SEPTEMBER/OCTOBER 1998

Fig. 14. Various road conditions.

(a)

Fig. 15. Estimation result of – curve of dry asphalt road.

Fig. 16. Estimation result of – curve of wet iron plate under the slip ratio
control.

the motor speed is suddenly increased by tire slip. Next, we

showed that the optimal slip ratio control has more advanced
performance. Such kinds of quick control are first realized
(b) only in EV’s. It is clearly shown that relatively sophisticated
Fig. 13. Experimental results of the slip ratio control. control theory can work well in actual experiments.
Advanced adhesion control is helpful for lateral control,
like yaw disturbance attenuation [6]–[9]. This is because the
VII. CONCLUSION
proposed optimal slip ratio control keeps the tire slip within
We have proposed a new field of “motion control of the small region. where both of the longitudinal and lateral
an EV.” The EV is a very interesting object, combining adhesion coefficients are still high enough.
the electrical and mechanical engineering fields from the
viewpoint of motion control. As an example, we proposed
advanced adhesion control utilizing the quick and precise
torque response of the electric motor. APPENDIX
We proposed the MFC and the optimal slip ratio control. We Configuration of UOT Electric March: We developed a
confirmed that the MFC can reduce its torque quickly when real test EV, the “UOT Electric March,” as shown in Fig.
HORI et al.: TRACTION CONTROL OF EV 1137

TABLE II
SPECIFICATION OFUOT ELECTRIC MARCH

Fig. 17. UOT Electric March.

Fig. 18. Configuration of UOT Electric March.

17. It is a converted car, the IC engine of which has been

replaced by an electric motor.
The front two wheels are driven by a l9-kW series-wound dc
motor through a 5-speed manual transmission and a differential
gear. The 1-quadrant dc chopper supplies power to the motor.
Its current limit is 400 A and it can produce maximum torque
over 100 N m, which is enough to perform the slip experiment.
Current and speed sensors are also implemented. To detect the
Fig. 19. Basic experiment on the current response.
vehicle speed, a speed sensor is implemented in the rear wheel.
The aim of our research is not in motor control itself, but
in motion control of an EV, where the type of motor is not a the Electric March while a graduate student in our laboratory.
problem. What is required is for the traction motor to generate They also thank T. Uchida and K. Yamazaki for their help in
torque quickly enough. Our development is based on this quick manufacturing of the vehicle.
and relatively precise torque generation. From this point of REFERENCES
view, the dc motor is the easiest device to confirm our idea,
in particular, for a basic experiment at a university. [1] K. Rajashekara, “History of electric vehicles in general motors,” IEEE
Trans. Ind. Applicat., vol. 30, pp. 897–904, July/Aug. 1994.
Fig. 18 shows the control system of the vehicle, and Table II [2] H. Sakai, Tire Engineering. Tokyo, Japan: Grand-Prix, 1987.
gives its specification. We used a note-type personal computer [3] M. Ito and K. Isoda, “The present and future trends of traction control
to realize all control algorithms. It not only executes the control system,” Jidosha-Gijutsu, Soc. Automotive Engineers Japan, vol. 46,
no. 2, pp. 32–37, 1992.
algorithm and puts out the voltage command to the chopper, [4] K. Ise et al., “The ‘Lexus’ traction control (TRAC) system,” Society
but also reads and records the sensor data. As the control of Automotive Engineers of Japan, Tokyo, Japan, Tech. Paper 900212,
algorithms are written by software (C-language), we can easily 1990.
[5] S. Yamazaki, T. Fujikawa, and I. Yamaguchi, “A study on braking
investigate various control strategies. and driving properties of automotive tires,” Trans. Soc. Automotive
Fig. 19 shows the basic experimental results of the current Engineers Japan, vol. 23, no. 2, pp. 97–102, 1992.
[6] J. Ackermann, “Yaw disturbance attenuation by robust decoupling of
controller. car steering,” in Proc. 13th lFAC World Congr., 1996, 1996, pp. l–6.
[7] Y. Wang and M. Nagai, “Integrated control of four-wheel-steer and yaw
ACKNOWLEDGMENT moment to improve dynamic stability margin,” in Proc. 35th IEEE-CDC,
1996, pp. 1783–1784.
The authors would like to express their sincere thanks to [8] N. Iwama et al., “Active control of an automobile—Independent rear
T. Furuya, Kansai Electric Power Company, for his work on wheel torque control,” Trans. SICE, vol. 28, no. 27, pp. 844–853, 1992.
1138 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 34, NO. 5, SEPTEMBER/OCTOBER 1998

[9] S. K. Sul and S. J. Lee, “An integral battery charger for four-wheel drive Yasushi Toyoda received the B.S. and M.S. degrees
electric vehicle,” IEEE Trans. Ind. Applicat., vol. 31, pp. 1096–1099, in electrical engineering from the University of
Sept./Oct. 1995. Tokyo, Tokyo, Japan, in 1995 and 1997, respec-
[10] C. Liu and H. Peng, “Road friction coefficient estimation for vehi- tively.
cle path prediction,” Vehicle System Dynamics Supplement, vol. 25. He joined Hitachi, Ltd. in 1997 and is currently
Amsterdam, The Netherlands: Swets & Zeitlinger, 1996, pp. 413–425. with the Power Device Development Center, Hitachi
[11] A. Daiss and U. Kiencke, “Estimation of tire slip during combined Works, Hitachi Corporation, Hitachi, Japan.
cornering and braking observer supported fuzzy estimation,” in Proc.
13th IFAC World Congr., 1996, pp. 41–46.
[12] L. R. Ray, “Nonlinear tire force estimation and road friction identifi-
cation: Simulation and experiments,” Automatica, vol. 33, no. 10, pp.
1819–1833, 1997.
[13] T. Furnya, Y. Toyoda, and Y. Hori, “Implementation of advanced
adhesion control for electric vehicle,” in Proc. IEEE Workshop Advanced
Motion Control, 1996, vol. 2, pp. 430–435.
[14] Y. Hori, Y. Toyoda, and Y. Tsuruoka, “Traction control of electric Yoshimasa Tsuruoka received the B.S. degree in
vehicle based on the estimation of road surface condition—Basic 1997 in electrical engineering from the University
experimental results using the test EV UOT Electric Mar.,” in Proc. of Tokyo, Tokyo, Japan. He is currently working as
IEEJ-IEEE Power Conversion Conf., 1997, vol. 1, pp. 1–8. a postgraduate student in information and commu-
nication engineering.
His current research interest is natural language
processing.
Yoichi Hori (S’81–M’83) received the B.S., M.S.,
and Ph.D. degrees in electrical engineering from the
University of Tokyo, Tokyo, Japan, in 1978, 1980,
and 1983, respectively.
He joined the Department of Electrical Engineer-
ing, University of Tokyo, in 1983 as a Research
Associate. Since 1988, he has been an Associate
Professor. During 1991–1992, he was a Visiting
Researcher at the University of California, Berkeley.
His research fields are control theory and its indus-
trial applications, in particular, to motion control,
mechatronics, power electronics, power systems, electric vehicles, etc.
Prof. Hori is a member of the Institute of Electrical Engineers of Japan,
Japan Society of Mechanical Engineers, Society of Instrument and Control
Engineers, Institute of Systems, Control and Information Engineers, Robotic
Society of Japan, Japan Society of Simulation Technology, Japan Society
of Mechanical Engineers, Society of Electric Vehicle Engineers, and Society
of Automotive Engineers of Japan. He received the Best Transactions Paper
Award for the IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS in 1993.