Professional Documents
Culture Documents
Abstract
An advanced traction control system can help limit wheel rotation and enhance vehicle stability. This article presents a
new traction control system under complicated situations, including the low slippery road surface and split-m road sur-
face. First, a 15-degree-of-freedom nonlinear vehicle dynamics simulation model is established. Then, the driving wheel
speed is regulated by adjusting the engine torque and the wheel brake pressure. The engine torque regulation is based
on a proportional–integral–derivative plus ant colony optimization controller, and the wheel brake pressure regulation is
based on a proportional–integral plus ant colony optimization controller. Finally, the proposed strategies are applied to
simulation and road tests. Results indicate that the algorithm exhibits high control accuracy and robust performance.
Compared with the traditional proportional–integral–derivative controller, the proposed strategies improve vehicle
acceleration performance and stability.
Keywords
Vehicle engineering, ant colony optimization, traction control system, parameter optimization, automotive control
Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 Advances in Mechanical Engineering
theory in sliding-mode. L Li et al.7 proposed a fuzzy improve a TCS using ACO. The proposed control
controller to solve the problem of coordination between approach was tested with low m slippery road surface
engine torque and active braking and verified it on and split-m road surface simulations in the simulation
uneven low-friction road conditions. and real vehicle test, which proves the effectiveness
TCS can meet the complex working conditions under of the method. Compared with the traditional TCS
active braking and engine-coordinated control.8 PID algorithm, the ACO-optimized TCS improves the
controllers have been widely used because of their sim- acceleration performance and reduces the number of
ple structure, convenient realization and stable and reli- adjustments.
able operation. It is very important how to show those The rest of this article is structured as follows. The
PID control systems clearly.9 The parameters of tradi- vehicle dynamic model is provided in section ‘‘Vehicle
tional PID controllers are optimized by manual meth- model.’’ The practical problem of the conventional
ods10 and are unsuitable for all situations. When a algorithm is discussed in section ‘‘Conventional TCS.’’
vehicle is in a complex environment, it will cause accel- Section ‘‘Controller design’’ describes the controller
eration difficulties and affect vehicle stability. design, which includes an engine controller and an
Optimization problems can make up for this short- active brake controller. In section ‘‘Improved ACO
coming, which can be found in many industries and in algorithm’’ the parameter optimization strategy for the
the scientific community. In contrast to classical mathe- PID controller using the ACO algorithm is described.
matical concepts, bionics algorithms (e.g. ant colony, In section ‘‘Result and discussion,’’ the results of simu-
genetic, and particle swarm algorithms) have been lation and real-car validation are shown in detail.
developed through simulation of natural ecological sys- Discussion and further research topics are also pre-
tems to solve complex optimization problems. These sented. Finally, section ‘‘Conclusion’’ summarizes the
algorithms greatly enrich the current optimization tech- article.
niques and provide practical solutions for combinator-
ial optimization problems that are difficult to deal
using traditional optimization techniques. Ant colony Vehicle model
optimization (ACO) is suitable for solving complex Body model
computing problems especially in optimizing PID para-
A nonlinear 15-degree-of-freedom (DOF) vehicle model
meters based on insect behavior under different envir-
is used to predict vehicle response under different road
onments.11,12 Ant algorithm has attracted increasing
conditions during various driving maneuvers, as shown
research attention since it was established and has been
in Figure 1.
successfully applied in traveling salesman problem
TCS controls the vehicle’s longitude dynamics but
(TSP),13,14 vehicle routing problem,15,16 and mobile ad
ignores several factors. In this article, the vehicle model
hoc networks.17,18 Several parameter optimization
includes longitude motion, lateral motion, vertical
methods have been proposed for ACO.19–22 Duan
motion, roll motion, pitching motion, and yaw motion.
et al.19 introduced an ACO algorithmic approach for
The vehicle dynamic equations can be described as
optimizing the parameters of a nonlinear PID control-
ler and proved that it has good control and robust per- follows
formance. Furthermore, ACO has been applied in X
dynamic systems. The system based on the adaptive ant m(u_ + vy w vz y) = Fxsij + mg sin u ð1Þ
colony (AAC) design also successfully optimizes the
dynamic model from preprocessing flight data.23,24 In
this study, ACO is used to design TCS. Given that
ACO can be easily combined with other local algo-
rithms, it can make up for the dynamic response that
PID cannot obtain and thus improve the stability and
acceleration of the vehicle.
This article proposed a combined solution to the
vehicle traction control problem. The wheel speed regu-
lation realized with engine torque control and active
braking control. A PID control system has been
designed for engine torque regulation and a
proportional–integral (PI) controller has been
employed for wheel brake pressure regulation. For
both control system, the controller gains have been
tuned by using a metaheuristic optimization method
named ACO. The significance of this article is to Figure 1. 15-DOF vehicle model.
Jin et al. 3
X
m(v_ + vz u vx w) = Fysij mg sin u cos u ð2Þ Conventional TCS
X During the driver’s full-throttle acceleration, the driving
m(w_ + vx v vy u) = (Fzsij + Fzdij ) mg cos u cos u
wheel torque exceeds the maximum tire adhesion force
ð3Þ between the tire and the pavement. The traditional TCS
X c algorithm is often faced with two problems. On the one
Jx v_ x = (Txij ) + (Fzsfl + Fzsrl Fzsfr Fzsrr ) ð4Þ hand, the brake control responds quickly, but the tor-
2
X que of the engine cannot be reduced immediately
Jy v_ y = (Tyij ) + b(Fzsrr + Fzsrl ) a(Fzsfl + Fzsfr ) because of the relatively low dynamic response of the
ð5Þ engine. On the other hand, controlling the torque and
X braking torque simultaneously is difficult. Therefore,
Jz v_ z = (Tzij ) + a(Fysfl + Fysfr ) b (Fysrr + Fysrl ) the reduction of engine torque and the generation of
c wheel brake torque require optimization using high-
+ ( Fxsfl + Fxsfr Fxsrl + Fxsrr ) precision control algorithms to reduce the rotation of
2
the wheel.
ð6Þ
ACO exhibits a great advantage in solving optimiza-
Wheel rotational equations can be expressed as tion problems. On the one hand, the ant algorithm is a
positive feedback and used as a search and optimization
dv tool in engineering because of its effectiveness in finding
Jw = Tq Fd rv Tb ð7Þ
dt very good solutions to difficult problems. On the other
hand, ACO is a heuristic algorithm of population and
has strong robustness and optimization ability. The ant
Tire model colony algorithm is a bionics algorithm based on ant
The tire model selected is the magic formula model pro- colony and has the advantages of good robustness.
posed by Professor Pacejka.25 This model is a mathe- This work applies ACO to optimize the conventional
matical expression based on the tire physical prototype, PID controller, which has the advantages of simple
which is used to accurately calculate the tire forces in structure, good robustness and, high reliability.
the longitudinal and lateral directions.6 The equation
of the tire longitudinal forces y can be defined as Controller design
y = D sinfC arctan½ B l E (B l arctan(B l))g Control scheme
ð8Þ The overall structure of the TCS is shown in Figure 2.
Given the lack of transverse differential lock control-
The slip ratio of tire can be expressed as follows ler’s braking application, the driving force can be trans-
mitted on both sides when the driving torque on both
v rv u
l= 3 100% ð9Þ sides of the differential is equal.1 The larger driving tor-
v rv
que causes the slipping of the low-side wheel, resulting
in velocity difference higher than 0. In this case, most
of the drive torque will be lost in the differential speed,
Engine model
the engine, and the transmission. The wheel speed dif-
An engine is a complicated system. Many factors affect ference in the driving wheel is 0, which is considered as
the engine torque output, but the most important are the desired value of TCS. The system monitors each
the throttle opening uth and crankshaft speed ne . wheel and adjusts the brake pressure and engine tor-
During vehicle motion, a lag exists between throttle que, thereby optimizing the stability and acceleration
input and torque output; as such, the time lag problem characteristics of the vehicle.
should be considered. Therefore, the first-order inertial In general, vehicle acceleration presents difficulties
environment is used to establish the engine model. In under typical road conditions, namely, low m pavement
this article, the dynamics model of the engine can be and the split-m pavement. Vehicle longitudinal velocity
depicted as is important for electronic stability control system; in
the meantime, vehicle longitudinal velocity is estimated
este1 based on wheel speed signals. When the vehicle enters
Te = Ts
te2 s + 1 ð10Þ the acceleration fault, the accuracy of speed estimation
Ts = f (uth , ne ) will be directly affected, which leads to a larger error in
slip calculation. This article proposes a control scheme
The engine output torque depends on throttle angle to solve the problem by maintaining the traction wheel
and rotation speed. speed.
4 Advances in Mechanical Engineering
wheel speed of the two-driving wheel is expressed as Assuming M is the number of ants, and M ants are
wd rw = (wl rw + wr rw )=2, control deviation is expressed randomly distributed on these nodes. In a discrete
as e(t) = vd rv (t) vo rv (t). Engine target torque is the space, ant movement is the transition from one state to
output torque of the engine to achieve the target wheel another. Pheromones will be stored on the route. The
speed. The PID control is expressed as strength of the pheromone accumulation of edge (i, j) is
denoted by t(i, j).
ðt Equations (15) and (16) indicate the state transition
de(t)
Ttar (t) = Kp e(t) + Ki e(t)dt + Kd ð11Þ rules, which are used to represent a real ant K located
dt
0 at node i and select the next node j. Equation (15) is a
greedy choice of technology and is the optimal combi-
nation of short distance and large pheromone level;
Active brake controller equation (16) balances this problem by allowing the
probability of the next node to be selected
The active brake pressurization rate is high, and the lag
(
time is short. The wheel responds very quickly to pres- arg max t(i, u) Dh(i, u)b if q q0
sure interference, which easily causes large fluctuation s= u2hk (i)
in wheel speed. 0 otherwise
When the traction controller detects that one or ð15Þ
more driving wheels rotate faster than the other wheels, 8 b
it calls active brake controller to apply brake friction to < Pt(i, j)½Dh(i, j) if s 2 hk (i)
t(i, u)½Dh(i, u)b
restrain wheel skid and reduce traction force. Braking pk (r, s) =
: u2hk (r)
Parameter Value
Figure 7. Velocity versus time in the split-m road surface: (a) velocity versus time with no control, (b) velocity versus time under
traditional control, and (c) velocity versus time under control.
traditional TCS improves the acceleration performance The engine torque adjustment is shown in
of the vehicle compared with the system without TCS. Figure 8(a). At the beginning, the engine output torque
However, the low-side wheels are still skidding, result- increases with increasing throttle opening. During skid-
ing in waste of energy from 3 to 8.5 s. This long-time ding of the side wheels, the engine output torque does
adjustment may cause damage to the controller in real not adjust to the low torque, as in low-profile pave-
vehicle test and cannot meet the control accuracy ments, but is maintained at 120 N m, thereby ensuring
requirements. The simulation results show that the the acceleration of the vehicle. The optimization para-
coefficients of adhesion are 0.8 for the left road and 0.2 meters of torque regulation are obtained by simulation:
for the right road in Figure 7(c). At 0.445 s, the TCS Kp = 75; Ki = 2; Kd = 0:8.
based on ACO optimization inhibits the transitional The active braking adjustment is shown in
slip of low-side wheels. The engine torque no longer Figure 8(b). To improve the acceleration of the vehicle
increases as the accelerator pedal opening increases. on the road surface, active pressure interference is only
Under the action of active braking, the TCS can effec- carried out for low-side driving wheels. In this process,
tively restrain the slip of the slippery side wheel and the low-side drive wheel is subjected to the active inter-
improve the speed by utilizing rough side wheel. From ference pressure, and the maximum interference pres-
Figure 7(b) and (c), the comparison between the tradi- sure reaches 12 bars. The optimization parameters of
tional control and the optimal control shows that the braking adjustment are obtained by simulation:
speed of the vehicle with traditional is 19 m/s at 8 s, Kp = 25; Ki = 1:8.
whereas the speed of the optimal control is 19 m/s at
5.3 s. Therefore, the proposed method is better than the
previous two methods on the split-m road surface. The Road test result
acceleration in the method with TCS is 80% higher An experimental study was carried out in the test field
than that without TCS and 50% higher than that in the of Heihe City of North China to verify the performance
traditional PID control method. of TCS on winter road condition. The simulation and
Jin et al. 9
Figure 8. Simulation results in the split-m road surface: (a) torque versus time under control and (b) cylinder pressure versus time
under control.
Figure 10. Velocity versus time in the low m slippery road surface: (a) wheel speed and vehicle velocity and (b) the winter test
results obtained on the snow road.
10 Advances in Mechanical Engineering
Figure 11. The winter test results obtained on split-m road surface: (a) wheel speed and vehicle velocity and (b) torque.
The target torque may be large in the short time after driving wheel is over sliding, the slip rate cannot be cal-
the intervention. With the sliding of the wheel, the con- culated. Therefore, slip rate should not be calculated
troller can quickly reduce the engine output torque and based on TCS control. The choice of target wheel speed
maintain a certain driving force. The pressure curve is can not only be used to observe wheel speed but also to
not drawn because the braking pressure is 0 during con- reflect slip rate. Based on this perspective, the advan-
trol process. The torque oscillation of the front wheel tage of this work compared with previous research is
has obvious smooth characteristics. The optimized that the acceleration performance of vehicles is better,
parameters obtained are Kp = 125; Ki = 2; Kd = 0:5. and the traction loss is smaller on low and separate
As shown in Figure 11, the vehicle starts from the roads. For example, in the real vehicle test of ice and
stationary state on the separation road. After the slip- snow road, the driving wheel only skips once, which
ping of the low-side wheel, TCS is activated, and engine effectively reduces the transition consumption of driv-
management system and hydraulic brake modulator are ing force.
controlled. The controller can exert braking force on The proposed strategy also uses ACO to optimize
the wheel in time so that wheel speed can be restored. the PI parameters of the torque controller in TCS.30
The brake pressure was intervened several times with Given that the pressure controller is not considered, the
the transition of the wheel, and the maximum value is single engine torque controller is difficult to meet the
15 bars. TCS improves the driving force of the low-side requirements of TCS split-m road surface. Therefore, in
wheel and effectively prevents the wheel from overturn- the simulation test of the split-m road surface, the vehi-
ing from 1 to 10 s, as shown in Figure 11(b). The engine cle speed was approximately 5.81 m/s at 5 s with ant
control exits frequently, which can affect ride comfort. colony PI controller, while it was approximately 19 m/s
The optimization parameters of torque regulation are at 5.3 s with the ACO-optimized TCS controller.
obtained by simulation: Kp = 80; Ki = 5; Kd = 0:7. In the study of the slip ratio as control target, the
The optimization parameters of braking adjustment are PID controller combined with the optimal longitudinal
obtained by simulation, Kp = 30; Ki = 2:7. slip rate identification,31 and the vehicle speed was
approximately 14 m/s at 4.5 s under low-adhesion road,
while it was approximately 19 m/s at 6 s with the ACO-
Discussion optimized TCS controller. The proposed method is
In contrast to other automotive engineering strategies, faster than the other methods. Hence, the control strat-
the new bionic strategy exhibits good adaptability to egy is excellent, and the optimization results are signifi-
terrain. This strategy can not only meet the general cant. The ant algorithm is also applicable and
operational requirements but also improve the dynamic reasonable for study of vehicle stability.
performance through TCS under severe conditions.
The control objectives of TCS are mainly divided
Conclusion
into slip rate and speed types. The phenomenon of vehi-
cle speed is corrected based on slip rate control.4–6,26–29 This article introduces a new TCS based on ACO for
In this study, we design the control target of TCS based engine torque and pressure controllers. The integrated
on speed control. As shown in equation (9), slip rate is strategy is more complex than the traditional PID con-
calculated based on wheel and vehicle speeds. When the trol. The ACO is used to optimize the PID/PI
Jin et al. 11
19. Duan HB, Wang DB and Yu XF. Novel approach to q uniform random number between 0 and 1.
nonlinear PID parameter optimization using ant colony q0 ratio of using greedy selection techniques
optimization algorithm. J Bionic Eng 2006; 3: 73–78. rv wheel radius
20. Pellegrini P, Stützle T and Birattari M. A critical analysis s Laplace transform factor
of parameter adaptation in ant colony optimization. te1 lay time
Swarm Intell 2012; 6: 23–48. te2 dynamic characteristic fitting coefficient
21. Eiben AE and Smith JE. Parameter control in evolution- u longitudinal velocity
ary algorithms. In: Eiben AE and Smith JE (eds) Intro- v lateral velocity
duction to evolutionary computing. Berlin: Springer, 2003,
w vertical velocity
pp.167–187.
22. Randall M. Near parameter free ant colony optimisation.
B tire stiffness factor
In: Dorigo M, Birattari M, Blum C, et al. (eds) Ant col- C shape factors
ony optimization and swarm intelligence. Berlin: Springer, D peak factor
2004, pp.374–381. E curvature factors
23. Lei X and Guo K. The model identification for small Fd driving force between road and tire
unmanned aerial rotorcraft based on adaptive ant colony Fzdij tire dynamic load
algorithm. J Bionic Eng 2012; 9: 508–514. Fzsij tire static load
24. Duan HB, Zhang XY, Wu J, et al. Max-min adaptive ant Fxs , Fys , projection of the suspension force to the
colony optimization approach to multi-UAVs coordi- Fzs vehicle body in the vehicle coordinate
nated trajectory replanning in dynamic and uncertain system, respectively
environments. J Bionic Eng 2009; 6: 161–173.
Jw wheels inertia moment
25. Pacejka HB. Tire and vehicle dynamics. 3rd ed. Oxford:
Jx roll inertia moment
Butterworth Heinemann, 2012.
26. Mansour K and Reza K. A new strategy for traction con- Jy pitch inertia moment
trol in turning via engine modeling. IEEE T Veh Technol Jz yaw inertia moment
2001; 50: 1540–1548. Kd differential coefficient
27. Borrelli F, Bemporad A, Fodor M, et al. An MPC/hybrid Ki integral coefficient
system approach to traction control. IEEE T Contr Syst Kp proportional gains
T 2006; 14: 541–552. Q problem dependent parameter
28. Shi J, Li X, Lu T, et al. Development of a new traction Tb braking torque
control system for vehicles with automatic transmissions. Te engine output torque
Int J Auto Tech 2012; 13: 743–750. Ts static output torque from the engine map
29. Jung H, Kwak B and Papk Y. Slip controller design for
chart
traction control system. Int J Auto Tech 2000; 1: 30–33.
30. Jing LI, Xue L and Jian Z. Regulation of PID controller
Txij four-wheel torque components in the X
parameters based on ant colony optimization algorithm axis
in vehicle traction control system. J Jilin Univ Sci 2008; Tyij four-wheel torque components in the Y
77: 769–772. axis
31. Jin LQ, Ling M and Yue W. Tire-road friction estimation Tzij four-wheel torque components in the Z
and traction control strategy for motorized electric vehi- axis
cle. PLoS ONE 2017; 12: 6. Tq longitudinal driving torques
g global pheromone decay coefficient
Appendix 1 Dhk (i, j) value of objective function
Dtk (i, j) quantity of pheromone in an ant
Notation hk (i) set of all the nodes
u body pitch angle
a distance between the front-axle and the uth throttle opening
center of gravity l slip ratio of tire
b distance between the rear-axle and center u body roll angle
of gravity v wheel speed
c distance between the front and rear axles vx roll rates
m sprung mass vy pitch rates
ne crankshaft speed vz yaw rates