V Semana Nacional de Ingeniera Electrnica

NONLINEAR OBSERVER FOR LATERAL VELOCITY USING ROLL ANGULAR ACCELERATION

Cuauhtemoc Acosta Lua*, Bernardino Castillo Toledo** and Stefano Di Gennaro*.

*Department of Electrical and Information Engineering and Center of Excellence DEWS
University of L'Aquila, Poggio di Roio, 67040 L'Aquila, Italy
E.mail: [email protected], [email protected]

**Centro de Investigacin y de Estudios Avanzados { CINVESTAV del IPN, Unidad Guadalajara

Av. Cientfica, Col. El Bajo, Zapopan, 45010, Jalisco, Mxico
E.mail: [email protected]

ABSTRACT
In this paper we present a nonlinear observer for
the estimation of the lateral velocity and the roll
position and velocity, based on measurements of
the longitudinal and lateral acceleration,
longitudinal velocity, yaw rate and the steer angle.
The nonlinear observer use a tyre-road friction
model by the Pacjeka magic formula and produce
the observer produce reliable estimates throughout
the linear and nonlinear handling region.
Exponentially stability of the observer is obtained.
The nonlinear observer has been validated by
numerical simulations to demonstrate the efficacy
of our nonlinear observer.
Keywords: Ground vehicles, Nonlinear Observer,
Vehicle Stability.
INTRODUCTION
The feedback control system for active safety in
automotive applications have entered production
cars. Many of these systems, which Electronic
Stability Program (ESP) [19], Vehicle Stability
Assist (VSA), Dynamic Stability Control (DSC),
and nonlinear control [1] and [15] have in common
that the control action depends on information
about vehicle velocity. However the lateral
velocity is rarely measured directly due to the cost
and space in the vehicle and must therefore
inferred form other measurement, such as
longitudinal and lateral acceleration, longitudinal
velocity, yaw rate and steer angle.
I.

7 al 9 de Octubre 2009, Ocotln, Jal. Mxico

Earlier works on observer for estimation of lateral

velocity without roll angular velocity are mainly
based on linear techniques [5], [4] or quasilinear
techniques for example [20], [17]. A nonlinear
observer linearizing the observer error dynamics is
proposed in [9] and [11]. The same type of
observer, in addition to an observer on forcing the
dynamics of the nonlinear estimation error to the
dynamics of a linear reference system is
investigated in [7]. Other types of observers linear
and nonlinear using the sliding mode techniques
are [3] and [14].
Other observer proposed found in literature are
based on the Estimation Kalman Filter (EKF). In
this case an EKF is used for estimating vehicle
velocity and tire forces in [13], thus without the
explicit use of friction models. An EKF based on a
tire-road friction model which also includes
estimation of the adhesion coef_cient and road
inclination angle is suggested in [16]. In [2], the
use of an EKF based on a nonlinear tire-road
friction model is considered, which also includes
estimation of cornering stiffness. The strategy
proposed in [10] combine dynamic and kinematics
models of the vehicle with numerical band-limited
integration of the equations to provide a side-slip
estimate. In [6] the side-slip angle is estimated
along with yaw rate in an approach which is
similar to the one considered herein, but without
yaw rate measurements.

39

V Semana Nacional de Ingeniera Electrnica

The goal of this work is to develop and efficiently

nonlinear observer for the vehicle velocity and roll
position using only the information available, i.e.
information from the sensor like longitudinal and
lateral acceleration, longitudinal velocity and yaw
rate in addition to the steer angle and principally
this observer guarantees the asymptotically
stability. A significative advantage of the proposed
approach over the EKF is that real-time solution of
the Riccatti differential equations is avoided then
our observer is implemented more efficiently
because the gains obtained in this observer are easy
for the implementation on the electronic control
unit.

ax , a y are longitudinal and lateral accelerations of

the vehicle, vx , v y are the longitudinal and lateral
velocity of CG of vehicle and \ , Z z are the yaw
angle and yaw rate, ms the so.called sprung mass,
h the center of gravity height. In (1), the vehicle
roll acceleration can be expressed as follows
bxD x  (k x  ms gh)D x  ms h  v y  Z z vx  (2)

J rDr

J r is the vehicle moment of inertia

momentum about the longitudinal axis, bx , k x are
the suspension roll damping and the suspension
roll stiffness respectively.
with

We proposed a kinematical model relating the

longitudinal velocity, lateral velocity, longitudinal
acceleration, lateral acceleration, yaw rate and
position and velocity roll dynamics. This model
contain an advantages that contain vehicle
parameters, and hence the kinematical model based
observer is unaffected by changes in the vehicle
parameters. The observer produce reliable
longitudinal, lateral velocity and position and
velocity roll angle throughout the linear and
nonlinear handling regions.
MATHEMATICAL PROBLEM
The dynamic models presented in this section are
used in the numerical simulations to estimate the
vehicle behavior. In the Figure 1, we show the
vehicle model used for the numerical simulation
within the study has four degrees of freedom:
longitudinal velocity vx , lateral velocity v y , yaw
II.

rate Z z and body roll D x . The vehicle longitudinal,

lateral and yaw dynamics can be expressed using
the so called bicycle model [8], [18].
m(vx  Z z v y )

ax  ms hZ zD x

m(v y  Z z vx )

a y  ms hZ zDx

J zZ z

(1)

l f Fy , f  lr Fy ,r

with m, J z the vehicle mass and inertia with

respect to the perpendicular axis, l f ,lr the distances
from the vehicle center of gravity (CG) to the tires,

7 al 9 de Octubre 2009, Ocotln, Jal. Mxico

Figure. 1 Linear bicycle model with roll degree

of freedom.
Hence, the system equations are
1
vx Z z v y   ax  ms hZ zZ x 
m

mh
v y Z z vx  k j a y  s (bxZ x  (k x  ms gh)D x )
Jr

Z z

1
 l f Fy , f  lr Fy ,r 
Jz

D x

Zx

(3)

mh

km bxZ x   k x  ms gh  D x  s a y
m

Jr
m
with km
and k j
.
mJ r  ms2 h 2
mJ r  ms2 h 2

Z x

40

V Semana Nacional de Ingeniera Electrnica

This model can be used to describe the vehicle

under the following assumptions
x Pitch dynamic are neglected.
x The system is rigid.
x We considered the passive suspension.

The dynamic of the estimation errors

ex vx  vx

ey

v y  v y

eD

D x  D x
Z x  Z x

eZ
The longitudinal and lateral accelerations have the
following expressions
ax P  Fx , f  Fx ,r 
ay

Fy , f , Fy ,r the front/rear lateral tyre forces, defined

as follows [12]

Di , j sin Ci , j arctan  Bi , jD j  i

x, y; j

f ,r

The constant Di , j , Ci , j , Bi , j can be determined

experimentally and the slip angles tyre are
v l Z
v l Z
D f Gd  y f z Dr  y r z
vx
vx
where G d is the driver angle.
In the following, ax , a yZ z and vx are supposed
measurable, and the aim is to determine an
observer to estimate the lateral velocity v y and the
position and velocity roll D x , Z x .
For, let we consider an observer of the following
form
vx
v y

1
 ax  ms hZzZx   k1 (vx  vx )
m

mh
Z z vx  k j a y  s (bxZ x  (k x  ms gh)D x )
Jr

Z z v y 

 k2 (vx  vx )

D x

Z x  k3 (vx  vx )

Z x

mh

km bxZ x   k x  ms gh  D x  s a y
m

 k4 (vx  vx )

ex
ey

P  Fy , f  Fy ,r 

where P is the tire-road friction coefficient and

Fx , f , Fx ,r the front/rear longitudinal tyre forces,

Fi , j

and its derivative

eD
eZ

1
ms hZ z ew
m
k1ex  k s kD eD  k s bx ew

k1ex  Z z ey 

(5)

k3ex  ew

k4 ex  km kD eD  kmbx ew
where ND k x  ms gh and the gains ki , i 1, , 4
can be selected to render them asymptotically
stable, under the following hypothesis.

Hypothesis 1: The yaw angular velocity Z z is

different from zero for all time instants t t t0 with
t0 the initial time.
For the previous hypothesis Z z t Zmin ! 0 for
t t t0 . In fact, let us consider the following
Lyapunov function candidate
1
1
V (t , e)
J ex2  ey2  G1sign(Z z )ex ey  eD2
2
2
(6)
1 2
 J 1eZ  G 2Z z ,min ey eD
2
with G1 , G 2 , J , J 1 ! 0 constants and the sign(Z z )
represent the signum function of the yaw rate.
Hence,
1

V (t , e) J ex k1ex  Z z ey  ms hZ z eZ
m

ey  k2 ex  k sND eD  k s bx eZ 

G1sign(Z z )ex  k2 ex  ksND eD  k s bx eZ 

1

G1sign(Z z )ey k1ex  Z z ey  ms hZ z eZ

m

G1G D (Z z )Z z ex ey  J 1eZ   k4 ex  kmND eD  kmbx eZ 

G 2Z z ,min ey  k3  eZ   eD  k3  eZ 
G1Z z ,min eD  k2 ex  ksND eD  ks bx eZ 
with G D the Dirac distribution.

7 al 9 de Octubre 2009, Ocotln, Jal. Mxico

41

V Semana Nacional de Ingeniera Electrnica

Choosing
Under Hypothesis (1), G D Z z  0 , so that
V (t , e)   J k  k G sign Z   e 2  G Z e 2
1

2 1

G1 !

ksND G 2Z z ,min eD  J 1kmbx eZ

(7)

  ms hZ z  ks bxG1sign Z z   J 1k4 ex eZ
m

D1  1  O1kmND  ! 0
O3 ksND  ks bx Z z ,min
O3

from (8)
1
G Z  JG1 Z z
k1
2
J  JG 2 Z z ,min  G1 1 z
G1G 22Z z2,min Z z  k sND G12G 2Z z ,min

 ksND ey eD  1  J 1kmND  ks bxG 2Z z ,min  eD eZ

k2

 ks bx  G1ms h Z z  G 2Z z ,min ey eZ
m

with Z z ,min min Z z , and

J k1  G1sign Z z  k2 G1 Z z

J
m

ms hZ z  ks bxG1sign Z z   J 1k4

(8)

D 2 ey eZ d D 2

Z z ,min
1

O2

ey2  D1

Z z ,min
Zz

eD2

ey2  D 2 O2 eZ2

e 2  D 3O3eZ2
O3 D
where Z z t Z z ,min ! 0 and from the Hypothesis (1)

D 3eZ eD d D 3

J 2G Z Z
J  JG Z z ,min  G1 2 z ,min z

and O2 , O3 ! 0 , one gets

Zz
1
 D 2 ey2
V (t , e) d G1Z z ex2  G1 Z z  D1

Z z ,min O2

Z
1
 k sND G 2Z z ,min  D1 z ,min  D 3 eD2

Z z O3

  J 1kmbx  O2D 2  O3D 3  eZ2

7 al 9 de Octubre 2009, Ocotln, Jal. Mxico

G12G 2Z z ,minZ z  k sND G1sign(Z z ) J 2  G12

1 J

 ms hZ z  k s bxG1sign(Z z )
k4

J1 m

where J !

Zz

2
2

Considering the following inequalities

D1ey eD d D1

J k sND G1G 2Z z ,min sign(Z z )

k3

k2G 2Z z ,min  ksND G1sign Z z   k3

Z z J 2  G12

J  JG Z z ,min  G1
2
2

t t t0

JZ z  k2  k1G1sign Z z   k3G 2Z z ,min

G 2Z z ,min 1

 ks bx ! 0
D1 
D2
O2

O2

G2 !

  JZ z  k2  k1G1sign Z z   k3G 2Z z ,min  ex ey

  k2G 2Z z ,min  ksND G1sign Z z   k3  ex eD

mO2  ms h 1

mO2
Z z ,min

G12
.
1  G 22Z z2,min

Therefore, one finally obtains

V (t , e) d G1 Z z ex2  OG ,1 Z z ey2

(9)

OG ,2Z z ,min eD2  OG ,3eZ2

with OG ,1 ! 0, OG ,2 ! 0 and OG ,3 1  J 1kmND  ! 0

then the equation (9) is negative definite, hence the
error system (5) is exponentially stable.
SIMULATION RESULTS
This section presents the behavior of the nonlinear
observer for the lateral velocity and roll position
and velocity. The simulation consists of a single
lane maneuver with the longitudinal velocity at
100 km/h Figure (2). The results obtained by the
kinematics model are shown in the Figure (3,4,5).
In the Figure (3a) we shown the effectiveness of
the nonlinear observer for the lateral velocity
III.

42

V Semana Nacional de Ingeniera Electrnica

because dynamic observer tracks the state v y ; in

the Figure (3b) we present the yaw rate Z z used in
this simulation. In the Figure (4) is shown the roll
position D x and velocity Z x and its observer
D x ,Z x . This observer is obtained only with
information based on measurements of the
longitudinal and lateral acceleration, longitudinal
velocity, yaw rate and the steer angle. Finally, in
the Figure (5) we present the errors where we can
see the error is identically to zero.

Figure 4. a) Position Roll D x (solid) and its

observer D x (dotted) [m]; b) Velocity roll Z x
(solid) and its observer Z x (dotted) [m].

Figure 2. a) Single lane change maneuver

G d [deg/s]; b) Longitudinal velocity vx (solid)
and its observer vx (dotted) [m/s].

Figure 5. Errors a) ex
eD

D x  D x and d) eZ

vx  vx ; b) ey

v y  v y ; c)

Z x  Z x .

CONCLUSIONS
This paper proposes the nonlinear observer for the
lateral velocity using the dynamics of roll position
and velocity. This observer is based in the
available information for the common sensor in the
vehicle, i.e. the longitudinal and lateral
acceleration ax , a y , longitudinal velocity vx , yaw
IV.

Figure 3. a) Lateral velocity v y (solid) and its

observer v y (dotted)[m/s]; b) Yaw rat Z z [rad/s].

7 al 9 de Octubre 2009, Ocotln, Jal. Mxico

rate Z z and the steer angle G d .

43

V Semana Nacional de Ingeniera Electrnica

Therefore, the observer is not sensitive to vehicle

parameter changes. The mathematical model used
in this article is the bicycle model together the
dynamics of passive suspension of utility vehicle.
Finally, we demonstrated the theory of a nonlinear
observer and its effectiveness with numerical
simulations where the error between the system
and the observer states tends exponentially to zero.
V.

[1]

[2]

[3]

[4]

[5]

[6]

[7]

REFERENCIAS
C. Acosta-Lua, B. Castillo Toledo and S.
Di Gennaro Output Feedback Regulation
of Electromagnetic Valves for Camless
Engines Proceedings of the 2007
American Control Conference, New York
City, USA, July 11-13, 2007
M. C. Best, T. J. Gordon and P.J. Dixon,
An extended adaptive Kalman filter for
real-time state estimation of vehicle
handling dynamics, Vehicle System
Dynamics, Vol 34, pp. 5775. 2000.
G. Baffet, J. Stphant and A. Charara,
Lateral
vehicle-dynamic
observers:
simulations and experiments, International
Journal of Vehicle Autonomous Systems,
vol. 5, num. 3-4, pp. 184-203, December,
2007.
J. Farrelly and P. Wellstead, Estimation of
vehicle lateral velocity, Proceedings of the
1996 IEEE international conference on
control applications pp. 552557, 1996.
Y. Fukada, Slip-angle estimation for
stability control, Vehicle Systems
Dynamics, vol 32, pp. 375388. 1999.
A. Hac and M. D. Simpson, Estimation of
vehicle side slip angle and yaw rate, In
SAE 2000 World Congress, Detroit, MI,
USA., 2000.
M. Hiemer, A. VonVietinghoff, U. Kiencke
and T. Matsunaga, Determination of
vehicle body slip angle with non-linear
observer strategies, In Proceedings of the
SAEWorld Congress, Paper no. 2005-010400, 2005.

7 al 9 de Octubre 2009, Ocotln, Jal. Mxico

[8]

[9]

[10]

[11]
[12]

[13]

[14]

[15]

[16]

[17]

[18]

G. J. Heydinger, W. R. Garrott, J. P.
Chrstos, and D. A. Guenther, A
methodology for validating vehicle
dynamics simulations, SAE, Paper
900128, 1990.
U. Kiencke and A. Daiss, Observation of
lateral
vehicle
dynamics,
Control
Engineering Practice, Vol 5(8), pp.
11451150, 1997.
J. Lu and T.A. Brown, Vehicle side slip
angle estimation using dynamic blending
and
considering
vehicle
attitude
information, Patent US 6671595, 2003.
U. Kiencke and L. Nielsen, Automotive
control systems, Berlin: Springer, 2000.
H. B. Pacejka, Tyre and vehicle
dynamics,
London:
ButterworthHeinemann
L. R. Ray, Nonlinear state and tire force
estimation for advanced vehicle control,
IEEE Transactions on Control Systems
Technology, Vol. 3(1), pp. 117124, 1995
J. Stphant, A. Charara and D. Meizel,
Evaluation of sliding mode observer for
vehicle
sideslip
angle
,Control
Engineering Practice, vol. 15, pp. 803-812,
2007.
P. Setlur, J. R. Wagner, D. M. Dawson and
D. Braganza, A Trajectory Tracking
SteerbyWire Control System for Ground
Vehicles, IEEE Transactions on Vehicular
Technology, Vol. 55, No. 1, pp. 7685,
2006.
A. Suissa, Z. Zomotor and F. Bttiger,
Method for determining variables
characterizing vehicle handling. Patent
US 5557520, 1996.
A.Y. Ungoren, H. Peng and H. Tseng, A
study on lateral speed estimation methods,
International
Journal
on
Vehicle
Autonomous Systems, Vol 2, pp. 126144.
2004.
J. Wong, Theory of Ground Vehicles.
New York: Wiley, 1978.

44

V Semana Nacional de Ingeniera Electrnica

[19]

[20]

A. T. Van Zanten, Bosch ESP system: 5

years of experience, In Proceedings of the
automotive
dynamics
and
stability
conference, pp. 354, Paper no. 2000-011633, 2000.
P. J. Venhovens, and K. Naab, Vehicle
dynamics estimation using Kalman filters,
Vehicle System Dynamics, Vol. 32, pp.
171-184. 1999.

APPENDIX
We considered simulations based on data from a
prototype vehicle obtained of the Ford Motor
Company. The nominal parameters are
m= 1880 kg.
J z = 830 kg m 2
lr= 1.53 m
lf= 1.38 m
l= f+lr m
P=
J r = 71 kg m 2
ms = 9 kg

Studies and Research Center, National Polytechnic

Institute, Guadalajara, Mxico, in 2003 and 2007,
respectively. Since April 2008, he has been with
the Department of Electrical and Information
Engineering, University of LAquila, LAquila,
Italy, where he was External Consulter in a project
between the Center of Excellence DEWS and Ford
Forschungzentrum Aachen. His research interest
focus on regulator theory, hybrid systems,
regulation theory, applications of nonlinear
control, automotive control systems.

VI.

g= .81 m/ s 2
h= 0.3 m
bx 1000 N m rad/s
k x 16000 N m/rad
B f , y 7.2
C f ,y

1.81

Df ,y

8854

Br , y

11

Cr , y

1.68

Dr , y

8394

AUTHORS
Cuauhtemoc Acosta-Lua was born in Lzaro
Crdenas,Michoacn, Mxico, in 1977. He
received the B.Sc. degree in electronic engineering
from the Morelia Technologic Institute, Morelia,
Mexico in 1996 and the M.Sc. and Ph.D. degrees
in electrical engineering from the Advanced
VII.

7 al 9 de Octubre 2009, Ocotln, Jal. Mxico

Bernardino Castillo-Toledo received the B.Sc.

degree in electrical engineering from the National
Polytechnic Institute (IPN), Mexico City, Mexico,
the M.Sc. degree from the Research and Advanced
Studies Center of the National Polytechnic Institute
(CINVESTAV-IPN), Guadalajara, Mexico, and the
Ph.D. degree from the University of Rome La
Sapienza, Rome, Italy, in 1981, 1985, and 1992,
respectively. He was a Lecturer with the School of
Electrical and Mechanics Engineering, IPN, from
1985 to 1989. From 1985 to 1995, he was with the
Automatic Control Section, Department of
Electrical Engineering, CINVESTAV-IPN, and,
since 1995, with CINVESTAV-IPN, Guadalajara
Campus. He held several research stages with the
University of Rome La Sapienza and University of
LAquila, LAquila, Italy, and was a visiting
Professor at the Laboratoire dAutomatique et
dAnalyse des Systemes, French Council for
Scienti_c Research, during 1999 to 2000, and at
the University of Compiegne, Compiegne, France,
during the first semester of 2002. His main
research interests include nonlinear control design,
the robust regulation problem, and the application
of artificial neural networks and fuzzy logic
techniques to control and fault diagnosis of
dynamical systems.
Stefano Di Gennaro received the M.S. degree
(summa cum laude) in nuclear engineering and the
Ph.D. degree in system engineering from the
University of Rome La Sapienza, Rome, Italy, in
1987 and 1992, respectively. Since October 1990,
he has been with the Department of Electrical and
Information Engineering, University of LAquila,
LAquila, Italy, where he was an Assistant
45

V Semana Nacional de Ingeniera Electrnica

Professor of Automatic Control and, since 2001,

has been an Associate Professor of Automatic
Control. He holds courses on automatic and
nonlinear controls. He is also with the Center of
Excellence DEWS Design methodologies of
Embedded controllers, Wireless interconnect and
Systems on- chip, University of LAquila. In 1986,
he was a Visiting Scientist at the Nuclear Research
Center ENEA, Casaccia. He has been a Visiting
Professor at the Laboratoire des Signaux et
Systmes, Centre National de la Recherche
Scientifique (CNRS), Paris, at the Department of
Electrical Engineering, Princeton University,
Princeton, NJ, at the Department of Electrical
Engineering and Computer Science, University of
California, Berkeley, and at the Advanced Studies
and Research Center, National Polytechnic
Institute, Unidad Ciudad de Mxico and Unidad
Guadalajara, Mxico. He is working in the area of
hybrid systems, regulation theory, and applications
of nonlinear control.

7 al 9 de Octubre 2009, Ocotln, Jal. Mxico

46