You are on page 1of 29

International Journal of Control

ISSN: 0020-7179 (Print) 1366-5820 (Online) Journal homepage: http://www.tandfonline.com/loi/tcon20

Lateral Motion Stability Control of Electric Vehicle


via Sampled-data State Feedback by Almost
Disturbance Decoupling

Qinghua Meng, Zong-yao Sun, Yong Shu & Tingting Liu

To cite this article: Qinghua Meng, Zong-yao Sun, Yong Shu & Tingting Liu (2017): Lateral Motion
Stability Control of Electric Vehicle via Sampled-data State Feedback by Almost Disturbance
Decoupling, International Journal of Control, DOI: 10.1080/00207179.2017.1367104

To link to this article: http://dx.doi.org/10.1080/00207179.2017.1367104

Accepted author version posted online: 16


Aug 2017.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=tcon20

Download by: [Australian Catholic University] Date: 17 August 2017, At: 19:38
Publisher: Taylor & Francis
Journal: International Journal of Control
DOI: https://doi.org/10.1080/00207179.2017.1367104

ARTICLE TEMPLATE
Lateral Motion Stability Control of Electric Vehicle via
Sampled-data State Feedback by Almost Disturbance Decoupling
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

Qinghua Menga and Zong-yao Sunb and Yong Shua and Tingting Liua
a
School of Mechanical Engineering, Hangzhou Dianzi University, Hangzhou, China; b

Institute of Automation, Qufu Normal University, Qufu,China

ARTICLE HISTORY
Compiled August 11, 2017

ABSTRACT
The paper presents a lateral motion stability control method for electric vehicle
driven by four in-wheel motors which considers time-variable vehicle speed and un-
certain disturbance caused by external factors. Firstly, an EV lateral motion dynam-
ics tracking control model is presented. Then in order to deal with the uncertain
disturbance in the lateral motion model, an Almost Disturbance Decoupling method
using sampled-data state feedback is proposed. Thirdly, a sampled-data state feed-
back controller is constructed based on the state feedback domination approach. The
proposed controller can attenuate the disturbances’ effect on the output to an arbi-
trary degree of accuracy. Simulation and test results under different vehicle speeds
show the effectiveness of the control method.

KEYWORDS
Electric Vehicle; Lateral Motion Stability; Sample-data State Feedback; Almost
Disturbance Decoupling

1. Introduction

Because of the advantages of Electric Vehicle(EV) driven by in-wheel motors in terms


of energy efficiency, environmental protection, performance benefits and so on, espe-
cially driven by four independent in-wheel motors, more and more researchers and
engineers pay more attention to the technologies such as vehicle motion control, en-
ergy optimization, and vehicle structural arrangement ( Wang et al. (2011), Nikam et
al. (2012), Xiong et al. (2012), Shuai et al. (2014), Mansouri et al. (2016)). Vehicle
motion stability control is one of these core technologies ( Kawashima et al. (2010),
Wu et al. (2013), He et al. (2013), Ko et al. (2015)). Unexpected disturbances are

CONTACT Qinghua Meng. Email: mengqinghua@hdu.edu.cn


generated because of maneuvers on severe road conditions, siding wind forces, differ-
ent tire pressures and so on when an EV runs, that may result in dangerous lateral
motions. It is difficult for drivers to react extremely quickly in dangerous conditions.
Consequently, improvement of EV lateral motion dynamics by active vehicle control
to avoiding such catastrophic situations has been and is continuing to be a subject
of active research (Hori (2004), Nam et al. (2012), Chen and Kuo (2014)). However,
most of these control algorithms are designed for conventional vehicle architecture. In
the recent years, a few researchers proposed several lateral motion control methods.
For example, a direct yaw-moment control system for a four-in-wheel motors-driven
electric vehicle was proposed in ( Geng et al. (2009)). Four wheel driven systems can
provide vehicle stability by using the braking and acceleration abilities of independent
in-wheel electric motors on each wheel in ( Sakai et al. (2002) and Li et al. (2009)).
Two different approaches, one was a novel integrated lateral stability control system
and the second was a regenerative braking based lateral stability control system, were
used in (Emırler et al. (2015)).
As aforementioned, when a vehicle runs in cruise condition, especially in high speed
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

cruise condition, many factors will generate uncertain disturbance to the body and
front steering wheels, which will influence the lateral stability of the electric vehicle
seriously. The aforementioned control methods usually did not include the uncertain
disturbance or considered the disturbance as known to simplify the model and com-
putation. In this paper, the parameters of the vehicle are obtained by sensors. So
this paper studies the problem of disturbance attenuation with internal stability by
designing a sampled-data state feedback controller, called Almost Disturbance Decou-
pling(ADD), to control the lateral motion stability of EV under high vehicle speed.
The control objective of the paper is to design a sampled-data state controller which
ensures Almost Disturbance Decoupling in the sense of L2 gain with global asymptotic
stability. Now the ADD problem has attracted more and more researchers, and great
achievements have been made for both linear and nonlinear systems ( Ozcetin et al.
(1992), Marino and Tomei (2000), Qian and Wei (2003), Chien et al. (2008)). Dis-
turbance observer-based control has been proven to be effective in compensating the
effects of unknown external disturbances for nonlinear systems in ( Yang et al. (2012),
Yang et al. (2013), Sun et al. (2016), Li et al. (2017)).
However, the aforementioned achievements were achieved using continuous-time
controllers, despite the fact that more and more systems are implemented digitally
in practice. It is necessary to study solving the ADD problems by sampled-data con-
trollers.
This paper proposes a controller via sampled-data state feedback for almost distur-
bance decoupling problem to keep the lateral motion stability of EV driven by four
in-wheel motors. Parameter L will be used to dominate the linearities, nonlinearities
and uncertain disturbance. The main contributions of this paper lie in the following
aspects.
(1) The time-variable vehicle speed and uncertain external disturbances are included
in the EV dynamic model.
(2) A sampled-data controller achieving global asymptotically stable is designed
without disturbance and dominating the disturbance’s effect on the output using
parameter L.
The paper is organized as follows. EV lateral motion dynamic model considering
time-variable vehicle speed and uncertain disturbances is presented in Section 2. Sec-
tion 3 presents the problem formulation and some preliminaries about sampled-data

2
feedback controller and ADD. A sampled-data controller is designed in Section 4.
Section 5 presents a torque allocation method for four in-wheel motors. Section 6
simulates and test the controller for lateral stability of an EV under different vehicle
speed, which is followed by the conclusion in Section 7.

2. The lateral motion dynamic Model of EV

The lateral motion model of EV driven by four in-wheel motors simplified to two
wheels model is shown in Figure 1. Different from the conventional car structures,
each wheel is independently driven by an in-wheel motor, so an external yaw moment
can be easily generated to regulate the car yaw and lateral motions because of the
fast and precise torque response of motors. To design the controller, the widely used
desired two-wheels vehicle lateral dynamics model, ignoring the pitch and roll motions,
can be expressed as follows:
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

Fyf + Fyr
β̇d = − γd
mvx
(1)
Lf Fyf − Lr Fyr + Mzd
γ̇d = ,
Iz

where βd is the desired vehicle sideslip angle, γd is the desired yaw rate, Fyf and Fyr
are the front and rear tire lateral forces, vx is the vehicle longitudinal speed, m and
Iz are car mass and yaw inertia, Lf and Lr are distances between center of gravity
and front wheel center and rear wheel center. Mzd is the desired external yaw moment
generated by longitudinal tire forces among the left and right side wheels.
Lateral forces of front and rear tires are obtained according to the following linear
tire model used in (Wang et al. (2015)):

Fyf = Cf αf , Fyr = Cr αr , (2)

where Cf and Cr are the front and real tire cornering stiffness values, αf and αr are
the wheel slip angles which can be calculated as

   
L f γd Lr γ d
αf = − β d + − δ(t) , αr = − βd − , (3)
vx vx

where δ(t) is the front wheel steering angle varies with time. Based on Equ.( 2) and
Equ.(3), system (1) can be rewritten as

(Cf + Cr )βd (Cr Lr − Cf Lf )γd Cf


β̇d = − + 2
− γd + δ(t)
mvx mvx mvx
(4)
(Cr Lr − Cf Lf )βd (Cf Lf + Cr Lr )γd Mzd Cf Lf
2 2
γ̇d = − + + δ(t).
Iz Iz v x Iz Iz

In practice, an EV is disturbed by external factors when it runs on road. So the


practical 2 DOF EV model can be written as

3
(Cf + Cr )βp (Cr Lr − Cf Lf )γp Cf
β̇p = − + 2
− γp + (δ(t) + Δδ(t))
mvx mvx mvx
(5)
(Cr Lr − Cf Lf )βp (Cf Lf + Cr Lr )γp Mzp Cf Lf
2 2
γ̇p = − + + (δ(t) + Δδ(t)) .
Iz Iz vx Iz Iz

In Equ.(5), βp is the practical vehicle sideslip angle, γp is the practical yaw rate,
Δδ(t) is uncertain disturbance which is generated by severe road conditions, siding
wind forces or different tire pressures and so on when a car runs, Mzp is the necessary
practical external yaw moment. The researchers generally regarded vehicle speed vx
as a constant in order to simply the analysis. In fact the vehicle speed varies when
the car runs. In order to reflect the real conditons, this paper takes consideration of
the case that the involved vx should be time-varing. In an analytical way, vx can be
formulated with uncertainties as the following form.
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

1 1
ρ1 (t) = , ρ2 (t) = 2 . (6)
vx vx

Although vx is time-varing, it has a limited highest value.Therefore

ρ1 (t) ≤ θ1 , ρ2 (t) ≤ θ2 , ρ1 (t) > 0, ρ2 (t) > 0. (7)

Substracting Equ.(4) from Equ.(5), we obtain

(Cf + Cr )eβ (Cr Lr − Cf Lf )eγ Cf


ėβ = − + 2
− eγ + Δδ(t),
mvx mvx mvx
(8)
(Cr Lr − Cf Lf )eβ (Cf L2f + Cr L2r )eγ ΔMz C f Lf
ėγ = − + + Δδ(t).
Iz Iz v x Iz Iz

where ėβ = β̇p − β̇d , ėγ = γ̇p − γ̇d , eβ = βp − βd , eγ = γp − γd . ΔMz = Mzp − Mzd is
the extra necessary moment to keep an EV stability.
Let x(t) = (x1 (t), x2 (t))T ∈ R2 be system state, u(t) ∈ R, y(t) ∈ R, and d(t) ∈ R
be the system input, output, and uncertain disturbance input respectively. Under the
following notations

x1 (t) = eβ , (9)
(Cr Lr − Cf Lf )ρ2 (t) − m
x2 (t) = eγ , (10)
m
(Cr Lr − Cf Lf )ρ2 (t) − m
u(t) = ΔMz , (11)
mIz
d(t) = Δδ(t), (12)

4
system (8) can be rewritten as follows

ẋ1 (t) = x2 (t) + φ1 (t, x(t)) + g1 d(t),


ẋ2 (t) = u(t) + φ2 (t, x(t)) + g2 d(t), (13)
y(t) = x1 (t),

where
Cf
g1 (t) = ρ1 (t), (14)
m
(Cr Lr − Cf Lf )ρ2 (t) − m Cf Lf
g2 (t) = , (15)
m Iz
Cf + Cr
φ1 (t, x(t)) = − ρ1 (t)x1 (t), (16)
m
[(Cr Lr − Cf Lf )ρ2 (t) − m] (Cr Lr − Cf Lf )
φ2 (t, x(t)) = x1 (t)
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

mIz
" #
Cr L2r + Cf L2f (Cr Lr − Cf Lf )ρ̇2 (t)
+ − ρ1 (t) + x2 (t). (17)
Iz (Cr Lr − Cf Lf )ρ2 (t) − m

3. Problem Formulation and Preliminaries

The objective of this paper is to design a sampled-data state feedback controller for EV
lateral motion stability such that the resulting closed-loop system is globally asymp-
totically stable at the origin when d(t) = 0 and also satisfies almost disturbance
attenuation for disturbance to keep the lateral stability. From Equ. ( 7), Equ.(16) and
Equ.(17), it is obvious that a constant c > 0 always exists such that

Cf + Cr

|φ1 (t, x(t))| = ρ1 (t)x1 (t) 6 c |x1 (t)| , (18)
m

|φ2 (t, x(t))| 6 c(|x1 (t)| + |x2 (t)|). (19)

On the other hand, it is also obvious that there always exists a known constant
G0 > 0 such that

|gi | 6 G0 , ∀t > 0, i = 1, 2. (20)

According to the aforementioned Equ.( 18), Equ.(19) and Equ.(20), we can design
an Almost Disturbance Decoupling(ADD) control via sampled-data state feedback.
Given a real number γ > 0, a linear sampled-data feedback controller u(t) is designed
as

x(tk+1 ) = Ax(tk ) + By(t)


(21)
u(t) = −Kx(tk ), ∀t ∈ [tk , tk+1 ), tk = kT, k = 0, 1, 2, . . . ,

where the time instant tk and tk+1 are the sampling points and T is the sampling
period, such that the following holds:

5
(1) when d(t) = 0, the closed-loop system ( 13) is globally asymptotically stable at
the equilibrium x = 0;
(2) For disturbance d(t), the response of the closed-loop system ( 13) starting from
the origin satisfies

Z t Z t
2
|y(s)| ds 6 γ 2
|d(s)|2 ds, (22)
0 0

where t is the duration of disturbance. Generally


Rt speaking, the energy of disturbance
generated to an EV is limited, that means 0 |d(s)|2 ds is bounded.

4. Controller Design

In this section, the work is to design a sampled-data state feedbak controller to solve
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

the Almost Disturbance Decoupling problem for system ( 1). The main result is de-
scribed as the following theorem.

Theorem 4.1. The problem of ADD for system ( 13) can be solved by a sampled-data
state feedback controller in the form of ( 21).

Proof
The contruction of the sampled-data state feedback controller can be divided into
three steps.The first step is to change the coordinates for system ( 13). The second step
is to construct a linear sampled-data controller. Finally, the scaling gain and sampling
period are determined to globally stabilize the system ( 13) with the help of domination
approach.

4.1. Pre-treatment of system (13): change of coordinates


First of all, a change of coordinates for system ( 13) is introduced to enable the comple-
tion of a state feedback solution. For a constant L ≥ 1 to be determined later, define
the following change of coordinates:
xi u
zi = i−1
, v = 2 , i = 1, 2. (23)
L L

Under this coordinates change, system ( 13) becomes

ż1 (t) = Lz2 (t) + φˉ1 (t, z(t)) + gˉ1 (t)d(t)


ż2 (t) = Lv(t) + φˉ2 (t, z(t)) + gˉ2 (t)d(t) (24)
y(t) = z1 (t),

where φˉi (t, z(t)) = φi (t, z(t))/Li−1 , i = 1, 2, gˉi (t) = gi (t)/Li−1 , i = 1, 2. From
Equ.(18), Equ.(19) and Equ.(20), it can be verified that for i = 1, 2
c 
φˉi (t, z(t)) 6 zˉ1 (t) + Li−1 zˉ2 (t) 6 c(z1 (t) + ∙ ∙ ∙ + z2 (t)), (25)
Li−1

6
G0
gi (t)| 6
|ˉ 6 G0 . (26)
Li−1

Define
         
z1 (t) 0 1 0   φˉ1 (∙ ) gˉ1 (∙ )
z(t) = ,A = ,B = ,C = 1 0 , φ(∙ ) = , G(∙ ) = ,
z2 (t) 0 0 1 φˉ2 (∙ ) gˉ2 (∙ )

under which system (24) can be rewritten as

ż(t) = LAz(t) + LBv(t) + φ(∙) + G(∙)d(t)


(27)
y(t) = Cz(t).

From (Qian and Du (2012) and Chu et al. (2015)), the system (27) with a sampler is
equivalent to the following discrete-time equation:
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

Z T
z(tk+1 ) = e LAT
z(tk ) + eLAs ds [LBv(tk ) + LHy(tk )] , (28)
0

where H = [a1 , a2 ]T .
The authors still use the equation ( 27) in the following section instead of ( 28) for
the convenience of stability analysis.

4.2. Construction of a Sampled-data state Feedback Control Law


From above, the controller law is constructed as

v(t) = v(tk ) = −Kz(tk ) = −k1 z1 (tk ) − k2 z2 (tk ), ∀t ∈ [tk , tk+1 ) , k = 0, 1, 2, . . . , (29)

where k1 and k2 are coefficients of the Hurwitz polynomial p2 (s) = s2 + k2 s + k1 and


both non-zero positive.
Construct a Lyapunov function V (z) = z T P z, where matrices P = P T > 0 and
A P + P A = −I. The derivative of V (z) along system (27) is
T

V̇ (z(t)) = −Lkz(t)k2 + 2Lz(t)T P Bv(tk ) + 2z T P φ(∙) + 2z T P G(∙)d(t)


6 −L(1 + 2KkP kB)kz(t)k2 + 2LKz(t)T (z(t) − z(tk ))P B + 2z T P φ(∙)
+ 2z T P G(∙)d(t). (30)

In the next setp, the paper will present that by choosing the scaling gain L and
sampling period T , the closed-loop system ( 27) can be rendered globally asymptotically
stable.

4.3. Determination of Gain L and Computation of T


Before determining L and T , it needs to obtain the estimates for some terms in ( 30).
According to (25), the following holds
q
kφ(∙ )k = φˉ21 + φˉ22 6 c1 kz(t)k, (31)

7

where c1 = c 1 + 2 ≥ 0. With the help of (31), we have

k2z T P φ(∙ )k ≤ 2c1 kP kkz(t)k2 . (32)

According to (26), the following holds

1
k2z T P G(∙ )kd(t) ≤ 4kP k2 G20 kz(t)k2 + kd(t)k2 . (33)
2

By (30) and (31), it is straightforward to verify that

kż(t)k ≤ c2 kz(t)k + c3 kz(tk )k + c4 kd(t)k, ∀t ∈ [tk , tk+1 ) . (34)

where c2 = LkAk+c1 , c3 = LkBKk and c4 = kG(∙)k. With the help of (34), integrating
(30) from tk to t yields
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

Z t
kz(t) − z(tk )k ≤ kz(τ )dτ
tk
Z t
≤ (c2 kz(τ )k + c3 kz(tk )k + c4 kd(t)k)dτ
tk
Z t Z t
≤ (c2 kz(τ ) − z(tk )k + (c2 + c3 )kz(tk )k)dτ + c4 kd(t)kdτ. (35)
tk tk

Define
Z t
M(t) = (c2 kz(τ )k + c3 kz(tk )k)dτ. (36)
tk

Taking derivative of M(t) yields

Ṁ(t) =c2 kz(τ )k + c3 kz(tk )k


≤c2 kz(τ ) − z(tk )k + (c2 + c3 )kz(tk )k
Z t
≤c2 M(t) + c2 c4 kd(τ )kdτ + (c2 + c3 )kz(tk )k. (37)
tk

By general solution of Homogeneous Linear Equations, Solving ( 37) for t ∈ [tk , tk+1 )
with M(tk ) = 0, we can get
Z t  Z τ  
c2 (τ −tk )
M(t) ≤e c2 (t−tk )
e c2 c4 kd(s)kds + (c2 + c3 )kz(tk )k dτ
tk tk
h i Z t
c 2 + c3
= kZ(tk )k e c2 (t−tk )
− 1 − c4 kd(s)kds
c2 tk
Z t
+ c4 ec2 (t−tk ) (e−c2 (τ −tk ) kd(τ )kd(τ ). (38)
tk

8
Substituting (38) into (35), the following holds
h i Z t
c2 + c3
kz(t) − z(tk )k ≤ kz(tk )k e c2 (t−tk )
− 1 − c4 kd(s)kds
c2 tk
Z t Z t
−c2 (τ −tk )
+ c4 e c2 (t−tk )
(e kd(τ )kdτ + c4 kd(τ )kdτ
tk tk
c2 + c3 h c2 (t−tk ) i
≤ e − 1 (kz(t) − z(tk )k + kz(t)k)
c2
Z t
+ c4 ec2 (t−tk ) kd(τ )kdτ. (39)
tk

 
If T is chosen small enough, denoting δ(t) = c2 +c3
c2 ec2 (t−tk ) − 1 , there exists
Z
δ(t − tk ) c4 ec2 (t−tk ) t
kz(t) − z(tk )k ≤ kz(t)k + kd(τ )kdτ. (40)
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

1 − δ(t − tk ) 1 − δ(t − tk ) tk

Substituting (32), (33), and (40) into (30) results in

δ(t − tk )
V̇ (Z(t)) ≤ − L(1 + 2KkP kB)kz(t)k2 + 2LKkP kB kz(t)k2
1 − δ(t − tk )
Z t
c4 ec2 (t−tk )
+ 2LkP kkBkz(t)k kd(τ )dτ
1 − δ(t − tk ) tk
1
+ 2c1 kP kkz(t)k2 + 4kP k2 G20 kz(t)k2 + kd(t)k2 . (41)
2

According to (24), we have

1 1 1
|y(t)|2 = 2 kCz(t)k2 ≤ 2 kz(t)k2 . (42)
γ2 γ γ

According to (41) and (42), we have

1
V̇ (Z(t)) + |y(t)|2 − kd(t)k2
γ2
δ(t − tk )
≤ − L(1 + 2KkP kB)kz(t)k2 + 2LKkP kB kz(t)k2
1 − δ(t − tk )
Z t
c4 ec2 (t−tk )
+ 2LkP kkBkz(t)k kd(τ )dτ
1 − δ(t − tk ) tk
+ 2c1 kP kkz(t)k2 + 4kP k2 G20 kz(t)k2
1 1
+ 2 kZ(t)k2 − kd(t)k2 . (43)
γ 2

Considering the stability of system ( 30), the L can be fixed appropriately as

1
L = 2 + 2c1 kP k + 4kP k2 G20 + . (44)
γ2

9
T is chosen as a small enough constant such that δ(t − tk ) ≤ δ(T ). Substituting (44)
into (43) and integrating both sides of ( 43) from tk to tk+1 result in
Z tk+1
1
(V̇ (z(t)) + |y(t)|2 − kd(t)k2 )dt
tk γ2
Z tk+1
≤ −2(1 + LKkP kB) kz(t)k2
tk
Z
δ(T ) tk+1
+ 2LΓ1 kz(t)k2 dt
1 − δ(T ) tk
Z tk+1 Z t
c 4 ec2 T
+ 2LΓ1 kz(t)k kd(τ )kdτ dt
1 − δ(T ) tk tk
Z tk+1 Z
1 tk+1
+ 2Lλmax (P ) 2
kz(t)k dt − kd(t)k2 dt, (45)
tk 2 tk
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

where

0 BK
Γ1 = λmax (P )
−HC BK .

According to Cauchy-Schwarz Inequality, we have


Z tk+1 Z t Z tk+1 Z t
kz(t)k kd(τ )kdτ dt ≤ kz(t)kdt kd(t)kdt
tk tk tk tk
Z Z
1 tk+1 1 t
≤ ( kz(t)kdt)2 + ( kd(t)kdt)2
2 tk 2 tk
Z tk+1 Z tk+1 Z Z tk+1
1 1 t
≤ 2
kz(t)k dt 1dt + 2
kd(t)k dt 1dt
2 tk tk 2 tk tk
Z tk+1 Z t
1 1
= T kz(t)k dt + T
2
kd(t)k2 dt. (46)
2 tk 2 tk

Substituting (46) into (45) results in


Z tk+1
1
(V̇ (z(t)) + 2 |y(t)|2 − kd(t)k2 )dt
tk γ
    Z tk+1
δ(T ) c 4 e c2 T
≤ −2 − 2LKBkP k + 2L +T Γ2 kz(t)k2 dt
1 − δ(T ) 1 − δ(T ) tk
  Z tk+1
c4 e c 2T
1
+ T LΓ2 − kd(t)k2 dt, (47)
1 − δ(T ) 2 tk

where

0 BK

Γ2 = kP k .
−HC BK

10
As aforementioned T is chosen as a small enough constant such that
 
δ(T ) c 4 ec2 T
− 1 − 2LKBkP k + 2L +T Γ2 < 0, (48)
1 − δ(T ) 1 − δ(T )

c 4 ec 2 T 1
T LΓ2 − < 0. (49)
1 − δ(T ) 2

When the aforementioned two criteria ( 48) and (49) are satified simultaneously, the
right side of (47) is negative definite, that also means
Z Z
tk+1
1 tk+1
(V̇ (z(t)) + 2 |y(t)|2 − kd(t)k2 )dt ≤ − kz(t)k2 dt, (50)
tk γ tk
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

from which it can be concluded that the equilibrium of the closed-loop system is
uniformly globally asymptotically stable when d(t) = 0.
Moreover, the following can be obtained based on ( 50)

X Z tk+1
n−1
1 2
X n−1 Z tk+1
V (z(tn )) − V (z(t0 )) + |y(t)| dt − kd(t)k2 dt ≤ 0. (51)
tk γ2 tk
k=0 k=0

When z(t0 ) = 0, based on the fact that V (z(tn )) ≥ 0 and V (0) = 0, it can be
concluded from (51) that

X Z tk+1
n−1
1 X n−1 Z tk+1

2
|y(t)|2 dt ≤ kd(t)k2 dt. (52)
tk γ tk
k=0 k=0

Because of t0 = 0 and letting tn−1 −→ t, the following inequality holds


Z t Z t
2
|y(τ )| dτ ≤ γ 2
kd(τ )k2 dτ. (53)
0 0

Then the almost disturbance decoupling in ( 22) are proved.


According to (48) and (49), the following hold.
 
1 + 2LKBkP k − T c4 ec2 T Γ2
δ(T ) ≤ min , 1 − 2T Lc4 e Γ2 .
c2 T
(54)
1 + 2LKBkP k + 2LΓ2

Define
 
1 + 2LKBkP k − T c4 ec2 T Γ2
W = min , 1 − 2T Lc4 e Γ2 .
c2 T
1 + 2LKBkP k + 2LΓ2

Based on (54), the sampling period T can be calculated as

1 (W + 1)c2 + c3
T ≤ ln . (55)
c2 c 2 + c3

11
Remark 1. Substituting the control law ( 29) into (28), there exists
 Z T  Z T
ẑ(tk+1 ) = e LAT
− e LAs
dsLBK z(tk ) + eLAs dsLHy(tk ) (56)
0 0

It is in the form of the discrete-time observer shown in ( 21). It is clear that there is
a relation between T and L from (55). From (44),it can be concluded that L can be
arbitrarily large, but a large L will lead to a smaller sampling period T . On the other
hand, (55) will give a very small T because of the conservative estimations used to
simplify the proof. it is no longer a hurdle in implementing the proposed sampled-data
cntroller using a very small sampling period.

5. Torque Allocation
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

Generally speaking, an EV will not skid easily with enough adhesion force. The sum
of every proportion between each tire’s added driving force and adhesion force should
be minimum if the EV runs more steadily, i.e.

Fx f l Fxf r Fxrl Fxrr


minJ = | |+| |+| |+| |. (57)
u f l Fz f l u f r Fzf r url Fzrl urr Fzrr

where Fxf l , Fxf r , Fxrl , Fxrr are the added driving forces of front-left tire, front-right
tire, rear-left tire and rear-right tire respectively, Fzf l , Fzf r , Fzrl , Fzrr are the vertical
forces of front-left tire, front-right tire, rear-left tire and rear-right tire respectively,
uf l , uf r , url , urr are the adhesion coefficients of front-left tire, front-right tire, rear-left
tire and rear-right tire respectively.
We assume that all the tires’ adhesion coefficients are equal. The vertical forces of
front wheels and real wheels are
9.8Lr
u f l F z f l = u f r F z f r = Fz f = m
2(Lf + Lr )
9.8Lf
url Fzrl = urr Fzrr = Fzr = m (58)
2(Lf + Lr )

F xf Fz f
Therefore, the stability of the EV is better when F xr = Fz r , i.e.

Lr
Fxf = Fx (59)
Lf r

where Fxf is the added force of front wheel, Fxr is the added force of rear wheel.
The sum of torques generated by every tire’s added force is equal to the added yaw
moment, i.e.

d
T z = Mz = (−Fxf l + Fxf r − Fxrl + Fxrr ), (60)
2

12
where d is the track width. Therefore,

−Lr
Fxf l = −Fxf r = Tz
(Lf + Lr )d
−Lf
Fxrl = −Fxrr = Tz (61)
(Lf + Lr )d

Considering the added force and the origin driving force of every wheel, the longitudial
force of every wheel should be

F´xi = Fti + Fxi . (62)

where Fti is the origin driving force of ith wheel, Fxi is the added force of ith wheel.
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

6. Simulation and Test Study

In this section the simulation of the control system ( 27) was conducted in MATLAB.
The EV parameters are listed in Table 1. Generally, the range of the high speed vx
is about from 80km/h to 120km/h. When an EV runs under a constant speed, the
vehicle speed is allowed to vary about ±3%.
From (13), the simulated EV’s lateral motion control system is

ẋ1 (t) = x2 (t) + 38.571ρ1 (t)x1 (t) + 18.571ρ1 (t)d(t)


 
8ρ̇2 (t)
ẋ2 (t) = u(t) + (0.0016ρ2 (t) − 1)x1 (t) + 21.6ρ1 (t) + x2 (t)
8ρ2 (t) − 7
+ (14.857ρ2 (t) − 13)d(t),

where u(t) is the control torque. sin(t)/(1 + t2 ) is used as the disturbance d(t) in the
simulaion, shown in Fig.2. There are c ≥ 1.996 and G0 ≥ 19.96 to meet (18), (19)and
(20). The sampled-data controller is

u(t) = L2 v(kT ) = −17L2 ẑ1 (kT ) − 13L2 ẑ2 (kT ). (63)

In this section, we carried out the EV stability control simulations under 120km/h
and 80km/h. In the simulations, the EV runs on the good city road. The initial values
are set z(0) = [0, 0]T and ẑ(0) = [0, 0]T in the numerical simulation. In the simulation,
there is no disturbance applied to the EV from time 0 to 1th second when the EV
runs, and the controller is not active. The disturbance is applied to the EV from 1th
to 3th second, and the controller works to control the EV’s lateral motion stability.
According to (9) and (10), we map controller’s states x1 and x2 into EV’s parameters
eβ and eγ . Fig.3 to Fig.5 show the simulation results when the speeds are 120km/h
and 80km/h. The simulation results show the effectiveness of the sampled-data state
feedback controller. The controller’s inputs are shown in Fig. 3. The controller can
stabilize the EV’s lateral motion under different vehicle speeds as shown from Fig. 4
to Fig.5. The system state eβ and the system state eγ are stabilized to zero within 2
seconds. The overshoots of vehicle sideslip angle and yaw rate are small and attenuate
rapidly. On the other hand, we can obtain that the max overshoots of eβ and eγ under
120km/h are both smaller than under 80km/h, that is correct because the eβ and eγ

13
need to be adjusted smaller under higher vehicle speed, otherwise the EV may become
unstable. The force allocations of every tire under 120km/h and 80km/h are shown in
Fig. 6 and Fig.7.
We also carried out tests to verify the controller’s effectiveness by an EV driven
by four in-wheel motors designed by ourselves, as shown in Fig. 8. We just tested
the controller under 50km/h for safety. The controller’s input is shown in Fig. 9. The
simulation and test results are shown in Fig. 10 and Fig.11, from which one can obtain
that the test results are agreeable with the simulation results.

7. Conclusions

In this paper, the lateral motion stability control of the EV driven by four in-wheels
motors with uncertain disturbance under high vehicle speed has been solved based on
ADD via a sampled-data state feedback controller, which is not only more practical in
real practice but also more difficult to design. Firstly, the lateral motion dynamic model
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

is built considering the time-variable vehicle speed and uncertain disturbance. Then
the sampled-data state feedback controller is constructed to solve the ADD problem for
nonlinear system which was solved usually by a continuous-time controller. The state
feedback domination approach is also used to dominate the uncertain nonlinearities
and disturbance by using a scaling gain. With the help of these tools, the sampling
period is calculated to guarantee global stability and disturbance attenuation. Then a
torque allocation method for four in-wheel motors of th EV is presented. Finally, the
simulations are conducted to prove the effectiveness of the sampled-data state feedback
controller for EV’s lateral motion stability under 120km/h and 80km/h vehicle speeds.
And the tests are carried out also under 50km/h to verified the simulation results. The
work of this paper proposes a new EV stability control method for the subject area
which has good anti-disturbance and robust performance, and can be used in the EV
more easily by using ADD method and linear sampled-data state feedback controller.
In this paper, the authors just take consideration of the case that the involed vx is
time-varing. In fact, the car mass m, cornering stiffness values Cf and Cr , and yaw
inertia Iz are also variable when an EV runs. Therefore, these parameters could be
considered as variables in the future research works despite many difficulties .

Acknowledgment(s)

This research was supported by Zhejiang Provincial Natural Science Foundation of


China under Grant No.LY16E050003.

14
References

Chen, B.C., and Kuo, C.C. (2014), “Electronic stability control for electric vehicle
with four in-wheel motors,” International Journal of Automotive Technology, 15,
573–580.
Chien, T.L., Chen, C.C., Huang, Y.C., and Lin, W.J. (2008), “Stability and Almost
Disturbance Decoupling Analysis of Nonlinear System Subject to Feedback Lin-
earization and Feedforward Neural Network Controller,” IEEE Transactions on
Neural Networks, 19, 1220–1230.
Chu, H., Qian, C., and Yang, J. (2015), “Almost disturbance decoupling for a class of
nonlinear systems via sampled-data output feedback control,” International Journal
of Robust and Nonlinear Control, 57, 2934–2939.
Emırler, M., Kahraman, K., Şentürk, M., Acar, O., Güvenç, B.A., Güvenç, L., and
Efendıoğlu, B. (2015), “Lateral stability control of fully electric vehicles,” Interna-
tional Journal of Automotive Technology, 16, 317–328.
Geng, C., Mostefai, L., Denai, M., and Hori, Y. (2009), “Direct Yaw-Moment Control
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

of an In-Wheel-Motored Electric Vehicle Based on Body Slip Angle Fuzzy Observer,”


IEEE Transactions on Industrial Electronics, 56, 1411–1419.
He, Z., Ou, Y., and Yuan, J. (2013), “Research on the Torque Dynamic Distribution
Algorithm of In-Wheel-Motor Electric Vehicle,” Lecture Notes in Electrical Engi-
neering, 198, 257–266.
Hori, Y. (2004), “Future vehicle driven by electricity and control-research on four-
wheel-motored” UOT Electric March II”,” IEEE Transactions on Industrial Elec-
tronics, 51, 954–962.
Kawashima, K., Uchida, T., and Hori, Y. (2010), “Rolling Stability Control Utilizing
Rollover Index for In-wheel Motor Electric Vehicle,” Ieej Transactions on Industry
Applications, 130, 655–662.
Ko, S., Ko, J., Lee, S., Cheon, J., and Kim, H. (2015), “A study on the road friction
coefficient estimation and motor torque control for an in-wheel electric vehicle,”
Proceedings of the Institution of Mechanical Engineers Part D Journal of Automobile
Engineering, 229, 611–623.
Li, F., Wang, J., and Liu, Z. (2009), “Motor torque based vehicle stability control
for four-wheel-drive electric vehicle,” in Vehicle Power and Propulsion Conference,
2009. VPPC’09. IEEE, IEEE, pp. 1596–1601.
Li, T., Sun, Z.Y., and Yang, S.H. (2017), “Output tracking control for generalised
high-order nonlinear system with serious uncertainties,” International Journal of
Control, 90, 322–333.
Mansouri, A., Smairi, N., and Trabelsi, H. (2016), “Multi-objective optimization of an
in-wheel electric vehicle motor,” International Journal of Applied Electromagnetics
and Mechanics, 50, 449–465.
Marino, R., and Tomei, P. (2000), “Adaptive output feedback tracking with almost
disturbance decoupling for a class of nonlinear systems ,” Automatica, 36, 1871–
1877.
Nam, K., Fujimoto, H., and Hori, Y. (2012), “Lateral stability control of in-wheel-
motor-driven electric vehicles based on sideslip angle estimation using lateral tire
force sensors,” IEEE Transactions on Vehicular Technology, 61, 1972–1985.
Nikam, S.P., Rallabandi, V., and Fernandes, B.G. (2012), “A High-Torque-Density
Permanent-Magnet Free Motor for in-Wheel Electric Vehicle Application,” IEEE
Transactions on Industry Applications, 48, 2287–2295.
Ozcetin, H., Saberi, A., and Sannuti, P. (1992), “Design for H almost disturbance de-

15
coupling problem with internal stability via state or measurement feedbacksingular
perturbation approach,” International Journal of Control, 55, 901–944.
Qian, C., and Du, H. (2012), “Global output feedback stabilization of a class of non-
linear systems via linear sampled-data control,” IEEE Transaction on Automatic
Control, 57, 2934–2939.
Qian, C., and Wei, L. (2003), “Almost disturbance decoupling for a class of high-order
nonlinear systems,” IEEE Transactions on Automatic Control, 45, 1208–1214.
Sakai, S.I., Sado, H., and Hori, Y. (2002), “Dynamic driving/braking force distribution
in electric vehicles with independently driven four wheels,” Electrical Engineering
in Japan, 138, 79–89.
Shuai, Z., Zhang, H., and Wang, J. (2014), “Design and evaluation on electric dif-
ferentials for over-actuated electric ground vehicles with four independent in-wheel
motors,” IEEE Transactions on Vehhicular Technology, pp. 55–66.
Sun, Z.Y., Li, T., and Yang, S.H. (2016), “A unified time-varying feedback approach
and its applications in adaptive stabilization of high-order uncertain nonlinear sys-
tems,” Automatica, 70, 249–257.
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

Wang, R., Chen, Y., Feng, D., Huang, X., and Wang, J. (2011), “Development and per-
formance characterization of an electric ground vehicle with independently actuated
in-wheel motors,” Journal of Power Sources, 196, 3962–3971.
Wang, R., Zhang, H., and Wang, J. (2015), “Robust lateral motion control of four-
wheel independently actuaed electric vehicles with tire force saturation considera-
tion,” Journal of the Franklin Institute, 352, 645–668.
Wu, F.K., Yeh, T.J., and Huang, C.F. (2013), “Motor control and torque coordination
of an electric vehicle actuated by two in-wheel motors,” Mechatronics, 23, 4660.
Xiong, L., Yu, Z., Wang, Y., Yang, C., and Meng, Y. (2012), “Vehicle dynamics control
of four in-wheel motor drive electric vehicle using gain scheduling based on tyre
cornering stiffness estimation,” Vehicle System Dynamics, 50, 1–16.
Yang, J., Chen, W., Li, S., and Chen, X. (2013), “Static disturbance-to-output decou-
pling for nonlinear systems with arbitrary disturbance relative degree,” International
Journal of Robust and Nonlinear Control, 23, 562–577.
Yang, J., Li, S., and Chen, W.H. (2012), “Nonlinear disturbance observer-based control
for multi-input multi-output nonlinear systems subject to mismatching condition,”
International Journal of Control, 85, 1071–1082.

16
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

y
F xf
F yr Mz V F yf
Fxr
x
b d

Lr Lf
a r g a f

Figure 1. The two wheels model of an EV driven by four in-wheel motors

17
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

0.5

0.4

0.3
d(t) (rad)

0.2

0.1

-0.1
0 10 20 30
Time (0.1s)

Figure 2. The disturbance used in the simulation

18
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

120km/h
10
80km/h

-5
0 5 10 15 20 25 30
Time (0.1s)

Figure 3. The sampled-data controller’s inputs under different vehicle speed

19
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

0.08
120km/h
0.06 80km/h

0.04

0.02

-0.02

-0.04
0 10 20 30
Time (0.1s)

Figure 4. The response of lateral motion states eβ of EV under different vehicle speed

20
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

1
120km/h
80km/h

0.5

-0.5
0 10 20 30
Time (0.1s)

Figure 5. The response of lateral motion states eγ of EV under different vehicle speed

21
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

44
The Front-left Tire
The Front-right Tire
force allocation (N)

42 The Rear-left Tire


The Rear -right Tire

40

38

36
0 5 10 15 20 25 30
Time (0.1s)

Figure 6. The force allocation of every tire under 120km/h

22
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

44
The Front-left Tire
The Front-right Tire
force allocation (N)

42 The Rear-left Tire


The Rear -right Tire

40

38

36
0 5 10 15 20 25 30
Time (0.1s)

Figure 7. the force allocation of every tire under 80km/h

23
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

Figure 8. The tested EV designed by ourselves

24
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

10

-5
0 10 20 30
Time (0.1s)

Figure 9. The sampled-data controller’s input under 50km/h

25
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

0.1
Simulation
Test

0.05

0 10 20 30
Time (0.1s)

Figure 10. the simulation and test results of lateral motion states eβ of EV under 50km/h

26
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

1
Simulation
Test

0.5

-0.5
0 10 20 30
Time (0.1s)

Figure 11. the simulation and test results of lateral motion states eγ of EV under 50km/h

27
Downloaded by [Australian Catholic University] at 19:38 17 August 2017

Table 1. The EV parameters in the simulation

Parameter Nominal Value Parameter Nominal Value

m 700 kg Iz 800 kg ∙ m2

Lf 0.8 m Lr 0.8 m

Cf 13000N/rad Cr 14000 N/rad

28

You might also like