Professional Documents
Culture Documents
Integrated Traction Control Strategy For 4-Motorized-Wheels Electric Vehicles With Improvement of Economy and Longitudinal Driving Stability
Integrated Traction Control Strategy For 4-Motorized-Wheels Electric Vehicles With Improvement of Economy and Longitudinal Driving Stability
net/publication/312437355
CITATIONS READS
0 126
2 authors:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Xudong Zhang on 31 January 2017.
ABSTRACT: This paper presents an integrated traction control strategy (ITCS) for 4-motorized-wheels elec-
tric vehicles. The purpose of this proposed strategy is to improve vehicle economy and longitudinal stability.
On high adhesion roads, economy optimization algorithm is applied to maximize motors efficiency by means
of the optimized torque distribution. On low adhesion roads, sliding mode control (SMC) algorithm is imple-
mented to guarantee the wheel slip ratio under the target slip ratio to make full use of road driving capacity.
Since the optimal slip ratio on different road conditions varies a lot, a quasi-optimal slip ratio is introduced
and set as the target value. The integrated traction control strategy is designed for combining the vehicle econ-
omy and longitudinal stability via three torque allocation cases: economy-based torque allocation, pedal self-
correcting torque allocation and inter-axle torque allocation. Finally, simulation results show that the proposed
strategy effectively reduces vehicle energy consumption, suppresses wheel skid and enhances the vehicle lon-
gitudinal driving performance.
∑ Ti = Td
i =1− 4
T1 + T3 ≥ T2 + T (5)
T ≤T
i motor
where Tmotor is the motor maximum torque output.
For the sake of simplicity, a few assumptions are
Figure. 1 Motor external characteristics and efficiency map listed:
(1) The slip ratio and rotational speed difference of
The motor’s torque response can be simplified as each wheel is quite small and hence, the motor rota-
first-order inertia as the following equation tional speed n equals beta·u/r.
Tcmdi (2) The torque output is identical if the motors are on
Ti = (1) the same axle.
τ ⋅ s +1 Despite its simplification, these assumptions are
where Ti is the motor’s torque output, Tcmdi is the still justified, since only longitudinal dynamic is dis-
torque command signal and τ is the time constant. cussed here and economy control will be activated
Since motor power is the function of motor effi- only when the controller is fully sure that the car
ciency, rotational speed and torque output, based on runs stable. The assumptions can be formulated as,
the efficiency map in Figure 1, the motor power can
T1 = T3 = T f / 2 = p ⋅ Td / 2
be easily calculated as
ni ⋅ Ti
, Driving condition T2 = T4 = Tr / 2 = (1 − p ) ⋅ Td / 2
Pi = η i (Ti ,n i ) (2) u
ni ⋅ T ⋅ η i (Ti ,n i ), n = beta ⋅
Regeneration condition r
where Pi and ni denote the motor power and rota- where p∈ [0.5, 1] is economy distribution coeffi-
tional speed; ηi is the motor efficiency. cient; Tf is the front axle driving torque; Tr is the rear
axle driving torque. Then the final economy objec-
tive function is obtained,
3 INTEGRATED CONTROL STRATEGY
J d = n
p ⋅ Td
+
(1 − p ) ⋅ Td
(6)
3.1 Economy-based Control η[0.5 p ⋅ Td , n] η[0.5(1 − p ) ⋅ Td , n]
pT (1 − p)Td
3.1.1 Objective functions establishment J r = n pTdη d , n + (1 − p)Tdη , n (7)
The relationship among the motor’s torque, RPM 2 2
and efficiency is shown in Figure 1. The motor effi-
ciency is quite distinguishing in different working
3.1.2 Solutions to objective functions
regions and especially when it works in the low
In this paper, genetic algorithm is applied to solve
speed or low torque output region, the motor has
these discontinuous objective functions. It can effec-
poor efficiency. If the driving torque can be in real-
tively prevent local optimization and find the solu-
time distributed among four motors according to
tions accurately and rapidly.
AVEC'16
The objective functions solution p, is given as a conditions. The objective function and inequality
two dimensional lookup table with current desired constraints for solving this quasi-optimal slip ratio
driving torque and motor rotational speed, which can are expressed as Equation (9).
avoid numerous online calculations and meet system
µi (λopt ) − µi (λqo )
real-time requirements. The economy-based distribu- min f (λ ) = ∑ (9)
tion coefficient p is shown in Figure 2. µi (λopt )
0.9
achieved under the proposed quasi-optimal slip ratio
0.8
is almost as same as the one under optimal slip ratio.
0.7
µ (λ ) = c1 (1 − e− c λ ) − c3λ 2
(8) which represents the designed sliding surface. In this
paper, exponential reaching law is selected and giv-
The friction model parameters of typical road sur-
en as
faces, and corresponding optimal slip ratios λopt and
maximum normalized driving forces are given in s& = − ks − ε sgn(s ) (13)
Table 1.
where c is the strictly positive constant gain and ε is
Table 1. Burckhardt friction model parameters reaching velocity factor.
Taking the derivative of Equation (12), and sub-
Roads c1 c2 c3 λopt µ(λopt)
stituting into Equation (13) gives
Dry asphalt 1.280 23.99 0.52 0.17 1.17
Wet asphalt 0.857 33.82 0.347 0.13 0.80 − ε sgn(s) = e& + ce = λ& + e(λ − λqo ) (14)
Dry concrete 1.197 25.16 0.537 0.16 1.09
Snow 0.194 94.12 0.065 0.06 0.19 Since for 4-motorized-wheels EVs each wheel is
Ice 0.05 306.3 0 0.03 0.05 independently controlled by a motor, we have the
wheel torque balance equation as
The fixed quasi-optimal slip ratio λqo is not the
optimal choice for one particular road condition.
J w w& = Tq ⋅ beta − Ft ⋅ r (15)
However µ(λqo) is most close to µ(λopt) on all road
AVEC'16
where Tq is defined as the torque which can make despite the total road adhesion is more than the driv-
tires reach the quasi-optimal slip ratio. er desired torque, however, if the car is running on a
Taking derivative of the wheel slip ratio that is joint road, the friction coefficient for front and rear
Equation (11) and substituting it into (15), the wheel wheels will be different. Or due to the vehicle load
slip ratio dynamic equation is obtained, distribution, the wheel with lower vertical load can
only be applied with relatively small traction. Oth-
u u&
λ& = (To ⋅ beta − F t ⋅r ) + (16) erwise, it may lead to a seriously tire skid. Converse-
J wω 2 r ω ⋅r ly, for the wheel with larger vertical load, it is neces-
sary to moderately increase its motor torque output
The sliding mode control law is derived by substi- to make full use of the road capacity.
tuting Equation (16) into Equation (14), By comparing different torque signals including
J wω 2u& driver desired torque, current motor torque output
Tq = −[ε sgn( s ) + ks + c(λ − λqo )] and quasi-optimal slip ratio torque, a suitable control
u ⋅ beta (17) case will be activated and kept operating until the
J ω ⋅ u& Ft ⋅ r vehicle running condition changes. Then the system
+ w +
u ⋅ beta beta needs to reselect the corresponding control case con-
stantly to guarantee the control strategy’s adaptabil-
Finally, to reduce high-frequency chattering due ity under different conditions.
to the non-deterministic switching control, a satura-
tion function sat(s/Φ)is applied instead of sgn(s),
where Φ is a positive constant which determines the Driver desired
torque Td
Current motor
state Ti, ni
Slip ratio
λi
N
The integrated control strategy aiming at the econo- Y
N
my optimization and stability control is proposed All wheels skid
60
80
60 40
40 20
20
0
0 200 400 600 800 1000 1200
0
0 200 400 600 800 1000 1200 time [s]
time [s]
Figure. 4 New European driving cycle Figure. 6 Thermal power and improvement
Figure 5 illustrates that vehicle working efficien- Table 4. Improvement results analysis
cy is significantly improved after adopting economy- Energy Consumption (kJ) Equivalent
Cycle Energy
based torque allocation. In some conditions, the im- Weight
Stage ITCS Saving
provement rate even reaches around 15%. Besides, Reduction
as Figure 6 shows, the economy-based torque alloca- Low 4WETD 919.2 5.26% 65.21 kg
tion also achieves lower thermal loss than the others, 870.9
Speed FWD 947.1 8.04% 96.34 kg
which can reduce motor’s heat load and prolong its High 4WETD 1267.0 8.06% 97.54 kg
service life. 1164
Speed FWD 1173.4 0.46% 17.65 kg
Table 4 lists the simulation results of economy Whole 4WETD 2186.2 6.88% 75.43 kg
improvement and equivalent weight reduction in 2036
Cycle FWD 2120.4 4.00% 51.22 kg
each driving cycle stage. The highest energy saving
occurs in high speed stage. Economy-based torque
allocation decreases 8.06% energy consumption 4.2 Simulation for stability-based control
compared with 4WETD. Lowest energy saving, In order to test the validity and robustness of the
0.46%, also occurs in high speed stage but it’s rela- stability performance of the proposed control strate-
tive to FWD. As for the whole driving cycle, econ- gy, a compressive simulation experiment is addition-
omy-based torque allocation can decrease 6.88% and ally carried out under strongly varying conditions,
4.00% energy consumption respectively, by compar- which contains the switching of different road condi-
ison to 4WETD and FWD. If we use weight to tions and driver desired torque. The road surface in-
measure the energy saving effect, it means 75.43 kg put for front tires in this simulation is listed in Table
and 51.22 kg weight reduction equivalently, which is 5. The time delay of the road switching for front and
difficult to achieve in engineering design phase. rear tires is also taken into consideration. In this
1
(a)
simulation, the accelerator pedal position is set at
4WEDT FWD ITCS 85%. After 9 s, the vehicle velocity reaches 60 km/h
motor efficiency [-]
0.9
and then the accelerator pedal is released to 28%
0.8 opening.
0.7
Table 5. Road surface input for front wheels
0 200 400 600 800 1000 1200 Distance (m) Tire/Road Friction Coefficient
time [s]
(b)
0 - 10 0.8
35
10 - 50 0.1
efficiency improvement [%]
30 4WEDT/ITCS
25
FWD/ITCS 50 - 80 0.2
20 80 - end 0.9
15
10
5
As shown in Figure 7, in the first 2.3 seconds, the
0
car is running on a high adhesion road. The traction
0 200 400 600 800 1000 1200
time [s] capacity provided by the road is much greater the
driver desired torque. Therefore, Case 1 economy-
Figure 5. Working efficiency and improvement based torque allocation is activated as shown in Fig-
ure 7(d). According to current torque and RPM sig-
AVEC'16
nals, the economy coefficient p is set as 0.5, which aiming at improving economy and longitudinal driv-
means even torque allocation is applied. And then, ing stability. According to the economy-based distri-
the front wheels move into the low adhesion area bution coefficient, driver desired torque is allocated
firstly and the rear wheels are still in the high adhe- reasonably among four driving motors, which max-
sion area. The control case switches from Case 1 to imizes the motor driving efficiency and improves
Case 3 inter-axles torque allocation as shown in Fig- overall vehicle economy. Meanwhile, pedal self-
ure 7(d). Based on sliding mode control, the torque correcting torque allocation and inter-axle torque al-
output of front motors should be reduced to prevent location are moreover designed to overcome vehicle
the slip ratio from skidding. At the same time, the stability problems on low adhesion road. By means
controller raises the rear motors output to ensure the of solving nonlinear programming problem, the qua-
vehicle dynamic performance to meet the driver’s si-optimal slip ratio on different road surfaces is ob-
demand as shown in Figure 7(b). When all the four tained as a fixed value. Based on sliding mode con-
wheels move into the low adhesion road, Case 2 pe- trol, the torque output of each motor is adjusted to
dal self-correcting torque allocation is activated. At keep the wheel slip ratio under the target slip ratio
about 6.4 seconds, the friction coefficient changes rapidly and stably. It can be concluded that the pro-
from 0.1 to 0.2, however the Case 2 keeps running, posed ITCS is able to balance the relationship be-
since its condition is still satisfied. Furthermore, it tween road/tire friction conditions, motor efficiency
should be noted that after the accelerator pedal posi- and driver desired torque; achieve vehicle economy
tion decreases from 85% to 28% at the 9th second, and longitudinal stability optimization.
Case 1 starts running, however the front and rear
motors’ outputs are different the torque outputs from
before. That is because through analysis of desired ACKNOWLEDGMENTS
torque and RPM information, economy coefficient p
is reset as 1, front-wheel-drive mode, to make the The work is supported by the BCV project (Berlin
vehicle achieve the optimal economy performance. City Vehicle Project) in Technical University of Ber-
(a)
lin. And the authors would like to thank China
100 Scholarship Council (CSC) for providing a scholar-
displacement [m]
100 rear mototrs [1] Hori, Y., Toyoda, Y., & Tsuruoka, Y. (1998). Traction
control of electric vehicle: Basic experimental results us-
50
ing the test EV “UOT Electric March”. Industry Applica-
0
tions, IEEE Transactions on, 34(5), 1131-1138.
0 1 2 3 4 5 6 7 8 9 10 [2] Esmailzadeh, E., Vossoughi, G. R., & Goodarzi, A.
time [s]
(c)
(2001). Dynamic modelling and analysis of a four motor-
20 ized wheels electric vehicle. Vehicle System Dynamics,
front tires
15 35(3), 163-194.
tire slip rate %
rear tires
10
[3] Yin, D., Oh, S., & Hori, Y. (2009). A novel traction con-
trol for EV based on maximum transmissible torque esti-
5
mation. Industrial Electronics, IEEE Transactions on,
0 56(6), 2086-2094.
0 1 2 3 4 5 6 7 8 9 10
time [s]
[4] He, H., Peng, J., Xiong, R., & Fan, H. (2014). An acceler-
(d) ation slip regulation strategy for four-wheel drive electric
3
vehicles based on sliding mode control. Energies, 7(6),
control case switch
2.5 3748-3763.
2 [5] Nam, K., Hori, Y., & Lee, C. (2015). Wheel Slip Control
for Improving Traction-Ability and Energy Efficiency of a
1.5
Personal Electric Vehicle. Energies, 8(7), 6820-6840.
1
0 1 2 3 4 5 6 7 8 9 10 [6] Yuan, X., & Wang, J. (2012). Torque distribution strategy
time [s] for a front-and rear-wheel-driven electric vehicle. Vehicu-
Figure. 7 Simulation results on variable conditions: (a) Vehicle lar Technology, IEEE Transactions on, 61(8), 3365-3374.
displacement; (b) motor torque output; (c) actual slip ratio; (d) [7] Smith DE, Starkey JM. Effects of Model Complexity on
control case switch. the Performance of Automated Vehicle Steering Control-
lers: Model Development, Validation and Comparison.
Vehicle System Dynamics. 1995 Mar;24(2):163–81.
5 CONCLUSIONS [8] X. Zhang, D. Göhlich, X.L. Wu. (2015). Optimal torque
distribution strategy for a four motorized wheels electric
vehicle. EVS28 International Electric Vehicle Symposium
In this paper, a novel integrated control strategy and Exhibition.
for 4-motorized-wheels electric vehicles is proposed,