Professional Documents
Culture Documents
A Driver Model-Based Direct Yaw
A Driver Model-Based Direct Yaw
A Driver Model-Based Direct Yaw
Abstract
In this article, a driver model–based direct yaw moment controller, selected as the upper controller, is developed, of
which the control target is determined through a reference driver model in accordance with the driver’s intention. The
sliding surface is chosen by the difference between the desired yaw rate and the real output yaw rate. Then, the desired
yaw moment is calculated by the sliding mode control. In the lower controller, a novel control torque distribution strat-
egy is designed based on the analysis of the tire characteristics. In addition, an admissible control set of the control tor-
ques is calculated in real time through an embedded tire model ‘‘UniTire.’’ Finally, a driver-in-the-loop experiment, via
the driving simulator, is conducted to verify the proposed direct yaw moment controller.
Keywords
Direct yaw moment control, sliding mode control, driver model, tire model
Introduction for which each wheel can be exerted driving and brak-
ing forces.
Direct yaw moment control (DYC) representing one Some control strategies and algorithms of the DYC
type of the electronic stability control (ESC) is used to system development have been presented in the current
apply the different longitudinal force between the inner literature, such as sliding mode control,4 linear quadra-
and outer wheels, which can stabilize the vehicle yaw tic regulator,5 fuzzy control,6,7 and model predictive
motion and increase the vehicle maneuverability.1,2 control.8 Besides, some integrated control including
There are now three ways to achieve the DYC function: DYC and steering was developed.9,10 There may be still
braking, driving, or both. Owing to its simple realiza- room for improvement as to two points of the current
tion, braking DYC is widely used as the control type; literature. For one thing, the driver’s intention can
quite the opposite, this mandatory deceleration can dis- draw conclusion by the way of measuring the steering
tribute the driver’s intention which leads to the acci- wheel angle in the current literature. This method fully
dent. On the contrary, the driving DYC can conquer
depends on the driver’s maneuver behavior; therefore,
this drawback,3 but it is implemented by a costly and
complex electronic differential.
The in-wheel motor electric vehicle (IEV) has four State Key Laboratory of Automotive Simulation and Control, Jilin
in-wheel motors, which can independently drive the University, Changchun, People’s Republic of China
wheel. The IEV has inherent advantages over tradi-
Corresponding author:
tional vehicles in DYC realization. Without any extra- Ye Zhuang, State Key Laboratory of Automotive Simulation and Control,
neous components, the DYC system tends to be more Jilin University, Changchun 130025, People’s Republic of China.
flexible and efficient due to the IEV in-wheel motors Email: yzhuang_cn@163.com
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
Simulator development
A driving simulator was built so as to validate the pro-
posed controller, as shown in Figure 1. The host/target
configuration is adopted in the simulator development.
In the host PC, the commercial vehicle dynamics soft-
ware CarSimÒ12 and the system-design platform
LabVIEWTM13 are applied to set up the software envi-
Figure 1. Driving simulator configuration. ronment. In the target PC, the LabVIEW Real-Time
Module provides the real-time operating system. The
the driver’s reaction delay and incorrect maneuver driver’s maneuver command is transmitted to the vehi-
directly influence the validity of the control target. For cle dynamics model through the NI PCI-6229 data
another, the control torques are mainly distributed in acquisition board.
proportion and coinstantaneously to the wheels. By A passenger car, with the main parameters listed in
this distribution strategy, the potential of the vehicle Table 1, is modeled in CarSim for the simulator.
stabilization cannot be entirely exploited. The validation plots of the developed vehicle model are
This problem has been preliminarily discussed by presented in Figure 2. A double lane change (DLC)
Chen et al. in previous work.11 In this article, a more maneuver is selected, with the vehicle velocity of
comprehensive analysis and a more objective validation 80 km/h. The validation results show (comes out) that
are fulfilled. The contributions of this work can be the CarSim model closely matches the experimental
listed as follows. data; therefore, the driving simulator is further used to
First, the control target is determined by a reference conduct a driver-in-the-loop experiment. (It is apparent
driver model, which precisely read the driver’s inten- that both the experiment and the CarSim have the
tion. Therefore, the negative impacts of the driver’s same steering wheel angle diagram. Although there do
reaction delay and incorrect maneuver are exist a little bit deviation in yaw rate, roll angle, and
counteracted. the lateral acceleration diagrams, the experiment has
Second, a novel control torque distribution strategy similar outlines in general, compared to the CarSim,
is implemented based on an embedded nonlinear tire which is within the acceptable range and can be used to
model ‘‘UniTire.’’ UniTire also computes the tire adhe- conduct a driver-in-the-loop experiment.)
sive limit in real time, which provides a control bound-
ary aiming to restrict the magnitude of the control Controller design
vector. Finally, a driver-in-the-loop validation experi-
ment is conducted. Vehicle dynamics model
This article is set forth as follows: in ‘‘Simulator A 7 DOF reference vehicle model is developed for the
development’’ section, the development of a driving controller design, including longitudinal motion, lateral
simulator is presented. In ‘‘Controller design’’ section, motion, yaw motion, and rotational motion of the four
the entire control method is described. The driver-in- wheels, which are shown in Figure 3.
the-loop experiment results on the driving simulator The equations of the vehicle motion are expressed as
Wang et al. 3
Figure 2. CarSim model validation: (a) steering wheel angle, (b) yaw rate, (c) roll angle, and (d) lateral acceleration.
Figure 4. UniTire tire model: (a) longitudinal force with slip ratio and (b) lateral force with slip angle.
Kx S x
fx = ð10Þ
Fxm
Ky S y
fy = ð11Þ
Fym
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
f = f2x + f2y ð12Þ
Figure 5. Block diagram of driver/vehicle system.
where Kx is the longitudinal slip stiffness, Ky represents
the cornering stiffness, Fxm is the potential extreme
value of the longitudinal force, and Fym is the potential to different vertical forces Fz is shown in Figure 4(a),
extreme value of the lateral force. while the relationship between lateral force Fy and the
The normalized resultant force has an exponential slip angle a with respect to different Fz values is shown
equation, expressed as in Figure 4(b).
Referred to the tire model, the following expressions
1 can represent the relationship between the longitudinal
F = 1 exp f E1 f2 E12 + f3 ð13Þ
12 tire force and lateral tire force
Furthermore, we can get the longitudinal force and
Fxij = ftxij Vx , Vy , g, vij , Tij ð16Þ
the lateral force by following equation
Fyij = ftyij Vx , Vy , g, vij , Tij ð17Þ
fx
Fx = F m Fz ð14Þ
f x Thus, equations (1) to (7) can be compactly rewritten
as
fy
Fy = F m Fz ð15Þ
f y G_ = f7 ðG, uÞ ð18Þ
where Fz is the vertical load, and mx and my are the where the vehicle state vector G and control input vec-
longitudinal and lateral friction coefficients, tor u are defined below
respectively.
The tire force comparisons of the passenger car G = ½ Vx Vy g vfl vfr vrl vrr T
between the UniTire output and the experiment data
are presented in Figure 4. The relationship between u = ½ Tfl Tfr Trl Trr T
longitudinal force Fx and the slip ratio k with respect
Wang et al. 5
Outline of proposed controller design the driver’s intention is meaningless. Another possible
A control target plays a key role in a control system. In scenario is due to the driver’s reaction delay, the con-
a DYC controller, the control target is generally related troller may miss the opportunity of controlling the vehi-
to the driver’s intention. Therefore, precisely under- cle and thus fail to stabilize it. Therefore, it is important
standing the driver’s intention plays a crucial role in to employ dSW to predict the driver’s intention. This
the DYC controller development. Since the driver can clarify the control target and enhance the control
model is established on the human driver and can, con- effect.
siderably, respond to the human driver’s behavior, the Unfortunately, dSW cannot be easily derived from
human driver’s intention can be read, which is of great dSW , since in the inverse transfer function of M(s), the
advantage. output dSW is ahead of the input dSW in the time
A preview driver model in a driver/vehicle system is domain; this is a noncausal and unrealizable process.
illustrated in Figure 516: P(s) is the driver’s preview Nevertheless, owing to the more and more advanced
strategy; H(s) is the driver’s control transfer function, image recognition and satellite navigation, detecting
in which C(s) is the driver’s regulation and calculation the road information ahead based on modern sensors is
block and M(s) is the maneuver transfer function, no longer a problem. Therefore, dSW is obtained from
which mainly represents the driver’s delay; B(s) is the the road information fa rather than dSW in the control-
feedback block; G(s) is the vehicle response; f is the cur- ler design. The road information is considered known
rent road information; fa is the road information ahead; in this article, and the acquisition algorithm would not
e is the path error; dSW represents the desired steering be discussed.
A hierarchical control architecture is utilized in this
wheel angle; dSW represents the actual steering wheel
article1 as illustrated in Figure 6. In the upper control-
angle, and y is the vehicle state vector.
According to the driver model, the driver’s intention ler, the desired yaw moment is determined by the refer-
can be directly reflected by the desired steering wheel ence driver model according to the road information.
angle dSW . However, the actual steering wheel angle In the lower controller, the actuator inputs are calcu-
dSW is the only measurable variable in the driver model. lated according to the desired yaw moment. Moreover,
Therefore, the current literatures approximately esti- a control boundary is computed in real time through an
mate the driver’s intention through the measured steer- embedded tire model ‘‘UniTire.’’
ing wheel angle dSW . The control target, desired yaw
rate, is set by
Upper controller
Gay Figure 7 is a single point block diagram of the optimal
g SW = dSW ð19Þ
Vx preview acceleration driver model,16 and the expression
of the optimal lateral acceleration can be represented
where Gay is the steady state cornering gain. by
In fact, it seems impossible that dSW will always
approximate to dSW , which can reflect the driver’s 2
intention. For instance, in some conditions, the vehicle €y = eðT P Þ ð20Þ
TP 2
runs into its nonlinear region, and its response changes
greatly compared with normal driving. The driver may with the lateral error e(TP ) defined by
not get used to this change and make an incorrect man-
euver due to panic. In such cases, using dSW to estimate e = f ðt + TP Þ yðtÞ TP y_ ðtÞ ð21Þ
6 Advances in Mechanical Engineering
where TP is the preview time and f (t + TP ) is the desired maneuver in the time domain. This advantage will fur-
path after TP . ther help the proposed system act earlier and timelier.
According to the driver model theory by Guo and Thus, the control target, desired yaw rate gdes , is
Fancher,17 the term 1 + Tc s is represented as the driv- related to dSW , by
er’s anticipation of the vehicle response information,
then Gay is the steady state cornering gain, expressed by Gay
g des = dSW ð26Þ
Vx
2
Vx
Gay = ð22Þ An approximate upper limit of the yaw rate can be
Li 1 + Ks Vx 2 written as
where i represents the steering system ratio and Ks is the mg
understeer gradient, defined by gdes ł ð27Þ
Vx
m a b This upper boundary of the control target is to guar-
Ks = 2 ð23Þ antee the desired yaw rate realization. Furthermore,
L Cr Cf
the precise control boundary is realized through a tire
The desired steering wheel angle dSW is thus related model in the lower controller.
to the lateral error e(TP ) by In order to get the vehicle yaw rate to track the con-
trol target more rapidly and accurately, sliding model
2Li 1 + Ks Vx 2 ð1 + Tc sÞ control is employed to deal with this control problem.
dSW = eðT P Þ ð24Þ Let the sliding surface be
Vx 2 TP 2
where K represents the steering wheel gain. s_ = g_ g_ des = f ðG Þ + g ðG Þ Myaw des ðG, uÞ g_ des ð29Þ
Figure 8 illustrates the relationships of K with with Myaw defined by
des
respect to Vx , K with respect to TP , and K with respect
to TP and Vx . Since the vehicle velocity Vx is always c c
Myaw des = Fxfr Fxfl cos d + ðFxrr Fxrl Þ ð30Þ
measured or estimated, and the preview time TP is 2 2
approximately a constant to a certain driver, K(TP , Vx )
where Fxij represents the longitude forces produced by
can be obtained in real time during driving. Therefore,
four wheels, s represents the steer angle, and c is the
dSW can be computed through the lateral error e(TP )
track width.
and the steering wheel gain K(TP , Vx ), as presented in
Therefore, the most ideal value of the desired yaw
equation (25). ^ yaw des that can satisfy s_ = 0 is
moment approximation M
The human driver’s steering wheel angle and the
controller prediction are compared on a certain path in 1
Figure 9. The human driver’s steering wheel angle is ^ yaw
M des = ½f ðG Þ g_ des ð31Þ
g ðG Þ
measured through the developed driving simulator. It
shows that the proposed method can precisely predict Here, we adopt the continuous relaxed condition to
the driver’s maneuver. Moreover, the predictive steer- avoid the shudder, and the desired yaw moment is
ing wheel angle is slightly ahead of the driver’s actual expressed by18
Wang et al. 7
Figure 8. Connections of steering wheel gain, preview time, and velocity: (a) K vs Vx at different TP values, (b) K vs TP at different
Vx values, and (c) K vs TP and Vx .
1 2
V= s ð33Þ
2
L
!
V = s s = s f ðGÞ + g ðGÞ Myaw des ðG, uÞ g des hs
= hs2 ł 0 ð34Þ
= hs2 ejsj ł 0
Myaw des
^ yaw
=M des hs ð32Þ ð36Þ
Here, we consider the Lyapunov function as Some vehicle states, including slip angle, tire forces,
and friction coefficient, required by the control law, can
8 Advances in Mechanical Engineering
1
Ks = af ar ð37Þ
ay L
Figure 11. Control torque distribution sequence: (a) understeer scenario and (b) oversteer scenario.
Wang et al. 9
Figure 13. Response of vehicle without control: (a) path, (b) steering wheel angle, (c) yaw rate, and (d) slip angle.
the other parameters, the comparison results are shown Figure 14(c) and (d) provides further evidence of the
in Figure 14, and the vehicle velocity is set as 120 km/h. superiority of the proposed DMbDYC system. The
During this DLC maneuver, both DMbDYC and yaw rate of the vehicle with DMbDYC is smooth and
SWbDYC stabilize the vehicle to pass the DLC path. the slip angle is small. By comparison, the SWbDYC
However, as shown in Figure 14, the DMbDYC approach fails to stabilize the vehicle. The yaw rate
approach is obviously much better than the SWbDYC changes sharply, and the slip angle is relatively large.
method. This means the vehicle is already somewhat unstable.
As shown in Figure 14(a), it is obvious that To sum up, the basic DYC controller is designed by
DMbDYC provides a much better control effect and the difference between the 2° model and the vehicle
the vehicle with the help of DMbDYC tracks the DLC model, which means that the error is not accurately
path accurately. The vehicle with the SWbDYC system, enough due to its simplicity, compared to the
however, has already slid out of the road. SWbDYC and the DMbDYC, which would lead the
The impact of the controller on the driver is control target a large fluctuation. Thus, the control sys-
shown in Figure 14(b): the driver steers the vehicle tem could not follow the desired yaw rate well and get
under the control of SWbDYC with a larger steering the worse response. And the SWbDYC choose the
wheel angle, which will probably bring the vehicle steering wheel angle to calculate the desired yaw rate,
response out of its linear region. In fact, during the which can present the driver’s intention. The
simulator driving, it is very hard to maneuver the vehi- DMbDYC desired the steering wheel angle to calculate
cle with the SWbDYC system, and the driver becomes the desired yaw rate, which can compensate the driver’s
panic due to the unfamiliar vehicle response. Under misoperation.
panic, it is possible for a driver to perform an incorrect
maneuver, which may lead to a very harmful and dan-
gerous result. In this sense, DMbDYC can reduce the Conclusion
possibility of an accident due to the small steering In the proposed controller, the reference driver model is
wheel angle. introduced to predict the driver’s intention according to
Wang et al. 11
Figure 14. Comparisons of driver model–based and steering wheel–based system: (a) path, (b) steering wheel angle, (c) yaw rate,
and (d) slip angle.
the road information. Since the control target is inde- When we drive on the low friction coefficient road,
pendent of the driver’s maneuver, the negative impacts the tire friction ellipse will shrink significantly.
by the driver’s reaction delay and incorrect maneuver Therefore, any control method can only provide a very
are reduced and counteracted. The driver-in-the-loop limit control effect. If possible, the most practical way
experiment results show that a much better control to stabilize the vehicle is by decelerating it in advance.
effect is achieved by the proposed DMbDYC system Once foreseeing the vehicle unstability caused by driv-
than the SWbDYC system. er’s maneuver, the driver model based controller
The longitudinal forces exerted in cornering alter the decelerates the vehicle before the maneuver happens.
vehicle steady state cornering characteristic. Taking The integrated control with the longitudinal control
advantage of this point, the distribution strategy of the added into the DYC system will be a subject of future
proposed controller reasonably determines the control work.
sequence to enhance the control effect and to reduce
the control power. Besides, the structure of IEV enables Declaration of conflicting interests
the distribution strategy to be more flexible. The IEV is
The author(s) declared no potential conflicts of interest with
not mandatorily accelerated or decelerated.
respect to the research, authorship, and/or publication of this
The tire model UniTire is embedded in the proposed
article.
controller to provide the real-time tire adhesive limit.
Since the tire forces can be always guaranteed to be in the Funding
friction eclipses, the DYC system can be always ensured
The author(s) disclosed receipt of the following financial
to be effective. In addition, the entire potential of each
support for the research, authorship, and/or publication
tire is separately exploited, such that the stable region of of this article: This paper is funded by the National
the full vehicle is maximized. Therefore, the control effect key research and development projects (Grant No.
of the proposed controller can be greatly improved. 2018YFB0104804).
12 Advances in Mechanical Engineering
References Appendix
1. Rajamani R. Vehicle dynamics and control. New York: Notation
Springer Science + Business Media, 2006.
2. Nagai M, Hirano Y and Yamanaka S. Integrated control a distance from center of gravity to front
of active rear wheel steering and direct yaw moment con- axle
trol. Vehicle Syst Dyn 1997; 27: 357–370. ay lateral acceleration at center of gravity
3. Osborn RP and Shim T. Independent control of all- b distance from center of gravity to rear axle
wheel-drive torque distribution. SAE paper 2004-01- c track width
2052. Cf front tire cornering stiffness
4. Yoshioka T, Adachi T, Butsuen T, et al. Application of Cr rear tire cornering stiffness
sliding-mode theory to direct yaw-moment control.
Fx =Fxij longitudinal tire force
JSAE Rev 1999; 20: 523–529.
Fxm tire longitudinal force potential extreme
5. Park K, Heo S and Baek I. Controller design for improv-
ing lateral vehicle dynamic stability. JSAE Rev 2001; 22:
value
481–486. Fy =Fyij lateral tire force
6. Boada BL, Boada MJL and Dı́az V. Fuzzy-logic applied Fym tire lateral force potential extreme value
to yaw moment control for vehicle stability. Vehicle Syst Fz tire vertical load
Dyn 2005; 43: 753–770. g standard gravity
7. Davoudi M, Menhaj MB and Davoudi M. A fuzzy based Gay steady state gain of vehicle
vehicle dynamic stability control (FDSC). SAE paper i steering system ratio
2006-01-3483. Iw rotational moment of inertia of wheel
8. Falcone P, Tseng HE, Borrelli F, et al. MPC-based yaw Iz yaw moment of inertia of vehicle
and lateral stabilisation via active front steering and K steering wheel gain
braking. Vehicle Syst Dyn 2008; 46: 611–628. Ks understeer gradient
9. Yoon J, Yim S, Cho W, et al. Design of an unified chas-
Kx tire longitudinal slip stiffness
sis controller for rollover prevention, manoeuvrability
Ky tire cornering stiffness
and lateral stability. Vehicle Syst Dyn 2010; 48:
1247–1268. L wheel base
10. Song J. Enhanced braking and steering yaw motion con- m total mass of vehicle
trollers with a non-linear observer for improved vehicle Myaw des desired yaw moment
stability. Proc IMechE, Part D: J Automobile Engineering r wheel effective radius
2008; 222: 293–304. s sliding surface
11. Chen Y, Hedrick JK and Guo K. A novel direct yaw Sx tire longitudinal slip ratio
moment controller for in-wheel motor electric vehicles. Sy tire lateral slip ratio
Vehicle Syst Dyn 2013; 51: 925–942. td neural delay time
12. Mechanical Simulation Corporation. CarsimÒ reference Tc driver correction time constant
manual, Version 8.02. Ann Arbor, MI: Mechanical Simu- Tij torque on wheel
lation Corporation, 2009.
Tn physical delay time constant
13. National Instruments Corporation. LabVIEWTM 2011
TP preview time
help. Austin, TX: National Instruments Corporation,
2011.
u vehicle control inputs
14. Guo K, Lu D, Chen S, et al. The UniTire model: a non- vx longitudinal velocity of wheel center
linear and non-steady-state tyre model for vehicle vy lateral velocity of wheel center
dynamics simulation. Vehicle Syst Dyn 2005; 43: 341–358. Vx longitudinal velocity of vehicle at center of
15. Guo K and Lu D. UniTire: unified tire model for vehicle gravity
dynamic simulation. Vehicle Syst Dyn 2007; 45: 79–99. Vy lateral velocity of vehicle at center of
16. Guo K and Guan H. Modelling of driver/vehicle direc- gravity
tional control system. Vehicle Syst Dyn 1993; 22: €y optimal lateral acceleration
141–184.
17. Guo K and Fancher P. Preview-follower method for af front tire slip angle
modelling closed-loop vehicle directional control. In: Pro- ar rear tire slip angle
ceedings of the 19th annual conference on manual control, g yaw rate
Cambridge, MA, 1 January 1983; pp.159–186. g des desired yaw rate
18. Slotine J-JE and Li W. Applied nonlinear control. Engle- g SW desired yaw rate with driver delay
wood Cliffs, NJ: Prentice Hall, 1991.
Wang et al. 13