See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/269293449

A LPV/H;∞ Fault tolerant control of vehicle roll dynamics under semi-active

damper malfunction

Conference Paper  in  Proceedings of the American Control Conference · June 2014

DOI: 10.1109/ACC.2014.6859081

CITATIONS READS

10 76

3 authors:

Fergani Soheib Olivier Sename

Laboratoire d’Analyse et d’Architecture des Systèmes (LAAS) Grenoble Institute of Technology
61 PUBLICATIONS   272 CITATIONS    372 PUBLICATIONS   4,699 CITATIONS   

SEE PROFILE SEE PROFILE

Luc Dugard
French National Centre for Scientific Research
431 PUBLICATIONS   6,442 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

LPV and Time-Delay Systems View project

Active side-stick View project

All content following this page was uploaded by Fergani Soheib on 28 September 2015.

The user has requested enhancement of the downloaded file.

 A LPV/Hinf fault tolerant control of vehicle roll
dynamics under semi-active damper malfunction
Soheib Fergani, Olivier Sename, Luc Dugard

To cite this version:

Soheib Fergani, Olivier Sename, Luc Dugard. A LPV/Hinf fault tolerant control of vehicle roll
dynamics under semi-active damper malfunction. 2014 American Control Conference (ACC
2014), Jun 2014, Portland, OR, United States. 2014. <hal-00940024>

HAL Id: hal-00940024

https://hal.archives-ouvertes.fr/hal-00940024
Submitted on 31 Jan 2014

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci-
entific research documents, whether they are pub-
lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
 CONFIDENTIAL. Limited circulation. For review only.
A LPV/H∞ fault tolerant control of vehicle roll dynamics under semi-active
damper malfunction
S. Fergani1 , O. Sename1∗ , L. Dugard1 ,
Abstract— This paper proposes a LPV/H∞ fault tolerant
control strategy for roll dynamics handling under semi-active
damper’s malfunction. Indeed, in case of damper’s malfunction,
a lateral load transfer is generated, that amplifies the risks of
vehicle roll over.
In this study, the suspension systems efficiency is monitored
through the lateral (or longitudinal) load transfer induced by
a damper’s malfunction.
The information given by the monitoring system is used
in a partly fixed LPV/H∞ controller structure that allows to
manage the distribution of the four dampers forces in order
to handle the over load caused by one damper’s malfunction.
The proposed LPV/H∞ controller then uses the 3 remaining
healthy semi-active dampers in a real time reconfiguration.
Moreover, the performances of the car vertical dynamics
(roll, bounce, pitch) are adapted to the varying parameter given
by the monitoring of the suspension system efficiency, which
allows to modify online the damping properties (soft/hard) to
limit the induced load transfer.
Simulations are performed on a complex nonlinear full
vehicle model, equipped by 4 magneto-rheological semi-active
dampers. This vehicle undergoes critical driving situations, and
only one damper is considered faulty at ones. The simulation
results show the reliability and the robustness of the proposed
solution.
Keywords: LPV/H∞ control, semi-active suspension, fault
tolerant control, .
I. I NTRODUCTION
Vehicle vertical dynamics are affected by many
interrelated sub-systems of the car aim at improving
passengers comfort and especially vehicle safety and road
holding. Among all sub-systems affecting the vertical
vehicle dynamics, suspension systems play a key role for
vehicle handling in critical situation since they ensure the
link between the wheels and the chassis, see [1]–[3]. Several
types of suspension systems have been developed and
commercialized. In the last decade, semi-active suspensions
have received a lot of attention by both academic and
industrial communities, see [4]–[7], since they provide the
best compromise between cost (energy, volume, and number
of sensors) and performance (road holding, comfort and
vehicle behaviour). In this work, a specific type of semi-
active suspension is under interest, namely, the Magneto
Rheological Dampers (MRDampers, see [8]–[10]).
1 GIPSA-lab, Control Systems Department, Grenoble
University - Grenoble INP, ENSE3 - Domaine Universi-
taire BP46, 38402 Saint Martin d’Hères - Cedex France.
olivier.sename@gipsa-lab.grenoble-inp.fr
Preprint submitted to 2014 American Control Conference.
Received September 27, 2013.
While some of the authors works have been concerned
with global chassis control using active or semi-active sus-
pension [11], [12], the fault tolerant control problem of such
systems has been considered only in [13] where a pre-defined
distribution of the suspension forces (computed from the
steady state behaviour) is used o compensate a damper oil
leakage.
This study focuses on the fault tolerant control recon-
figuration of MR semi-active dampers. Indeed, few works
have been concerned with the control reconfiguration in the
presence of suspension system malfunctions or failures.
While detecting a damper malfunction, the proposed strategy
aims at keeping the vehicle stability and performance through
an adequate distribution of the 3 remaining healthy actuators.
The characteristics of magneto rheological dampers allow
to compensate the lack of the vertical force in the faulty
suspension corner by reconfiguring the global suspensions
control.
To solve that problem a new LPV/H∞ fault tolerant
control is introduced to manage the deterioration of the
vertical dynamics by using a varying parameter that
coordinate the use of the healthy dampers. The main idea
involves 2 steps. First, a monitoring system is introduced
to evaluate the state of health of the suspension system.
Here, the load transfer induced b a damper malfunction is
considered, but different methods could be integrated in the
proposed control strategy (observers, parity space, ...). Then
the global suspension control is scheduled according to
the monitor parameter to adapt on-line the damper control
distribution, and the performances of the suspension systems
as well (in term of comfort and road holding).
To achieve these objectives, the authors have chosen to
fix the structure of the LPV/H∞ controller by making the
LMI’s orthogonal with parameters dependency, as follow:
 H∞ 
uf l (t)
uH∞ (t)
 fr
uH∞ (t)
 = U (ρ)Cc0 (ρ) xc (t) (1)
rl
| {z }
uH
rr (t)
∞ Cc (ρ)
The suspension forces distribution is obtained through the
matrix U (ρ):
 
ρ1 0 0 0
 0 ρ2 0 0 
U (ρ) = 
 0 0 ρ3 0 
 (2)
0 0 0 ρ4
 CONFIDENTIAL. Limited circulation. For review only.

where ρi are the varying parameters given by the
considered suspensions monitoring strategies.
Remark 1: This kind of structure has been used by the
authors for vehicle dynamics control with braking, steering
and suspension actuators [14], [12].
main dynamical equations are given in equation (4),
where Ftxi = Ftxil + Ftxir , Ftyi = Ftyil + Ftyir ,
Ftzi = Ftzil + Ftzir are the tire forces (based on Pacejka
tire non linear model) and Fszi = Fszil +Fszir , (i = {f, r}).

Here, this approach is extended to account for suspension

actuator’s malfunction. Since roll dynamics affect very much B. Vertical modeling
the vehicle behaviour, the authors have chosen to schedule
the suspension control using the lateral load transfer as a
The model used for the controller synthesis is the linear
varying parameter (ρl ). The controller output matrix shows
vertical 7-DOF model. It includes several vertical dynamics
the dependency on this varying parameter and ensures the
as the chassis acceleration z̈s , the four wheels accelerations
suspension efforts reconfiguration, as follows:
z̈usij , the roll bounce acceleration θ̈ and the pitch acceler-
ation φ̈.
 
1 − ρl 0 0 0
 0 ρl 0 0
U (ρl ) =   (3)
 0 0 1 − ρl 0 
0 0 0 ρl
C. Semi-active Magneto-rheological damper

The paper is organised as follows: Section 2 briefly

In this study, the proposed strategy is applied to a vehicle
presents the vehicle and MR damper models used for
equipped by four semi-active MR dampers. There are various
synthesis and validation purposes. Section 3 is devoted to
approaches to model semi-active dampers. In the parametric
the main contribution of the paper, i.e a LPV/H∞ fault
model of [15], the hysteresis loop force-velocity is well
tolerant control of vehicle roll dynamics. The performance
modeled by an hyperbolic tangent function.
analysis is done in Section 4 with time domain simulations
performed on a complex nonlinear full vehicle model.
Conclusions and future works are given in the last section.
ms zs
Paper notations: ks FMR
Throughout the paper, the following notations will be m us z us
adopted: indices i = {f, r} and j = {l, r} are used kt
to identify vehicle front, rear and left, right positions re- zr
spectively. Then, index {s, t} holds for forces provided by
suspensions and tires respectively. {x, y, z} holds for forces
and dynamics in the longitudinal,
q lateral and vertical axes
respectively. Then let v = vx2 + vy2 denote the vehicle Fig. 1. QoV model for a semi-active suspension in a vehicle.
speed, Rij = R − (zusij − zrij ) the effective tire radius,
m = ms + musf l + musf r + musrl + musrr the total vehicle
mass. The model parameters are those of a Renault Mégane The MR damping force is given by:
Coupé, obtained during a collaborative study with the MIPS
laboratory in Mulhouse, through identification with real data,
see [14]. FM R = Ifc tanh (a1 żdef + a2 zdef ) + b1 żdef + b2 zdef (5)
II. F ULL VEHICLE MODELING
A. Full vehicle model where the electric current is bounded between
The model (4) used in this work is a nonlinear full 0 ≤ Imin ≤ I ≤ Imax ≤ 2.5. Imin and Imax depend
vehicle model. Details of this model and the corresponding on the MR damper specifications. Experimental data
parameters can be found in [14]. It involves several car obtained from a commercial MR damper are used to
chassis dynamics: vertical (zs ), longitudinal (vx ), lateral model the nonlinearities of this actuator by using (5). The
(vy ), roll (θ), pitch (φ) and yaw (ψ). It also models the parameters of the MR damper model used in this analysis
vertical and rotational motions of the wheels (zusij and ωij are: f c = 600.9, a1 = 37.8, a2 = 22.1, b1 = 2830.8 and
vij −Rij ωij cos βij
respectively), the slip ratios (λij = max(v ) b2 = −7897.2.
ij ,Rij ωij cos βij )
and the center of gravity side slip angle (βcog ) dynamics
as a function of the tires and suspensions forces. The The QoV system dynamics, given in a state-space repre-

Preprint submitted to 2014 American Control Conference.

Received September 27, 2013.
 CONFIDENTIAL. Limited circulation. For review only.




 v̇x = − Ftxf cos(δ) + Ftxr + Ftyf sin(δ) /m − ψ̇vy


v̇y = − Ftxf sin(δ) + Ftyr
+ Ftyf cos(δ) /m + ψ̇vx




z̈s = − Fszf + Fsz
r + Fdz /ms




 z̈usij = Fszij − Ftzij /musij





θ̈ = (Fszrl − Fszrr )tr + (Fszf l − Fszf r )tf + mhv̇y /Ix (4)
φ̈ = Fszf lf − Fszr lr − mhv̇x )/Iy





ψ̈ = lf (−Ftxf sin(δ) + Ftyf cos(δ)) − lr Ftyr + (Ftxf r − Ftxf l )tf cos(δ) − (Ftxrr − Ftxrl )tr + Mdz /Iz




(Rij Ftxij − Tbfij )/Iw

ω̇ij =





 β̇
cog = (Ftyf + Ftyr )/(mvx ) + ψ̇

sentation, is written as: z3 zr zs z1

Wzr Wzs (ρl )
Full vehicle model
    
żs 0 1 0 0 zs
 z̈s  − ksm+b2 − mb1 ks +b2
ms
b1
ms
  żs  with LPV MR dampers
 = s s z2
Wθ (1 − ρl )
  
żus   0 0 0 1  zus  zus
Model
ks +b2 b1
z̈us mus mus − ks +k
mus
t +b2
− mbus
1
żus
| {z } | {z } | {z } Fsij
ẋ A x
ay
0
0
 
θ
 −·ρfc
 
ms 0  I Wf ilter Σ(ρ1, ρ2)
żdef
+  
zdef
0 0  zr
·ρfc kt |{z}
m m
| us {z us } u
B   ρ1, ρ2
    zs
y1 1 0 −1 0   żs 

=
y2 0 1 0 −1 zus  ρl
|{z} | {z } ż ρl Varying parameter
y us K(ρ1, ρ2, ρl )
C
(6) generation
where, ρ = tanh [a1 żdef + a2 zdef ] ∈ [0, 1] is a varying
parameter, the accelerometers of the sprung (z̈s ) and un-
sprung mass (z̈us ). These measurements are related to the Fig. 2. Global chassis control implementation scheme.
comfort and road holding performances, that depend on the
semi-active damper properties and obviously on the road
irregularities.
one is used for the control reconfiguration and adaptation to
III. D ESIGN OF T HE LPV/H∞ FAULT TOLERANT critical driving situations with damper malfunction, the two
CONTROL OF VEHICLE ROLL DYNAMICS UNDER others parameters are needed to account for the dissipativity
SEMI - ACTIVE DAMPER MALFUNCTION and saturation of the semi-active MR , as in [16].
In this paper, a new LPV/H∞ fault tolerant control strat- A. LPV QoV model ρ1 and ρ2 :
egy is based on the monitoring of the semi-active dampers.
When a fault is detected on one of the four semi-active Indeed, the semi-active force is modeled following [8], as:
dampers (i.e a lack in the vertical forces), the roll dynamics
are amplified, causing vehicle instability and increasing car Fsa = b1 (żsi − żusi ) + b2 (zsi − zusi ) + I · fc · ρ (7)
| {z } | {z }
roll-over risks. To manage this instability, the proposed passive semi−active
LPV/H∞ suspension control is scheduled thanks to ρl the
where I is the electric current to control the semi-
load transfer generated by the roll bounce of the vehicle
active force based on the desired performances and ρ =
ρl (by comparing the righ/letf forces) and tunes the 3
tanh [a1 żdef + a2 zdef ] ∈ [0, 1] represents the nonlinearities
remaining healthy dampers to achieve fault compensation
of the shock absorber. In the control synthesis for FTC,
without reaching saturation. In addition, the performance
the varying parameters ρ1 and ρ2 allow to ensure that
objectives are set thanks to this varying parameter ρl which
the suspension control meets the semi-activeness and the
is included in the considered weighting functions on chassis
saturation damper’s constraints, respectively. Then, in this
displacement Wzs and the roll dynamics of the car Wθ .
paper the suspension in each corner is modeled as:
Scheduling parameters: 
ẋlpv = Alpv (ρ1 , ρ2 ) xlpv + B1 uc + B2 w
(8)
This strategy given in Fig. 2 includes 3 varying parameters. ylpv = C1 xlpv

Preprint submitted to 2014 American Control Conference.

Received September 27, 2013.
 CONFIDENTIAL. Limited circulation. For review only.

where 
xs
T suspension systems, ρl → 1, penalizing the provided output
xlpv = ,
 xf  suspension force on the faulty corner, changing the level of
As + ρ2 Bs2 Cs2 ρ1 Bs Cf saturation depending on the detected fault. Also, an overload
Alpv (ρ1 , ρ2 ) = ,
0 Af
     T appears on the right side. To managed that, the lacking
0 Bs1 Cs
B1 =
Bf
, B2 =
0
, C1 =
0
suspension effort is compensated by the 3 healthy dampers
to stabilise the vehicle. Indeed, left suspensions are set to
Cf xf "hard" to handle the overload caused by the loss of one
ρ1 = tanh(Cs2 xs ) tanh( ) CF1x ,
F1
tanh(Cs2 xs )
f f of the right side dampers. On the other side, suspensions
ρ2 = Cs2 xs are relaxed and tuned to "soft" for the remaining healthy
xs , As , Bs , Bs1 , Bs2 , Cs and Cs2 are the state and matrices actuators (since the overload is on the other side) and a level
of a state-space representation of the QoV model by includ- of saturation is applied to the faulty one depending of the
ing the MR damper model in (6) and considering zdef and degree of deterioration detected. This distribution is handled
żdef as output; xf , Af , Bf , Cf are the state and matrices of thanks to the specific structure of the suspension controller,
a representation of the low-pass filter Wf ilter = ωf /(s+ωf ) given as follows :
which is added to the system to make the control input
ẋc (t) = Ac (ρ1 , ρ2 , ρl )xc (t) + Bc (ρ1 , ρ2 , ρl )y(t)

matrices parameter independent. 



B. LPV controller structure scheduled by ρl :

  H∞ 
uf l (t)
Ks (ρ) :=
The third scheduling parameter, ρl , acts in the presence  uH ∞
f r (t) = U (ρ )C 0 (ρ , ρ ) x (t)
l 1 2 c
  H c
∞

u rl (t)

of damper malfunction, which can be seen directly on the 

 | {z }
lateral load transfer of the vehicle. This parameter, defined uH ∞
rr (t) Cc (ρ1 ,ρ2 )
(12)
as follows, allows the right/left suspension control reconfig-
where xc (t) is the controller state, Ac (ρ1 , ρ2 , ρl ),
uration:
 Bc (ρ1 , ρ2 , ρl ) and Cc (ρ1 , ρ2 , ρl ) controller scheduled

 F zl
= m s × g/2 + m s × h × a y /lf by ρl while ρ1 and ρ2 ensure the semi-activeness of the
dampers. uH∞ (t) = [uH H∞ H∞ H∞
 ∞
f l (t)uf r (t)url (t)urr (t)] the



Fzr = ms × g/2 − ms × h × ay /lr

input
(9) control of the suspension actuators and y(t) = zdef (t).

C. The suspension control problem formulation

ρl = |(δf l Fzf l + δrl Fzrl ) − (δf r Fzf r + δrr Fzrr )|




/|(Fzf l + Fzrl + Fzf r + Fzrr )|; In this study, a 7 DOF vehicle model is considered, (see

(II-B)and augmented with LPV damper model (7) for each
with δij : the suspension systems efficiency given by the
corner of the vehicle.
considered monitoring system, Fzij : the vertical forces, ay
The suspension control with performance adaptation (see
lateral acceleration, ρl ∈ [0 1]: the monitoring parameter.
[17]) is presented. The following H∞ control scheme is
The innovative solution which aims at stabilizing the vehicle
considered, including parameter varying weighting functions.
in the presence of damper failure is the following: the 2
+2ξ11 Ω11 s+Ω11 2
where Wzs = ρl ss2 +2ξ 2 is shaped in order to
controller has a partly fixed structure obtained by by making 12 Ω12 s+Ω12
reduce the bounce amplification of the suspended mass (zs )
the LMIs structure orthogonal with a parameter dependency
between [0, 12]Hz.
on the control output matrix, as follow: 2
+2ξ21 Ω21 s+Ω21 2
 H∞  Wθ = (1 − ρl ) ss2 +2ξ 22 Ω22 s+Ω22
2 attenuates the roll bounce
uf l (t) amplification in low frequencies.
uH∞ (t) Wu = 3.10−2 shapes the control signal.
 fr  = U (ρl )Cc0 (ρl ) xc (t) (10)
uH∞ (t) Remark 3.1: The parameters of these weighting functions
rl
| {z }
H∞
urr (t) C c (ρl ) are obtained using genetic algorithm optimization as in [16].

The suspension forces distribution is obtained with the matrix According to Fig. 2, the following parameter dependent
U (ρl ):   suspension generalized plant (Σgv (ρ1 , ρ2 , ρl )) is obtained:
1 − ρl 0 0 0
 0 ρl 0 0
 ξ˙ = A(ρ1 , ρ2 , ρl )ξ + B1 w̃ + B2 u

U (ρl ) =   (11)
 0 0 1 − ρl 0 
0 0 0 ρl Σgv (ρ1 , ρ2 , ρl ) := z̃ = C1 (ρ1 , ρ2 , ρl )ξ + D11 w̃ + D12 u
y = C2 ξ + D21 w̃ + D22 u

The parameter ρl defined in (9) generates the adequate (13)
suspension forces distribution, depending on the load transfer where ξ = [χvert χw ]T ; z̃ = [z1 z2 z3 ]T ;
(left  right) caused by the critical situation. w̃ = [zrij Fdx,y,z Mdx,y ]T ; y = zdefij ; u = uH ∞
ij ;
This suspension tuning is achieved as follows: When one and χw are the vertical weighting functions states.
of the suspension dampers is faulty, a load transfer is then
generated and influences the vehicle stability and handling. One of the main interesting contributions is the use of
When a malfunction is detected on one of the left front the parameter ρl that schedules the distribution of the left &

Preprint submitted to 2014 American Control Conference.

Received September 27, 2013.
 CONFIDENTIAL. Limited circulation. For review only.
Lateral Transfer Ratio
right suspensions on the four corners of the vehicle and tune
Vehicle with faulty damper
the suspension dampers smoothly. This is done thanks to the 1
LPV strategy
LPV frame work, from "soft" to "hard" to improve the car 0.5
performances according to the driving situation.

LTR
0

In this synthesis, the authors wish to stress that a very

−0.5
interesting innovation is the use of a partly fixed structure
controller with a parameter dependency (ρl ) on the con- −1

trol output matrix, combined with the scheduling of the

0 4 8 12
weighting functions by the use of the varying parameter ρl , t [s]

on the chassis displacement (zs , considered as a comfort

indicator) and the roll motion (θ, a road holding indicator). Fig. 3. Lateral load transfer
This allows to tune various actuators controllers, depending
on the driving situation, by a hierarchical activation to opti- Fig. 3 shows the lateral load transfer generated by the
mize their use (coordinate framework with smooth transition driving scenario; based on it, the scheduling parameter ρl is
between different performance objectives even if they are calculated.
contradictory). Suspension damper’s forces
1000
Front right ’Faulty’ damper
front rear damper
The LPV system (13) includes 3 scheduling parameters 750 Rear left
Rear left
and can be described as a polytopic system, i.e, a convex 500

combination of the systems defined at each vertex of a 250

polytope defined by the bounds of the varying parameter. The

ij
0

synthesis of the controller is made within the framework of Fs −250

the H∞ control of polytopic suspensions, (for more details, −500

see [18]). −750

−1000
Remark 2: All controllers presented along the paper are 0 1 2
t[s]
3 4 5 6

synthesized in the LPV/H∞ framework. This design is

achieved, thanks to the LMI-based H∞ resolution. Fig. 4. Suspension damper’s forces: the faulty and healthy dampers efforts

In Fig. 4, the 4 semi-active dampers efforts provided by

IV. S IMULATION R ESULTS
Time domain simulations are performed on the full
nonlinear vehicle model given in Section II-A. For sake of
completeness, the results of the proposed LPV/H∞ fault
tolerant control are denoted "LPV strategy" in red and
compared to the "vehicle with the damper failure" in blue.
To test the efficiency of the proposed LPV/H∞ FTC of
the designed fault tolerant LPV/H∞ controller are given.
It is clear that the failure occurs on the front rear damper
which can not provide more then 30% of the nominal
force of the healthy MR dampers. Also, it can be seen that
the dampers forces distribution is scheduled, following the
varying parameter ρl (generated by monitoring the lateral
transfer ratio). The suspensions forces provided on the right
side of the vehicle are larger than those on the right side, due
to the big load supported by their dampers. Moreover, the
force provided by the front right damper is greater than the
one provided by the rear right one, because it compensates
the load due to the front left damper.
Roll motion
2
vehicle with faulty damper

1.5
vehicle roll dynamics under semi-active damper malfunction,
LPV strategy
the following scenario is used: 1

0.5
θ

1) The vehicle runs at 80km/h in straight line on wet 0

road (µ = 0.5, where µ is a coefficient representing −0.5

the adherence to the road). −1

2) The front right damper of the vehicle is considered −1.5

0 1 2 3 4 5 6
faulty (a failure of 70% on the nominal behaviour of t [s]

the healthy dampers).

3) A 5cm bump on the left wheels (from t = 0.5s to Fig. 5. Roll motion of the vehicle θ
t = 1s),
4) A Another bump on the right wheels (from t = 3s to Fig. 5 represents one of the main results of the paper.
t = 4s), The roll dynamics are clearly attenuated by the proposed

Preprint submitted to 2014 American Control Conference.

Received September 27, 2013.
 CONFIDENTIAL. Limited circulation. For review only.
Front right chassis displacement Right rear chassis displacement
0.02
LPV/H∞ FTC strategy. This allows to maintain a good road 0.015
vehicle with faulty damper
0.015
holding and stability of the vehicle. vehicle with faulty damper
0.01
LPV strategy
0.01 0.005
LPV strategy
0.005 0
−3 Chassis displacement in CoG Chassis acceleration in CoG
x 10 1.5

Zsfr

zsrr
25 Vehicule with faulty dampers 0 −0.005
Vehicule with faulty damper
1
LPV strategy −0.005 −0.01
20
0.5
−0.01 −0.015
15
LPV strategy 0 −0.015 −0.02

ZsFull
Z Full

10
−0.5 −0.02
s

−0.025
0 1 2 3 4 5 6 0 1 2 3 4 5 6
5 t [s] t [s]
−1

0
−1.5
Fig. 12. Chassis displacement in Fig. 13. Chassis displacement in rear
−5 front right zsf r .
−2 right zsrr .
0 1 2 3 4 5 6 0 1 2 3 4 5 6
t [s] t [s]

Front left chassis displacement Left rear chassis displacement

0.06 0.05
Fig. 6. Chassis displacement in CoG Fig. 7. Chassis acceleration in CoG vehicle with faulty damper
vehicle with faulty damper
zs . z̈s . 0.05 0.04

0.04 LPV strategy

LPV strategy 0.03
0.03
0.02
0.02
From Fig. 6 and Fig. 7, it can be noticed that the developed

zsfl

zsrl
0.01
0.01
strategy in addition on enhancing vehicle roadholding, it im- 0
0
proves passengers comfort by reducing chassis acceleration −0.01
−0.01

z̈s and displacement zs while driving. −0.02 −0.02

−0.03 −0.03
0 1 2 3 4 5 6 0 1 2 3 4 5 6
t [s] t [s]
−3 Front right wheel displacement −3 Right rear wheel displacement
x 10 x 10
4 4
vehicle with faulty damper Vehicle with faulty damper
3 3 Fig. 14. Chassis displacement in Fig. 15. Chassis displacement in rear
LPV strategy
LPV strategy
front left zsf l . left zsrl .
2
2

1
1
rr
zusfr

0 Front right chassis acceleration Right rear chassis acceleration

us

6 6
Z

0
−1 vehicle with faulty damper
4 Vehicle with faulty damper 4
−1
−2
LPV strategy
2 LPV strategy
2
−2
−3

0 0
ddt

−3
zsrrddt
−4
0 1 2 3 4 5 6 0 1 2 3 4 5 6
sfr

t [s] t [s]
Z

−2 −2

Fig. 8. Wheel displacement in front Fig. 9. Wheel displacement in rear −4 −4

right zusf r . right zusrr . −6 −6

−8 −8
0 1 2 3 4 5 6 0 1 2 3 4 5 6
t [s] t [s]

Front left wheel displacement Left rear wheel displacement Fig. 16. Chassis acceleration in front Fig. 17. Chassis acceleration in rear
0.05 0.05
vehicle with faulty damper Vehicle with faulty damper
right z̈sf r . right z̈srr .
0.04 LPV strategy LPV strategy
0.04

0.03 0.03
Front left chassis acceleration Left rear chassis acceleration
4 2
vehicle with faulty damper
Zusrl
fl

0.02 1.5
us

0.02 3 vehicle faulty damper

Z

1
0.01 0.01 2
LPV strategy LPV strategy
0.5
1
Zsflddt

Zsrlddt

0 0
0
0
−0.01 −0.01 −0.5
0 1 2 3 4 5 6 0 1 2 3 4 5 6
t [s] t [s] −1
−1

−2
−1.5
Fig. 10. Wheel displacement in front Fig. 11. Wheel displacement in rear
left zusf l . left zusrl . −3
0 1 2 3 4 5 6
−2
0 1 2 3 4 5 6
t [s] t [s]

Fig. 18. Chassis acceleration in front Fig. 19. Chassis displacement in rear
In Fig. 8, 9, 10, 11, the four wheels bounce of the vehicle left z̈sf l . left z̈srl .
are shown. It can be seen also that the improvements brought
by the designed controller on the left side are better than on Figures from Fig. 12 to Fig. 19 show various comfort
the right side, due to the larger damping forces supplied on performances on each corner of the vehicle (chassis dis-
this side to handle the load transfer. placement, acceleration, resp). It is clearly noticed that the

Preprint submitted to 2014 American Control Conference.

Received September 27, 2013.
 CONFIDENTIAL. Limited circulation. For review only.

performance objectives are differently reached, depending on [11] S. Fergani, O. Sename, and L. Dugard, “Performances improvement
the suspension forces distribution and reconfiguration given through an LPV/H∞ control coordination strategy involving braking,
semi-active suspension and steering systems,” in Proceedings of the
by the proposed LPV/H∞ fault tolerant control. This allows 51th IEEE Conference on Decision and Control (CDC), Maui, Hawaii,
to handle the damper’s failure effect on the vehicle dynamics USA, December 2012.
in several driving situations. [12] ——, “A new LPV/H∞ global chassis control through load transfer
distribution and vehicle stability monitoring,” in Proceedings of IFAC
Joint conference 5th Symposium on System Structure and Control,
V. C ONCLUSION 11th Workshop on Time-Delay Systems 6th Workshop on Fractional
This paper has presented a new LPV/H∞ fault tolerant Differentiation and Its Applications, Grenoble, France, February 2013.
[13] J. Tudon-Martinez, S. Varrier, R. Morales-Menendez, R. Ramirez-
control strategy which handles vehicle roll dynamics under Mendoza, D. Koenig, J.-J. Martinez, and O. Sename, “Fault toler-
damper malfunction. It proposes a new structure of the ant control with additive compensation for faults in an automotive
controller, by making the corresponding LMIs orthogonal damper,” in Networking, Sensing and Control (ICNSC), 2013 10th
IEEE International Conference on, 2013, pp. 810–814.
with a parameter dependency on the controller matrix output. [14] C. Poussot-Vassal, O. Sename, L. Dugard, P. Gáspár, Z. Szabó,
The varying parameter used in the developed strategy is and J. Bokor, “Attitude and handling improvements through gain-
obtained by monitoring the lateral transfer ratio caused by scheduled suspensions and brakes control,” Control Engineering Prac-
tice, vol. 19, no. 3, pp. 252 – 263, 2011.
the roll bounce of the vehicle. This allows to online recon- [15] S. Guo, S. Yang, and C. Pan, “Dynamical Modeling of Magneto-
figure the provided suspensions forces in the four corners Rheological Damper Behaviors,” J. of Intell. Mater., Syst. and Struct.,
of the vehicle to reach the desired performance objective. vol. 17, pp. 3–14, 2006.
[16] A. L. Do, O. Sename, and L. Dugard, "LPV modeling and control
Simulations performed on a complex nonlinear model have of semi-active in automotive systems", 15th chapiter, pp381 − 411,
shown the efficiency of the proposed approach. in "Control of Linear Parameter Varying Systems with Applications",
The authors stress that using the LPV framework allows to C. Scherer and J. Mohammadpour, Eds. Springer,New-York,.
[17] S. Savaresi, C. Poussot-Vassal, C. Spelta, O. Sename, and L. Dugard,
simplify the implementation procedure. The next step of this Semi-Active Suspension Control for Vehicles. Elsevier - Butterworth
work is being started with the implementation of this strategy Heinemann, 2010.
on a test benchmark, available at Gipsa-lab in Grenoble, [18] C. Scherer, “Mixed H2 /H∞ control for time-varying and linear
parametrically-varying systems,” International Journal of Robust and
developed in collaboration with a high-technology start up Nonlinear Control, vol. 6, no. 9-10, pp. 929–952, november 1996.
"SOBEN". It consists of vehicle equipped with four semi-
active Electro-Rheological dampers. Different road profile
could be generated separately on each wheel and online
control can be implemented.
R EFERENCES
[1] D. Fischer and R. Isermann, “Mechatronic semi-active and active
vehicle suspensions,” Control Engineering Practice, vol. 12, pp. 1353–
1367, august 2003.
[2] D. Hrovat, “Survey of advanced suspension developments and related
optimal control application,” Automatica, vol. 33, no. 10, pp. 1781–
1817, october 1997.
[3] M. Ieluzzi, P. Turco, and M. Montiglio, “Development of a heavy
truck semi-active suspension control,” Control Engineering Practice,
vol. 14, pp. 305–312, june 2005.
[4] E. Abdellahi, D. Mehdi, and M. M. Saad, “On the design of active
suspension system by H∞ and mixed H2 /H∞ : An LMI approach,”
in Proceedings of the IEEE American Control Conference (ACC),
Chicago, Illinois, june 2000, pp. 4041–4045.
[5] P. Gáspár, I. Szaszi, and J. Bokor, “Iterative model-based mixed
H2 /H∞ control design,” in Proceedings of the UKACC International
Conference on Control, Swansea, United Kingdom, 1998, pp. 652–
657.
[6] J. Lu and M. DePoyster, “Multiobjective optimal suspension control to
achieve integrated ride and handling performance,” IEEE Transaction
on Control System Technology, vol. 10, no. 6, pp. 807–821, november
2002.
[7] R. Takahashi, J. Camino, D. Zampieri, and P. Peres, “A multiobjective
approach for H2 and H∞ active suspension control,” in Proceed-
ings of the IEEE American Control Conference (ACC), Philadelphia,
Pennsylvania, june 1998, pp. 48–52.
[8] J. Lozoya-Santos, R. Morales-Menendez, R. Ramirez-Mendoza,
J. Tudón-Martínez, O. Sename, and L. Dugard, “Magnetorheological
Damper - An Experimental Study,” J. of Intell. Mater. Syst. and Struct.,
vol. 23, pp. 1213–1232, 2012.
[9] X. Dong, M. Yu, C. Liao, and W. Chen, “Comparative Research on
Semi-Active Control Strategies for Magneto-rheological Suspension,”
Nonlinear Dynamics, vol. 59, pp. 433–453, 2010.
[10] K. Hudha, H. Jamaluddin, P. Samin, and R. Rahman, “Effects of
Control Techniques and Damper Constraint on the Performance of a
Semi-Active Magnetorheological Damper,” Int. J. Vehicle Autonomous
Systems, vol. 3, pp. 230–252, 2005.

Preprint submitted to 2014 American Control Conference.

Received September 27, 2013.
View publication stats