Vehicle System Dynamics

International Journal of Vehicle Mechanics and Mobility

ISSN: 0042-3114 (Print) 1744-5159 (Online) Journal homepage: https://www.tandfonline.com/loi/nvsd20

Vehicle dynamic estimation with road bank angle

consideration for rollover detection: theoretical
and experimental studies

H. Dahmani, M. Chadli, A. Rabhi & A. El Hajjaji

To cite this article: H. Dahmani, M. Chadli, A. Rabhi & A. El Hajjaji (2013) Vehicle
dynamic estimation with road bank angle consideration for rollover detection:
theoretical and experimental studies, Vehicle System Dynamics, 51:12, 1853-1871, DOI:
10.1080/00423114.2013.839819

To link to this article: https://doi.org/10.1080/00423114.2013.839819

Published online: 30 Sep 2013.

Submit your article to this journal

Article views: 809

View related articles

Citing articles: 6 View citing articles

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

https://www.tandfonline.com/action/journalInformation?journalCode=nvsd20
Vehicle System Dynamics, 2013
Vol. 51, No. 12, 1853–1871, http://dx.doi.org/10.1080/00423114.2013.839819

Vehicle dynamic estimation with road bank angle

consideration for rollover detection: theoretical and
experimental studies
H. Dahmani∗ , M. Chadli, A. Rabhi and A. El Hajjaji
Information and Systems Laboratory (EA 4290), University of Picardie Jules Verne, Modelling,
MIS, 7 Rue du Moulin Neuf 80000 Amiens, France

(Received 15 November 2012; accepted 11 August 2013 )

This article describes a method of vehicle dynamics estimation for impending rollover detection.
This method is evaluated via a professional vehicle dynamics software and then through experimental
results using a real test vehicle equipped with an inertial measurement unit. The vehicle dynamic states
are estimated in the presence of the road bank angle (as a disturbance in the vehicle model) using a
robust observer. The estimated roll angle and roll rate are used to compute the rollover index which
is based on the prediction of the lateral load transfer. In order to anticipate the rollover detection, a
new method is proposed in order to compute the time-to-rollover using the load transfer ratio. The
used nonlinear model is deduced from the vehicle lateral dynamics and is represented by a Takagi–
Sugeno (TS) fuzzy model. This representation is used in order to take into account the nonlinearities of
lateral cornering forces. The proposed TS observer is designed with unmeasurable premise variables
in order to consider the non-availability of the slip angles measurement. Simulation results show that
the proposed observer and rollover detection method exhibit good efficiency.

Keywords: vehicle rollover; vehicle dynamics; TS fuzzy model; observer; robustness; LMI

1. Introduction

In recent years most modern vehicles have been equipped with driver assistance systems in
order to improve driver and passenger security. For example the anti-lock braking systems to
improve braking performance and electronic stability programmes to stabilise yaw motion.
Safety and driver assistance systems greatly reduce potential injury risk during vehicle acci-
dents. Today, airbags are integrated in most vehicles. This has resulted in a decreasing number
of serious injuries during vehicle crashes. In the past few years, research has introduced new
systems such as side and roof airbags which cover even more accident situations than merely
frontal crashes. However, vehicle rollover still remains without adequate occupant protection.
While rollover occurs only in 3% of all passenger car accidents, they contribute to 33% of
the number of fatal accidents in the USA.[1] These figures show the potential danger to the
passengers during rollover. The reduction of vehicle rollover occurrence is an important part
in providing increased passenger safety.

∗ Corresponding author. Email: hamdi.dahmani@u-picardie.fr

© 2013 Taylor & Francis

1854 H. Dahmani et al.

Rollover can occur during typical driving situations and poses a real threat for vehicles.
Examples are excessive speed during cornering, obstacle avoidance and severe lane change
manoeuvres. In such cases, rollover occurs as a direct result of the lateral wheel forces induced
during these manoeuvres.[2–4] In early studies on the detection of vehicle rollover, the concept
of a static rollover threshold was used, but this is only useful at steady state. In [5–7], time-to-
rollover (TTR) is proposed to estimate the time until rollover occurs and a direct yaw moment
control using differential braking is performed. Hac et al. described a rollover index (RI)
using a model-based roll estimator in [8]. In [9], an RI combining the lateral dynamics model-
based estimator and vertical dynamics model-based estimator is proposed. Traditionally, some
estimation of the vehicle load transfer ratio (LTR) has been used as a basis for the design of
rollover prevention systems.[10] The LTR quantity can be simply defined as the difference
between the normal forces on the right- and left-hand sides of the vehicle divided by their
sum. Clearly, LTR varies within [−1, 1], and for a perfectly symmetric car that is driving
straight, it is zero. In this work we use the dynamic load transfer ratio LTRd as RI which
is computed using the estimated roll angle and roll rate. In order to anticipate the detection,
the LTRd rate is used to compute the TTR. A model-based roll state estimator is designed
in the presence of the road bank angle using the H∞ approach and taking into account the
unmeasurable premise variables.[11,12] The nonlinear model, as derived from three-degree-
of-freedom vehicle lateral dynamics, is represented by a Takagi–Sugeno (TS) fuzzy model [13]
which is very efficient in order to take into account the lateral force nonlinearities.[12,14,15]
This representation has been widely used and studied these last years (see, for example,
[15–20]).
The structure of the paper is as follows. Section 2 introduces the used model represented
by a TS fuzzy model, the parameter identification and the model validation using the CarSim
software. Section 3 presents a rollover detection study, the LTRd and the TTR are discussed
in the presence of the road bank angle. A TS model-based roll state estimator using the H∞
approach is then designed in Section 4. Section 5 contains simulation results and shows the
RI computed in different driving scenarios using the CarSim software package and then an
experimental vehicle.

2. Vehicle modelling and parameter identification

The model used in this work describes vehicle lateral and roll dynamics, which is obtained
by considering the well-known single-track (bicycle) model with a roll degree of freedom
(Figure 1). The three-dimensional model with road bank angle and nonlinear tyre character-
istics of the four wheeled vehicle behaviour can be described by the following differential
equations[21–23]:

m(vβ̇ + vψ̇ − φ¨v h) = 2Fyf + 2Fyr − mgφr ,

Iz ψ̈ = 2Fyf lf − 2Fyr lr , (1)
Ix φ¨v = ms gh(φv + φr ) + ms ay h − Kφ φv − Cφ φ˙v ,

with β denoting the sideslip angle, ψ, φv and φr the vehicle yaw, the roll angle and the road
bank angle, respectively, Ff the cornering force of the two front tyres and Fr the cornering
force of the two rear tyres. For further description of the parameters appearing in the vehicle
dynamics model, refer to Table 1. The Pacejka tyre model [24] gives the cornering forces Fyf
 Vehicle System Dynamics 1855

Figure 1. Vehicle parameters description.

and Fyr as a function of tyre slip angles by the following nonlinear expressions:

Fyf = Df sin[Cf tan−1 Bf (1 − Ef )αf + Ef tan−1 (Bf αf )],

(2)
Fyr = Dr sin[Cr tan−1 Br (1 − Er )αr + Er tan−1 (Br αr )],

with

lf ψ̇
αf = δf − − β,
v
(3)
lr ψ̇
αr = − β,
v

with δ being the front steering angle, αf the slip angle of the front tyres and αr the slip
angle of the rear tyres (Figure 1). Coefficients Bi , Ci , Di and Ei (i = f, r) depend on the tyre
characteristics, road adhesion coefficient and the vehicle operational conditions. The above
Pacejka model describes such phenomena, but is hardly usable since it depends on many
nonlinearities and varying parameters that need to be known.
Lateral forces are assumed in other studies to be proportional to the slip angle, i.e.

Fyf = Cf αf ,
(4)
Fyr = Cr αr .

It is obvious that when the slip angles are very small, the obtained linear model works
very well (Figure 2). However, in case of important slip angles, the nonlinear model must be
considered.[25]
1856 H. Dahmani et al.

Table 1. Vehicle model parameter deffinitions.

Parameter Description Unit

β Sideslip angle at CG (rad)

ψ̇ Yaw rate (rad/s)
φv Roll angle (rad)
φ˙v Roll rate (rad/s)
δf Front steering angle (rad)
ms Sprung vehicle mass (kg)
m Vehicle mass (kg)
v Vehicle speed (m/s)
Ix Roll moment of inertia at CG (kgm2 )
Iz Yaw moment of inertia at CG (kgm2 )
lr Distance from CG to rear axle (m)
lf Distance from CG to front axle (m)
T Vehicle track width (m)
h CG height from roll axis (m)
Cφ Combined roll damping coefficient (Nm s/rad)
kφ Combined roll stiffness coefficient (Nm/rad)

5000

4000

3000
Fuzzy model
2000 Nonlinear model
Linear model
Lateral force (N)

1000

−1000

−2000

−3000

−4000

−5000
−0.5 0 0.5
Slip angle (rad)

Figure 2. Comparison of the tyre models.

2.1. TS fuzzy representation of the vehicle model

In this work we take into account the nonlinearities of the cornering forces by considering a
TS fuzzy representation of the tyre model described by the following rules:


Fyf = Cf1 αf ,
If |αf |is M1 then
Fyr = Cr1 αr ,
 (5)
Fyf = Cf2 αf ,
If |αf | is M2 then
Fyr = Cr2 αr ,

where Cfi , Cri are the front and rear tyre cornering stiffness which depend on the road friction
coefficient and the vehicle parameters. M1 (M2 ) is a fuzzy set for small (large) slip angles.
 Vehicle System Dynamics 1857

Remark 1 Since αf and αr have similar values (the same fuzzy sets), the proposed rules are
made only for αf . This assumption allows reducing the number used membership functions
and therefore the number of parameters to identify.

The overall forces are obtained by

Fyf = μ1 (|αf |)Cf1 αf + μ2 (|αf |)Cf2 αf ,
(6)
Fyr = μ1 (|αf |)Cr1 αr + μ2 (|αf |)Cr2 αr ,
with μj (j = 1, 2) being the jth bell curve membership function of fuzzy set Mj . They satisfy
the following properties:

2
μi (|αf |) = 1,
i=1 (7)
0 ≤ μi (|αf |) ≤ 1, i = 1, 2.
The expressions of the membership functions used are:
ωi (|αf |)
μi (|αf |) = 2 , i = 1, 2, (8)
i=1 ωi (|αf |)

with
1
ωi (|αf |) = . (9)
(1 + |((|αf | − ci )/ai )|)2bi
Using the above approximations of nonlinear lateral forces by TS rules, nonlinear model (1)
can be represented by the following TS fuzzy model:

2

ẋ(t) = μi (|αf |) Ai x(t) + Bi δf (t)) + Bw φr (t),
i=1 (10)
y(t) = Cx(t),
with
⎡ σI ρi Ixeq ⎤ ⎡ Cfi Ix ⎤
i xeq hCφ h(ms gh−kφ )
− − 1 − 2 mIx veq
⎢ mIx v mIx v2 Ix v Ix v ⎥ ⎢ ⎥
⎢ ρi ⎥
Ai = ⎢ − Iτziv 0 0 ⎥ , Bi = ⎢
⎢
2 CIfizlf ⎥
⎥,
⎢ m hσ
I z ⎥ ⎣2 mmI
s hCfi ⎦
⎣− s i ms hρi
−
Cφ (ms gh−kφ ) ⎦
x
mIx mIx v Ix Ix
0 0 1 0 0
g
Bw = [− 0 0 0]T , C = [0 1 0 0],
v
with x(t) = [β(t) ψ̇(t) φ˙v (t) φv (t)]T is the state vector of the model, Ixeq denotes the
equivalent roll moment of inertia of the vehicle about the roll axis, which is given by
Ixeq = Ix + ms h2 (11)
and σi , ρi , τi are auxiliary variables introduced in order to simplify the model description; they
are defined as follows:
σi = 2(Cri + Cfi ),
ρi = 2(lr Cri − lf Cfi ), (12)
τi = 2(lf2 Cfi + lr2 Cri ).
1858 H. Dahmani et al.

Table 2. Simulation vehicle parameters.

Parameter Value Unit

ms 1592 (kg)
m 1832 (kg)
v 20 (m/s)
Ix 614 (kg m2 )
Iz 2488 (kg m2 )
lr 1.77 (m)
lf 1.18 (m)
T 1.5 (m)
h 0.559 (m)
Cφ 6000 (Nm s/rad)
kφ 48,000 (Nm/rad)

Table 3. Parameters of the membership functions.

a1 a2 b1 b2 c1 c2 Cf1 Cf2 Cr1 Cr2

0.0852 3.8722 0.6741 22.8174 0.0218 3.8529 96240 829.15 107180 650.44

Steering wheel angle (deg)

25

20

15
Steering wheel angle (deg)

10

−5

−10

−15

−20

−25
0 1 2 3 4 5 6 7
Time (s)

Figure 3. Steering wheel angle used in the fishhook test.

2.2. Parameter identification of the vehicle model

Using an identification method based on the Levenbenrg–Marquadt algorithm [14] combined

with the least-square method, parameters of the membership functions (ai , bi and ci ) and
stiffness coefficient values are obtained. For the vehicle parameters defined in Table 2 and a
dry road, the obtained values are given in Table 3.
The quality of the parameter identification results is examined in comparison with the
simulation results obtained from the professional vehicle dynamics software CarSim.[26]
One of the comparison results is shown in Figure 4, where the simulation is carried out
under a fishhook manoeuvre (Figure 3) at a speed of 50 km/h. The simulation results show
that the identified TS model shows a good representation of the actual states measured using
CarSim.
 Vehicle System Dynamics 1859

(a)Sideslip angle (rad) (c)Roll rate (rad)

0.01 0.1
Sideslip angle (rad)

Vehicle model

Roll rate (rad)

0.005 0.05
CarSim
0 0

−0.005 −0.05

−0.01 −0.1
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Time (s) Time (s)
(b)Yaw angle (rad) (d)Roll angle (rad)
0.1 0.04
Yaw angle (rad)

Roll angle (rad)

0.05 0.02

0 0

−0.05 −0.02

−0.1 −0.04
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Time (s) Time (s)

Figure 4. Comparison of the TS model with CarSim at 50 km/h.

3. Rollover detection study

3.1. Lateral load transfer

Lateral load transfer is the change in the normal force acting on the tyres due to both the
acceleration of the centre of gravity (CG), and the shifting of the position of the CG in the y
direction due to the movement of the suspension. Figure 5 illustrates lateral load transfer in
the vertical plane.

3.2. The LTR and LTRd in the presence of the road bank angle

The LTR can be defined as the difference between the normal forces on the right tyres and the
left tyres of the vehicle divided by their sum. Assuming no vertical motion exists, the LTR is
given by
Fzl − Fzr Fzl − Fzr
LTR = = , (13)
Fzl + Fzr mg
with Fzl and Fzr are, respectively, the vertical left and right tyre forces. It is apparent that LTR
varies within [−1, 1], and for a perfectly symmetric car which is driving straight, it is zero.
The extrema are reached in the case of a wheel lift-off of one side of the vehicle, in which
case LTR becomes 1 or −1 depending on the side which lifts off.
The estimation of the LTR is very difficult since normal force sensors are expensive. An
expression for LTR which depends on the roll states and vehicle parameters can be obtained.
This is denoted by LTRd . In order to derive LTRd we resolve weight (ms g) and pseudo-force
(ms ay ) into components in the vehicle-fixed y and z directions; the following dynamics are
obtained:

ms ay h + ms gh(φv + φr ) − Cφ φ˙v + Kφ φv = 0. (14)

We can also write a torque balance for the sprung and unsprung masses about the left tyre roll
axis as follows:

T T
Fzr T + ms ay h + ms ghφr − ms g − hφv − mu g = 0. (15)
2 2
1860 H. Dahmani et al.

Figure 5. Vehicle roll model.

By substituting Equations (14) and (15) into Equation (13), we obtain the following expression
for LTRd :
2
LTRd = (Cφ φ˙v + Kφ φv ). (16)
mgT

Remark 2 In the above section it is shown that the presence of the road bank angle affects
the LTRd through the roll angle and roll rate.

3.3. TTR computation

Even though the LTRd can be used as a RI to accurately detect the tyre lift off, it can be pointed
out that the LTRd is not able to predict or detect an impending rollover with anticipation. The
TTR is one of the most efficient indicators in order to anticipate the rollover detection. It is
defined as the time remaining before wheel lift off will occur, which gives a clear indication
of the beginning of rollover. A TTR computation is proposed in [27,28] by assuming that the
input steering angle stays fixed at its current position in the foreseeable future; it is defined
as the time it takes for the vehicle sprung mass to reach its critical roll angle. In order to take
account of the steering angle rate, two advanced versions of the TTR are developed, based on
the fifth-order linear vehicle dynamic model in [29].
In this study the TTR is computed as follows: assuming that the LTRd increases or decreases
at its current rate in the near future, compute the time taken by the LTR to reach 1 or −1 using
the following equations:
1 − LTRd
TTR = if LTRd > 0, (17)
RLTR
−1 − LTRd
TTR = if LTRd < 0, (18)
RLTR
with RLTR being the LTRd rate, which is obtained from a filtered differential signal of the
LTRd .
 Vehicle System Dynamics 1861

Under normal driving conditions there is no risk of rollover and the LTRd is close to zero. In
this case the TTR is usually very large and increases excessively. For example, if the vehicle
is driven straight, there is no roll motion at all. Therefore, the TTR approaches infinity. For
implementation considerations, TTR is saturated at 10 s when the LTRd is too small.

4. Model-based roll state estimator with road bank angle consideration

As shown in the above section, in order to compute the LTRd it is necessary to know the roll
angle and roll rate which are difficult and very expensive to measure directly. Roll angle and roll
rate can be estimated from measurable signals such as the yaw rate and the vehicle parameters.
In this study both the roll angle and the roll rate are estimated in the presence of the road bank
angle as bounded unknown input disturbances in the used TS vehicle model. Vehicle dynamic
variables such as the yaw rate, steering angle and vehicle velocity are measured (Figure 6).

4.1. TS estimator design conditions

Using Equation (10), a TS model-based estimator for the estimation of the roll angle and the
roll rate in the presence of the unknown road bank angle is represented as follows:


2

˙ =
x̂(t) μi (|α̂f |) Ai x̂(t) + Bi δ(t) + Li (y(t) − ŷ(t)),
i=1 (19)
ŷ(t) = C x̂(t).

By using measurable signals such as the tyre steering angle and yaw rate and consider-
ing unmeasurable premise variables, the roll angle and the roll rate can be estimated with
Equation (19). The aim of the design is to determine gain matrices Li , which guarantee the
asymptotic convergence of x̂(t) towards x(t). Let us define the state estimation error as follows:

e(t) = x(t) − x̂(t). (20)

The dynamic of the state estimation error is governed by


2 
2
ė(t) = μi (|α̂f |)μj (|αf |)((Ai − Li C)e(t) + Aij x(t) + Bij δ(t)) + Bw φr (t), (21)
i=1 j=1

Figure 6. Vehicle state estimator.

1862 H. Dahmani et al.

with
Aij = Aj − Ai , Bij = Bj − Bi . (22)
Let us define
   
e(t) δ(t)
xe (t) = , w= . (23)
x(t) φr (t)
The augmented system formed from system (10) and state estimation error (21) can now be
expressed as

2 
2
ẋe (t) = μi (|α̂f |)μj (|αf |)(Āij xe (t) + B̄ij w(t)), (24)
i=1 j=1

with
   
A − Li C Aij Bij Bw
Āij = i , B̄ij = . (25)
0 Aj Bj Bw

Remark 3 Since w(t) is constituted of the steering angle and the road bank angle, it can be
logically assumed to have a finite energy.

The TS estimator gains have been computed by considering the effect of the road bank angle
on the state estimation errors. One possible method is to minimize the L2 gain (H∞ norm)
from disturbances to the estimation errors.
The L2 gain between vector w(t) and estimation error e(t) are defined by the following
quantity:
e(t)
γ = sup . (26)
w(t)
By the definition of the supremum and the L2 gain, Equation (26) can be expressed as
 ∞  ∞
e(t) e(t) dt ≤ γ
T 2
w(t)T w(t) dt. (27)
0 0

Theorem 1 If there exists positive and symmetric matrices P1 and P2 , matrices Mj and
positive scalar γ satisfying the following LMI for i, j = 1, 2, :
⎡ ⎤
i P1 Aij P1 Bij P1 Bw
⎢ ATij P1 j P2 B j P 2 Bw ⎥
⎢ ⎥
⎢ ⎥ < 0, (28)
⎣ BijT P1 BjT P2 −γ 2 I 0 ⎦
BwT P1 BwT P2 0 −γ 2 I

with

i = ATi P1 + P1 Ai − Mi C − C T MiT + I, (29)

j = ATj P2 + P2 Aj (30)

then the estimation error (20) converges asymptotically towards zero and satisfies the H∞
performance (27). The observer gains are given by Li = P1−1 Mi .
 Vehicle System Dynamics 1863

Proof Let us consider the following Lyapunov function candidate:

V (xe ) = xe (t)T Pxe (t) (31)

with P = PT > 0. The state estimation error can be written as follows:

e(t) = Ce xe (t) (32)

with
Ce = [I 0]. (33)
System (24) is stable and the H∞ gain of the transfer from unknown input vector w(t) to the
state estimation errors e(t) is bounded by γ > 0 if the following condition holds:

J∞ = V̇ (xe ) + e(t)T e(t) − γ 2 w(t)T w(t) < 0. (34)

Substituting Equation (24) into Equation (34) gives


2 
2
μi (|α̂f |)μj (|αf |)(xe (t)T ĀTij Pxe (t) + xe (t)T PĀij xe (t) + w(t)T B̄ijT Pxe (t)
i=1 j=1

+ xe (t)T PB̄ij w(t)) + xe (t)T xe (t) − γ 2 w(t)T w(t) < 0. (35)

Inequality (35) can be expressed as an equivalent inequality as follows:


2 
2  T   
x Xij PB̄ij xe
μi (|α̂f |)μj (|αf |) e <0 (36)
i=1 j=1
w B̄ijT P −γ 2 I w

with
Xij = ĀTij P + PĀij + CeT Ce . (37)
According to the convex sum property of the activation functions, inequality (36) holds if the
following conditions are satisfied:
 
ĀTij P + PĀij + CeT Ce PB̄ij
< 0, ∀ i, j = 1, 2. (38)
B̄ijT P −γ 2 I

These constraints are nonlinear. In order to get LMI conditions, let us consider the following
particular form of matrix P :
 
P 0
P= 1 . (39)
0 P2
By substituting Equations (25) and (39), inequality (38) can be written as
⎡ ⎤
i P1 Aij P1 Bij P 1 Bw
⎢ A T P1 j P2 B j P 2 Bw ⎥
⎢ ij ⎥
⎢ T ⎥<0 (40)
⎣ Bij P1 BjT P2 −γ 2 I 0 ⎦
BwT P1 BwT P2 0 −γ 2 I
1864 H. Dahmani et al.

with

i = (Ai − Li C)T P1 + P1 (Ai − Li C) + I, (41)

j = ATj P2 + P2 Aj . (42)

Using variable change Mi = P1 Li , condition (40) is linear in variables P1 , P2 and Mi ,

which leads to the equivalent condition given by Equation (28). Indeed, it suffices to satisfy
Equation (28) to guarantee V̇ (t) < 0 with the γ -attenuation (27). 

4.1.1. LMI relaxation

To obtain less-conservative LMI conditions of Theorem 1, we use the technique developed in

[16], the following results are proposed.

Corollary 1 If there exists matrices P1 > 0 and P2 > 0, matrices Qij , Mj and scalar γ such
that the following LMIs hold

ii + Qii < 0, i = 1, 2, (43)

ij + ji + Qij + Qji < 0, i < j, (44)
 
Q11 Q12
T
> 0, (45)
Q12 Q22

with
⎡ ⎤
j P1 Aij P1 Bij P1 Bw
⎢ T
Aij P1 i P2 B i P1 Bw ⎥
⎢ ⎥
ij = ⎢ ⎥<0 (46)
⎣ BijT P1 BiT P2 −γ 2 I 0 ⎦
BwT P1 BwT P2 0 −γ 2 I

then estimation error (20) is stable and satisfies the H∞ performance (27). The observer gains
are given by Li = P1−1 Mi .

5. Simulation results

The above developed approach has been tested under different driving scenarios using CarSim
software and then an experimental vehicle equipped with the RT2500 inertial measurement
unit.

5.1. CarSim software validation

The developed model-based estimator has been implemented in a professional simulator in

order to be tested under different driving tests. Two fishhook tests are conducted with different
steering wheel angles (Figure 7). The input steering angle used in test 2 is defined such that
the wheel lift-off occurs at 2.8 s, whereas in test 1 no wheel lift-off occurs. In this simulation,
the vehicle is driven at a constant speed of 110 km/h in a 6% banked road. Using the model
 Vehicle System Dynamics 1865

0.08
test 1
0.06 test 2

0.04

Steering angle (rad)

0.02

−0.02

−0.04

−0.06

−0.08
0 1 2 3 4 5 6 7 8
Time (s)

Figure 7. The used steering angles in test 1 and test 2.

parameters given in Table 2, the resolution of LMI constraints (28) the LMI toolbox of Matlab
leads to the following Pi and Mi matrices:
⎡ ⎤
0.0003 0.0145 0.0022 0.0087
⎢0.0145 1.6132 0.2351 0.4366⎥
P1 = 104 ⎢⎣0.0022 0.2351 0.0343 0.0632⎦ ,
⎥

0.0087 0.4366 0.0632 1.1349

⎡ ⎤
0.4271 0.0101 −0.001 0.0788
⎢ 0.0101 0.0961 0.0033 0.0410⎥
P2 = 103 ⎢⎣−0.001 0.0033 0.0025 0.0059⎦ ,
⎥

0.0788 0.0410 0.0059 0.1457

M1 = 104 [−0.099 − 5.277 − 0.782 − 8.482]T ,
M2 = 105 [0.007 1.096 0.160 − 0.417]T .

Then, the following observer gains are obtained:

L1 = [222.130 131.155 − 920.682 − 8.385]T ,

L2 = [192.989 75.474 − 468.130 − 8.131]T .

The vehicle states estimated using the designed observer are compared with the actual states
measured in CarSim for the conducted tests (Figures 8 and 9). In both tests, the designed
estimator shows good performance. However, between 3 and 3.6 s (the moment of the wheel
lift-off), the estimation is not quite that good. This is due to the vehicle model which does not
take into account the vehicle behaviour after the rollover. Figure 10 shows the simulation results
of the LTRd computation for the two tests. The TTR shows good efficiency for the rollover
detection, but the proposed rollover indicator, which is the TTR, shows bigger anticipation in
the rollover detection (Figure 10). This advantage is very interesting since the rollover has to
be avoided in a matter of seconds.

5.2. Experimental tests

This part is devoted to the definition of the experimental test used to validate the developed
aproaches in vehicle dynamic estimation and rollover detection. The used test vehicle has been
1866 H. Dahmani et al.

(a) Sideslip angle (rad) (c) Roll rate (rad/s)

0.1 0.5
Sideslip angle (rad)

CarSim

Roll rate (rad/s)

0.05 Estimate

0 0

−0.05

−0.1 −0.5
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Time (s) Time (s)
(b) Yaw rate (rad/s) (d) Roll angle (rad)
0.5 0.1
Yaw rate (rad/s)

Roll angle (rad)

0.05

0 0

−0.05

−0.5 −0.1
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Time (s) Time (s)

Figure 8. Simulation results of the vehicle state estimator in test 1.

(a)Sideslip angle (rad) (c)Roll rate (rad/s)

0.1 1
Sideslip angle (rad)

Roll rate (rad/s)

0.05 CarSim 0.5

Estimate
0 0

−0.05 −0.5

−0.1 −1
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
time (sec) time (sec)

(b)Yaw rate (rad/s) (d)Roll angle (rad)

0.5 0.2
Yaw rate (rad/s)

Roll angle (rad)

0.1
0
0
−0.5
−0.1

−1 −0.2
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
time (sec)
time (sec)

Figure 9. Simulation results of the vehicle state estimator in test 2.

equipped with the inertial measurement unit RT2500 which includes precision Inertial and
GPS Navigation systems to measure the motion, position and orientation of the vehicle. Thus,
the RT2500 allows the measurement of all vehicle dynamics without simplifying assumptions.
It works on any type of vehicle and in any orientation provided that it has been configured
properly. The RT2500 must be firmly attached to the chassis of the vehicle in order to prevent
motion from distorting the measurements. A metal rod was installed between the rear seats
of the vehicle on which the central support is fixed. It can also simply be fixed on a plane
surface fixed to the vehicle chassis (Figure 11). The acquisition of the data is done by a laptop
connected to the RT2500 via an ethernet cable. A GPS antenna is also fixed on the roof of the
vehicle and connected to the RT2500.
In order to test the vehicle dynamics estimator and the rollover risk indicators in real
situations, experimental tests have been conducted. Using our vehicle (Renault Laguna II)
equipped with the inertial measurement unit RT2500, two tests have been conducted on the
track of our partner (School J.B. Delambre, Amiens). This track is 100 m long and 8 m wide,
ideal to simulate tests in the slalom and obstacle avoidance tests. The steering wheel angle is
measured from the Laguna CAN bus. On the other hand, using the RT2500 unit we obtained
the following variable measurements:
 Vehicle System Dynamics 1867

1 10
test1
test2 9 test1
test2
0.5 8

0 6

TTR (s)
d
LTR

−0.5 4

−1 2

−1.5 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Temps (s) Temps (s)

Figure 10. Comparison of the LTRd and TTR in test 1 and test 2.

Figure 11. The test vehicle on the track of ‘School Delambre’.

• Longitudinal vehicle speed v.

• Lateral vehicle speedvy .
• Yaw rate ψ̇.
• Sideslip angle β.
• Lateral acceleration ay .
• Vehicle roll angle φ.
• Vehicle roll rate φ̇.

Longitudinal vehicle speed, vehicle lateral acceleration and vehicle yaw rate are also
available as measurements from the Laguna II computers.
In this part we present the results of two slalom tests which were made to validate the TS
observer and the rollover risk indicators.
Two slalom tests were conducted to simulate close situations of the rollover. The first
scenario involves several slalom races over a period of 9.5 s, the second scenario is of a shorter
duration (3.5 s) with major steering angles. Both test 1 and test 2 are conducted with small
variations of the vehicle speed as shown in Figure 12. Figures 13 and 14 show the steering
1868 H. Dahmani et al.

50 50

45 45

40 40

35 35

Vehicle speed (km/h)

30 30

25 25

20 20

15 15

10 10

5 5

0 0
0 2 4 6 8 10 0 0.5 1 1.5 2 2.5 3 3.5
Time (s)

Figure 12. Vehicle speeds evolution in test 1 and test 2.

0.15 49.9135

0.1
49.9135
Steering wheel angle (rad)

0.05
49.9135
Latitude (°)

0
49.9135
−0.05

49.9135
−0.1

49.9135
−0.15

−0.2 49.9135
0 2 4 6 8 10 2.3208 2.321 2.3212 2.3214 2.3216 2.3218 2.322 2.3222
Time (s) Longitude (°)

Figure 13. Steering wheel angle and vehicle trajectory in test 1.

0.15 49.9135

0.1 49.9135
Steering wheel angle (rad)

0.05 49.9135

0 49.9135
Latitude (°)

−0.05 49.9135

−0.1 49.9135

−0.15 49.9135

−0.2 49.9135

−0.25 49.9135
0 0.5 1 1.5 2 2.5 3 3.5 2.3213 2.3213 2.3214 2.3215 2.3215 2.3215 2.3216 2.3217 2.3217
Time (s) Longitude (°)

Figure 14. Steering wheel angle and vehicle trajectory in test 2.

wheel angle and vehicle trajectory, respectively, in test 1 and test 2. The estimation of the
vehicle dynamics by the proposed observer compared with measurements from the RT2500 is
illustrated in Figure 15. These results show a satisfactory accuracy of the designed observer.
Estimation errors are larger for the sideslip angle and the roll angle, this can be explained by the
low values of these variables (less than 0.06 rad) compared with the yaw rate for example which
reaches 0.6 rad/s. Estimation errors and disturbances are thus more felt on the lower variables.
 Vehicle System Dynamics 1869

0.08 0.8
Measured Measured
Estimated Estimated
0.06 0.6

0.04 0.4
Sideslip angle (rad)

Yaw rate (rad/s)

0.02 0.2

0 0

−0.02 −0.2

−0.04 −0.4

−0.06 −0.6

−0.08 −0.8
0 2 4 6 8 10 0 2 4 6 8 10
Time (s) Time (s)
0.08 0.5
Measured Measured
Estimated Estimated
0.06 0.4

0.3
0.04 Roll rate (rad/s)
Roll angle (rad)

0.2
0.02
0.1
0
0
−0.02
−0.1

−0.04 −0.2

−0.06 −0.3
0 2 4 6 8 10 0 2 4 6 8 10
Time (s) Time (s)

Figure 15. Vehicle dynamics estimation in test 1.

0.8 1.2

0.6 1

0.4 0.8
LTRd

LTRd

0.2 0.6

0 0.4

−0.2 0.2

−0.4 0

−0.6 −0.2
0 2 4 6 8 10 0 0.5 1 1.5 2 2.5 3 3.5
Time (s) Time (s)

Figure 16. The LTR estimation for the two tests.

In order to estimate the rollover indicator, the lateral transfer ratio (LTRd ) is calculated for
the two scenarios. This value is shown in Figure 16, in the two scenarios the LTRd increases at
each turn and therefore at each situation closer to the vehicle rollover. In the second scenario,
at time 0.8 s, the LTRd slightly exceeds the threshold value of 1 which corresponds to the
beginning of a rollover. Figure 17 shows the TTR calculated for the two scenarios. Compared
with the LTRd , this index enables to predict rollover situations with more anticipation. As soon
as the TTR value reaches zero (for example, at time 0.75 s) the rollover should be considered
as imminent. This situation is detected by the LTRd at moment 0.8 s.
1870 H. Dahmani et al.

14 7

12 6

10 5
TTR (s)

TTR (s)
8 4

6 3

4 2

2 1

0 0
0 2 4 6 8 10 0 0.5 1 1.5 2 2.5 3 3.5
Time (s) Time (s)

Figure 17. TTR in the two tests.

These results show the effectiveness of the developed techniques for the vehicle dynamic
estimation and the rollover detection. However, improvements should be made to avoid bad
estimations in some situations. For instance, the longitudinal dynamics can be considered in
order to take into account the longitudinal velocity variations. The changes of the model after
the tyre lift off must also be considered in order to make the estimation in such situations more
accurate.

6. Conclusions and future works

A model-based roll state estimator is designed in the presence of the road bank angle using
the H∞ fuzzy observer approach. The nonlinear three-degrees-of freedom vehicle model is
represented by a TS fuzzy model. This representation takes into account the nonlinearities
introduced by the lateral forces. The designed fuzzy observer shows good performance even
in the presence of unknown road bank angle. Design conditions are formulated in LMI terms
that are easy to solve using convex optimisation tools. A dynamic approximation of the lateral
transfer ratio is used to compute the TTR which shows good accuracy and good anticipation
in the detection of of impending rollover. First, the proposed fuzzy observer and the RI have
been tested through the CarSim software in two different tests. An experimental test is then
conducted under real-world conditions using a Laguna car equipped with a RT2500 inertial
measurement unit. In further works, we will extend the results by considering more complex
vehicle models in order to take into account the vehicle behaviour after the wheel lift-off and
vehicle speed variations.

Funding

This work was supported by the ‘Conseil Régional de Picardie’ and the European Regional
Development Fund within the framework of the project ‘SEDVAC’.

Supplemental data

Supplemental data for this article can be accessed here http://dx.doi.org/10.1080/

09593330.2013.839819
 Vehicle System Dynamics 1871

References

[1] N.H.T.S.A Technical Report. Motor vehicle traffic crash injury and fatality estimates. NCSA (National Center
for Statistics and Analysis) Advanced Research and Analysis; 2003.
[2] Solmaz S, Corless M, Shorten R. A methodology for the design of robust rollover prevention controllers for
automotive vehicles: part 2-active steering. Proceedings of the American Control Conference; 2007 Jul 11–13;
New York, USA.
[3] Jongchul J., Taehyun S. and Gertsch, J. Vehicle full-state estimation and prediction system using state observers.
IEEE Trans Veh Technol. 2009;58:4078–4087.
[4] Ling-Yuan H, Tsung-Lin C. A vehicle roll-stability indicator incorporating roll-center movements. IEEE Trans
Veh Technol. 2009;58:2651–2652.
[5] Chen B, Peng H. Rollover warning of articulated vehicles based on a time-to-rollover metric. Proceedings of
ASME International Congress and Exposition;1999 Nov; Knoxville, TN.
[6] Chen B, Peng H. Differential-braking-based rollover prevention for sport utility vehicles with human-in-the-loop
evaluations. Veh Syst Dyn 2001;36(4–5):359–389.
[7] Ungoren AY, Peng H. Rollover propensity evaluation of an SUV equipped with a TRW VSC system. SAE
Transaction, 2001; 2001-01-0128.
[8] Hac A, Brown T, Martens J. Detection of vehicle rollover. SAE Transaction, 2004; 2004-01-1757.
[9] Yoon J, Cho W, Koo Bg, Yi K. Unified chassis control for rollover prevention and lateral stability. IEEE Trans
Veh Technol. 2009;58:596–609.
[10] Odenthal D, Bünte T, Ackermann J. Nonlinear steering and braking control for vehicle rollover avoidance.
Proceedings of the European Control Conference; 1999; Karlsruhe, Germany.
[11] Chadli M, Maquin D, Ragot J. Observer-based controller for Takagi–Sugeno modèels. Proceedings of the IEEE
International Conference on Systems, Man and Cybernetics; 2002; Tunisie.
[12] Dahmani H, Chadli M, Rabhi A, El Hajjaji A. Fuzzy observer for detection of impending vehicle rollover with
road bank angle considerations. Control & Automation (MED), 2010 18th Mediterranean Conference on; 2010
Jun 23–25. p. 1497–1502, doi:10.1080/00423114.2011.642806
[13] Takagi T, Sugeno M. Fuzzy identification of systems and its applications to modeling and control. IEEE Trans
Syst, Man, Cybernetics. 1985;15:116–132.
[14] Dahmani H, Chadli M, Rabhi A, El Hajjaji A. Fuzzy observer for detection of impending vehicle rollover with
road bank angle considerations. IEEE MED; 2010 Jun 23–25; Marrakech, Morocco.
[15] Oudghiri M, Chadli M, El Hajjaji A. Robust observer-based fault tolerant control for vehicle lateral dynamics.
Int J Veh Design. 2008;48(3–4):173–189.
[16] Xiaodonga L, Qinglingb Z. New approaches to H∞ controller designs based on fuzzy observers for TS fuzzy
systems via LMI. Automatica 2003;39:1571–1582.
[17] Guerra T, Kruszewski A, Vermeiren L, Tirmant H. Conditions of output stabilization for nonlinear models in
the Takagi–Sugeno’s form. Fuzzy Sets Systems. 2006;157(9):1248–1259.
[18] Tanaka K, Kosaki T. Design of a stable fuzzy controller for an articulated vehicle. IEEE Trans Syst, Man,
Cybernetics, Part B. 1997;27(3):552–558.
[19] Chadli M, Akhenak A, Ragot J, Maquin D. State and unknown input estimation for discrete time multiple model.
J Franklin Inst. 2009;346(6):593–610.
[20] Alexander L, Rajamani R, Starr P, Donath M. Road bank angle considerations in modeling and tilt stability
controller design for narrow commuter vehicles. In Proceedings of the 2006 American Control Conference;
2006 Jun 14–16; Minneapolis, MN, USA.
[21] Ackermann J, Bartelett A, Kaesbauer D, Sienel W, Steinhauser R. Robust control with uncertain parameters.
London: Springer; 1993.
[22] Ryu J, Christian Gerdes J. Estimation of vehicle roll and road bank angle. Proceedings of the 2004 American
Control Conference; 2004 Jun 30–Jul 2; Boston, MA.
[23] Kidane S, Alexander L, Rajamani R, Starr P, Donath M. Road bank angle considerations in modeling and tilt
stability controller design for narrow commuter vehicles. Proceedings of the 2006 American Control Conference;
2006 Jun 14–16; Minneapolis, MN, USA.
[24] Pacejka HB, Bakker E, linder L. A new tire model, an application in vehicle dynamics studies. SAE paper; 1989.
890089.
[25] Dahmani H, Chadli M, Rabhi A, El Hajjaji A. Road angle considerations for detection of impending vehicle
rollover. IFAC AAC; 2010; Munich, Germany.
[26] CarSim software. http:www.carsim.com
[27] Chen BC, Peng H. Rollover warning of articulated vehicles based on a time-to-rollover metric. Proceedings of
the 1999 ASME International Congress and Exposition; 1999 Nov; Knoxville, TN.
[28] Chen BC, Peng H. Differential braking based rollover prevention for sport utility vehicles with human-in-the-loop
evaluations. Veh Syst Dyn. 2001;36(4–5):359–389.
[29] Yua H, Güvençb L, Özgüünera. Heavy duty vehicle rollover detection and active roll control. Veh Syst Dyn.
2008;46(6):451–470.