Escolar Documentos
Profissional Documentos
Cultura Documentos
Abstract: Driving safely can be ensured by a better understanding of the risk and critical
situations. This can be achieved by a good knowledge of road attributes and vehicle dynamic
behaviors. This paper proposes two algorithms: the first one is dedicated to a new estimation
process which was designed to estimate the vehicle roll and road bank angles, in order to make
the road trajectory more realistic. This estimator is based on an unknown input sliding mode
observer and a LPV model. The second concerns a robust LPV state output feedback control
designed via LMI optimization and gain-scheduling method. The steering control strategy is
designed to simulate the non linear four wheel vehicle model under higher dynamic demands.
The steering vehicle control and the observer developed here have been validated experimentally
using the data acquired on the laboratory vehicle Peugeot 307 developed by INRETS-MA.
These algorithms were developed for an application known as Itinerary Rupture Diagnosis:
to evaluate the physical limits of a vehicle negotiating a bend.
data and discussed, then the speed extrapolation tests are of the vehicle. This vehicle model take advantage of the
presented. Section 6 presents some concluding remarks. main suspension compliances (i.e. toe in/out with lateral
Note that the experimental data used here were acquired forces), and all the parameters of the tyre Pacejka model
from the laboratory vehicle Peugeot 307 developed by which are obtained on test bed.
INRETS-MA see (Lechner et al. [2008]). Some measure-
ments performed by the VANI vehicle (a high efficiency 3. ROAD BANK ANGLE ESTIMATION USING
tool developed by the French Public Works network to UNKNOWN INPUT SLIDING MODE OBSERVER
record road attributes) were also used as a reference for
the bank angle (Dupre et al. [1998]). In this part of the work, a LPV model and an unknown
input sliding mode observer are presented as follows:
3.1 First model used to design unknown input observer
651
Preprints of the 18th IFAC World Congress
Milano (Italy) August 28 - September 2, 2011
In this approach, we assume that the components of the The steering vehicle control proposed here is designed in
unknown inputs u2 (t) are bounded, and that there is a two steps: the first is performed by solving a convex LMI
positive real number such as: optimization problem to design state feedback using the
ku2 (t)k f or all t (3) equation (6) (see gray boxes of the figure 4)). The second
step concerns the interpolation of the state feedback gains
Let x
(t) be an estimation of x(t), and e(t) be the estima- calculated in the first step using the Gain-Scheduling
tion error e(t) = x
(t) x(t). method (Stilwell et al. [1999], Bett [2005]) (see blue boxes
of the figure 4)).
The observability of pair (A, C) implies the existence of a
L np , such as the eigenvalues of (A LC) are located
in the open left half plane. The dynamical equation of the
estimated states x(t) using unknown input observer (see
figure 3) can be written as follows:
= A
x x + B1 (t)u1 + L(t)(y y) B2 (t)E (
y , y, ) (4)
where is a design parameter and k.k denotes the
standard Euclidean norm. The function E (e(t), ) is an
important element of the unknown input observer.
( F (Ce(t))
if F Ce(t) 6= 0
E (e(t), ) = r = kF (Ce(t))k (5)
fg
0 if F Ce(t) = 0
Fig. 4. Diagram block of steering vehicle control
652
Preprints of the 18th IFAC World Congress
Milano (Italy) August 28 - September 2, 2011
1.5 K(Vx)d/dt
Jmin = xT (0)Pj x(0) (9) K(Vx)y
1 K(Vx)
Gains K3 Variable
formulation of the inequality (10) is: 6 K(2,3)(Vx) Gains
K(3,4)(Vx)
5 Variable
xT (0) Gains K2
x(0) P
0 (11) 4 K(1,2)(Vx)
j 3
K
2 1
5 10 15 20 25 30 35 40 45
For the LPV model (6) and cost function (8), if there is Longitudinal speed [m/s]
653
Preprints of the 18th IFAC World Congress
Milano (Italy) August 28 - September 2, 2011
Index
[Deg.s1]
7
[Deg]
0.5 5
6.5
1 Measured 0 Interpolation method
Estimated 6
1.5 5 0 50 100 150 200 250 300 350 400 450
0 100 200 300 400 0 100 200 300 400 Distance travelled [m]
[m]
0 100 200 300 400 0 100 200 300 400 70
Distance travelled [m] Distance travelled [m] 0.02
60
0.04
Y [m]
40
since the laboratory vehicle is fitted with a 3D sonic Yaw angle error
0.4
30
anemometer.
20 0.2
[Deg]
2 10 0
0 0 0.2
[Deg]
2 10 0.4
0 100 200 300 400 0 200 400
X [m] Distance travelled [m]
4
Measured by VANI
Estimated
6
0 50 100 150 200 250 300 350 400 450
Distance travelled [m] Fig. 10. Trajectories: reference and tracking errors
Figures 11 compare performances of the system composed
Fig. 7. Road bank angle: estimated and measured of the two models (LPM and NL4WM) and steering
control: the main dynamic parameters are much close
The results shown in figures 8 to 14 were obtained using to the measurements for this trial involving high loads
the block diagram of the figure 1, and the road bank (lateral acceleration up to 6 m/s2 at the some bends).
estimated angle with unknown input sliding mode observer
Lateral acceleration Lateral speed
is considered in this trial. The Figures 8 and 10 show the 10 Measured 2
[km/h]
MNL4R
control using the LPV (dotted line) and the NL4W (solid 0 2
line) models in terms of trajectory tracking: the latter, 5 4
providing a more realistic vehicle behavior, produces fewer 0 100 200 300 400 0 100 200 300 400
Yaw rate Longitudinal acceleration
errors than the LPV bicycle model in relation to lateral 20 2
[m.s2]
10 0
steering input very close to the actual driver steering angle 0 2 Measured
654
Preprints of the 18th IFAC World Congress
Milano (Italy) August 28 - September 2, 2011
figures the vehicle model run with speed increment equal these two algorithms were presented. These two algorithms
to +27 km/h is in understeer situation. are used simultaneously to performe speed extrapolation
trials and explore the limit of behavior of the vehicle in
160
Longitudinal speed : Measured and NL4WM + Control law critical situations under high dynamic loads. The different
tests performed and the experimental tests highlight the
140
robustness of the algorithms developed here. These algo-
[km/h]
20
Vx + 27 km/h vol. 38, 2006, pp 357-364.
0 M. Corless and J. Tu. State and input estimation for a
20 class of uncertain systems Automatica, vol. 34, 1998, pp
0 50 100 150 200 250 300 350 400
Distance travelled [m] 757-764.
S. Hui and S. H. Zak. Observer design for systems with
Fig. 12. Speed extrapolation tests of the NL4WM coupled unknown inputs, Int. J. Appl. Math. Comput. Sci., vol.
with steering control: Extrapolated speed profiles and 15, 2005, pp 431-446.
steering angles L. Imsland, H. F. Grip, T. A. Johansen, T. I. Fossen, J. C.
Kalkkuhl and A. Suissa. On non-linear unknown input
10
NL4WM + Control law: Lateral Accelerations observers: applied to lateral vehicle velocity estimation
8
Measured = Vx
Vx + 9 km/h
on banked roads Int. Jour. of Cont., vol. 80, pp. 1741-
Vx + 18 km/h 1750, 2007.
6 Vx + 27 km/h
C. Edwards, S. K. Spurgeon and R. J. Patton. Sliding
4
mode observers for fault detection and isolation Auto-
[m.s2]
2
matica, vol. 36, pp. 541-553, 2000.
0 S. Boyd, L. El Ghaoui, E. Feron and V. Balakrishnan. Lin-
2 ear Matrix Inequalities in System and Control Theory
4
0 50 100 150 200 250 300 350 400
SIAM, Philadelphia, 1994.
Distance travelled [m] D. J. Stilwell and W. J. Rugh. Interpolation of Observer
State Feedback Controllers for Gain Scheduling IEEE
Fig. 13. Speed extrapolation tests of the NL4WM coupled Trans. aut. cont., vol. 44:6, pp. 1225-1229, June 1999.
with steering control: Lateral accelerations C. J. Bett. The electrical engineering handbook: controls
NL4WM + Control law: Lateral forces in function of wheel Sideslip angles
and systems: Gain-Scheduled Controllers Elsevier Aca-
In reference Pacejka model (black) Vertical Load: 1, 3, 5 and 7 KN demic, pp. 1107-1114, 2005.
5000 5000 V. Cerone, M. Milanese and D. Regruto. Combined
Fy Right Front [N]
Fy Left Front [N]
Vx + 27 km/h
Tieber. A passenger car driver model for higher lateral
0 0
accelerations Vehicle System Dynamics, vol. 45, no. 12,
pp. 1117-1129, 2007.
5000 5000
D. Lechner Embedded laboratory for vehicle dynamic
10 5 0 5
Sideslip Left Rear [Deg]
10 10 5 0 5
Sideslip Right Rear [Deg]
10 AVEC 2008, Kobe, Japan, 5-9 October 2008.
D. Lechner, C. Naude and L. Menhour DIARI: An em-
bedded application to determine the maximal achievable
Fig. 14. Speed extrapolation tests of the NL4WM coupled speed in a bend. Proc. of AVEC, UK, August 2010.
with steering control: tire lateral operating points G. Dupre, P. Flachat, G. Gratia, M. Latorre and J. C.
Olivier Detection dalerte securite liees `a des dysfonc-
6. CONCLUSION tionnements de linfrastructure routi`ere Bulletin des
Laboratoires des Ponts et Chaussees, vol. 213, 1998.
This paper addresses steering control and road bank angle L. B. Walcott and S. Zak State observation of nonlinear
observer. Steering control is based on a LMI optimiza- uncertain dynamical systems IEEE Trans. Automat.
tion and gain-Scheduling methods. The road bank angle Contr., vol. 32, pp. 166-170, 1987.
estimation is based on an unknown input sliding mode H. B. Pacejka. Tire and vehicle dynamics Elsevier,
observer. The most important steps in the designing of Netherlands, 2005.
655