TSSD 2009
TSSD 2009
TSSD 2009
1861-5252/
Transactions on
Systems, Signals & Devices
Vol. 4, No. 1, pp.1-139
F
Stabilization Through Output Feedback
Control for Uncertain Switched Discrete
Time Systems
3
OO
E. Maherzi,1 M. Besbes,1 J. Bernussou,2 and R. Mhiri1,3
1
Réseaux et Machines Electriques (RME)
Institut Supérieur des Sciences Appliquées et de Technologie,
centre urbain Tunis-nord, B.P. 676, 1080 Tunis, Tunisia.
1. Introduction
The literature has shown a growing interest on switched systems since
switched control systems exist widely in engineering technology and so-
cial systems [7, 8, 20]. Switched systems are an important class of hybrid
systems defined by a set whose elements are dynamic continuous (or dis-
crete) time models and a commutation law which governs, in time, the
jumps between the elements, defining a non stationary dynamic system.
Many important progress and remarkable achievements have been made
2 E. Maherzi et al.
F
for systems with specific switching laws [19], stability analysis for systems
with arbitrary switching laws [13] and design of stabilizing switching laws
[14]. Many contributions to analyze stability of arbitrary switched sys-
tems use conservative arguments, the most pessimistic ones assuming
the existence of a common lyapunov function [17, 18]. Even if these con-
ditions are easily tractables, they can be used in a very few applications.
Some recent results are given in Daafouz et al., 2002 [2] where a suffi-
Varying systems.
OO
cient (but relatively non restrictive compared to the quadratic approach)
stability condition for discrete switched systems is provided using the
polyquadratic approach recently proposed by Daafouz and Bernussou
2001 [1] for stability analysis and stabilization control of Linear Time
2. Problem formulation
We consider an uncertain discrete time switched system where the so-
called subsystems are uncertain with a polytopic uncertainty, as roughly
illustrated in (Fig.1) where the uncertainty domains are polytopic shaped
F
with different number of vertices to cope with maximal generality. The
model can be stated as follows:
M
X Nl
X
x(k + 1) = ξl (k) αi(l) Ai(l) x(k) (1)
l=1 i(l)=1
Dl =
OO
where l is the switching index. M the number of uncertain systems
domain and Dl is the uncertainty domain for subsystem l defined by:
Aα Aα =
Nl
X
i(l)=1
αi(l) Ai(l) , αi(l) ≥ 0,
i(l)=1
αi(l) (k) = 1
1 if the state matrix is defined into Dl domain
ξl (k) =
0 else
AM 1 AM 2
AM NM SSM
F
[4] to propose the following criteria.
H=
XM
l=1
OO
trices G1 ...GM of appropriates dimensions solutions of the LM Is:
Gl + GTl − Si(l)l GTl ATi(l)l
Ai(l)l Gl Si(j)j
> 0, ∀(l, i(l), i(j), j) ∈ (e × el × ej × J)
Gl + GTl − Si(l)l GTl ATi(l)l
>0
Ai(l)l Gl Si(j)j
F
again
Nj
X
Gl + GTl − Sl GTl ATl
αi(j) >0
Al Gl Si(j)j
i(j)=1
replacing
PN j
"
OO
Gl + GTl − Sl
Al Gl
by Sj we obtain:
Gl + GTl − Sl
Al Gl
GTl ATl
Sj
>0
#
>0
F
y(k) = l=1 ξl (k)(Cl x(k)).
where [Al , Bl , Cl ] ∈ Dl , with:
X Nl Nl
X
Dl = αi(l) [Ai(l) , Bi(l) , Ci(l) ], αi(l) ≥ 0, αi(l) (k) = 1
i(l)=1 i(l)=1
OO
The stabilization problem of the switched system through static out-
put feedback consists in determining a control law of the form:
u(k) = Kl y(k)
x(k + 1) = (Alα + Blα Kl Clα )x(k)
x(k + 1) = Aelα x(k) such A
elα = (Alα + Blα Kl Clα )
F
Theorem 4 The system (4) can be locally polyquadratically stabilized
through output feedback if there exist M symmetric positive definite ma-
trices S1 ...SM , M matrices G1 ...GM and M matrices U1 ...UM of appro-
priates dimensions solutions of the LMIs:
Gl + GTl − Sl (Ai(l)l Gl + Bi(l)l Ul Ci(l)l )T
PM
H = l=1 N l
OO
Ai(l)l Gl + Bi(l)l Ul Ci(l)l
Vl Ci(l)l = Ci(l)l Gl
∀(l, i(l), j) ∈ (e × el × J)
e = {1...M }, el = {1...N l}
J function of l.
The output feedback gain is then defined by:
Kl = Ul Vl−1
Sj
> 0,
(6)
ε(k) = x(k) − x
b(k) (8)
F
Using (7), the dynamic estimation error is:
M
X
ε(k + 1) = ξl (k)(Ai(l) − Ll Cl ε(k)) (9)
l=1
OO elα ε(k), A
ε(k + 1) = A elα = Alα − Ll Clα
T
ATi(l)l Gl − Ci(l)l Fl Si(j)j
∀(l, i(l), i(j), j) ∈ (e × el × ej × J) (11)
PM
H = l=1 N l
e = {1...M }, el = {1...N l}, ej = {1...N j}
J function of l.
And the observation matrix is given by:
PR
Ll = (GTl )−1 FlT (12)
The observer (7) is obviously not feasible since for uncertain subsystems
the dynamic A matrix is not known. As an attempt to overcome this
difficulty one may think in choosing for the observer a dynamic matrix
Anom which is inside the uncertainty domains at each time (for instance,
the mean value with respect to the matrices associated with the vertices
of the different uncertainty polytopic domains). To close the loop and
realize a dynamic output control a natural way is then to compute with
the polyquadratic approach a state feedback as in [4] and apply this
feedback gain with the observer state, mimicking what is done with the
separation principle. Of course the closed loop stability is not implied
by such an approach and has to be checked using, for instance, the
polyquadratic analysis approach.
Stabilization of uncertain switched discrete time systems 9
F
Remark 2 We can note that for the local polyquadratic stability de-
scribed by (2), condition (11) becomes: System (10) is locally polyquadrat-
ically stable if there exist M symmetric positive definite matrices, S1 ...SM ,
M matrix F1 ...FM and M matrices G1 ...GM of appropriate dimension
solution of the LMIs:
"
T
OO
Gl + GTl − Sl (ATi(l)l Gl − Ci(l)l
T
Ai(l)l Gl − Ci(l)l Fl
∀(l, i(l), j) ∈ (e × el × J)
tation becomes:
PM
xb(k + 1) = l=1 ξl (k)(Abl x
b(k) + Bl u(k) + Ll (y(k) − yb(k)))
PM
yb(k) = l=1 ξl (k)Cl x
b(k) (14)
ε(k) = x(k) − xb(k)
PR
Using (14), the dynamic estimation error is:
M
X XNl
ε(k + 1) = ξl (k)( bl )x(k) + (A
αi(l) ((Ai(l) − A bl − Ll Cl ε(k))) (15)
l=1 i=1
u(k) = Kl x
b(k) (16)
then:
M
X XNl
x(k + 1) = ξl (k)( αi(l) (Ai(l) + Bl Kl )x(k) − Bl Kl ε(k)) (17)
l=1 i=1
10 E. Maherzi et al.
F
ε(k + 1) ε(k)
Ai(l)l + Bl Kl −Bl Kl (18)
with Φi(l)l = bl − Ll Cl
∆i(l)l A
b
and ∆i(l)l = Ai(l)l − Al
Gl1 Gl2
Using theorem 1 and replacing Gl by The system described
with "
(.)T Si(j)j
OO
>0
Gl3 Gl4
by (18), is polyquadratically stabilizable with the state feedback gains Kl
and the observation gains Ll if there exist H symmetric positive definite
matrices S11 ...SMN and M matrices G1 ...GM of appropriates dimensions
solutions of:
#
bl − GT Ll Cl
GTl1 (Ai(l)l + Bl Kl ) + GTl3 ∆i(l)l −GTl1 Bl Kl + GTl3 A
(.) = l3
bl − GT Ll Cl
GTl2 (Ai(l)l + Bl Kl ) + GTl4 ∆i(l)l −GTl2 Bl Kl + GTl4 A l4
and ∆i(l)l = Ai(l)l − A bl
(19)
7. Study cases
PR
We propose to design a switched observer for the example proposed
by Kothare, Balakrishnan and Morari 1996 [3], taken from a benchmark
where the physical system is constituted by two masses m1 , m2 vehicles
connected through a spring with stiffness k.
x1 x2
m k m
u
m1 m2
F
x4 (k)
y(k) = x2 (k)
1 0 0 0.1 0
0 1 0.1
0 0
A=
−0.1k/M1
, B=
0.1k/M1 0 1 0.1/M1
0.1k/M2
OO −0.1k/M2 1
C= 0 1 00
0 0
M 1 = m1 , M 2 = m2
M1 = m1 + m, M2 = m2
M 1 = m1 , M 2 = m2 + m
M1 = m1 + m, M2 = m2 + m
PR
4 3 2 1 0
F
gain K is computed for m = 1.1
K1 = −44.7395 36.0760 −14.5427 −5.3980
K2 = −105.9129 86.6016 −30.8068 −13.0689
OO
K3 = −40.2343 32.0069 −14.4219 −4.7267
Fig. 4. Trajectory of x2
Stabilization of uncertain switched discrete time systems 13
F
OO
PR
14 E. Maherzi et al.
8. Conclusion
The problem of designing a stabilizing output feedback is a complex
one for switched systems with uncertain sub-systems. It is known that
dynamic output control design is difficult when the uncertainties are
F
structured ones, which is the case for polytopic uncertainty. In this
paper we propose sufficient conditions using polyquadratic stability to
find robust output feedback controls for this kind of systems. In the case
of dynamic output control with observer the design still incorporates
a level of heuristic due to the choice of the dynamic A matrix for the
observer ; This point deserves further research, for instance, on the design
OO
of a non observer based switched dynamic output controller with the
probable drawback of the definition of bilinear matrix inequalities BMI
for the control design and the loss of state estimation which is a good
feature of the observer based output control since beside the control it
also provides some possibility for detection and supervision.
References
[1] J. Daafouz and J. Bernussou. Parameter dependent Lyapunov functions
for discrete time systems with time varying parametric uncertainties. Sys-
tems & Control Letters, 43:355-359, 2001.
[2] J. Daafouz, G. Millerioux and C. Iung. A poly-quadratic stability based
approach for linear switched systems. Int. Journal of Control, 75:1302-
1310, 2002.
[3] M. V. Kothare, Balakrishnan and Morari. Robust Constrained Model
Predictive Control using LMI. Automatica, 32:1361-1379, 1996.
[4] E.Maherzi, J.Bernussou and R.Mhiri. Stability and Stabilization for un-
PR
certain switched systems, a polyquadratic Lyapunov approach. Int. Jour-
nal on Sciences and Techniques of Automatic control , 1:61-74, 2007.
[5] J. Daafouz, Riedinger, and C. Iung. Stability analysis and control synthe-
sis for switched systems: A switched Lyapunov function approach. IEEE
Trans. on Automatic Control, 47(11):1883-1887, 2002.
[6] P. Perez and J. Geromel. H2 control for discrete-time systems optimality
and robustness. Automatica, 22(4):397-411, 1993.
[7] E. Skafidas, R. J. Evans, A. V. Savkin and I. R. Petersen. Stability results
for switched controller systems. Automatica, 35:553-564, 1999.
[8] M. Mariton. Jump linear systems in automatic control. New York: Marcel
Dekker, 1990.
[9] Z.Sun, S.S. Ge and T.H. Lee. Controllability and reachability criteria for
switched linear systems. Automatica, 38(5):775-786, 2002.
[10] G. Xie and L. Wang. Controllability and stabilizability of switched linear-
systems. Systems & Control Letters, 48(2):135-155, 2003.
Stabilization of uncertain switched discrete time systems 15
F
[13] D. Liberzon and A.S. Morse. Basic problem in stability and design of
switched systems. IEEE Control Systems Magazinz, 19(5):59-70, 1999.
[14] Z. G. Li, C.Y. Wen and Y.C. Soh. Stabilization of a class of switched
systems via designing switching laws. IEEE Trans. on Automatic Control,
46(4):665-670,2001.
[15] A. Giua, C. Seatzu and C. Van Der Mee. Optimal control of switched
autonomous linear Systems. Conf. on Decision and Control, 40th CDC,
OO
:2472-2477, Orlando-USA, 2001.
[16] A. Raymond and R.A. DeCarlo. Optimal control of switching systems.
Automatica, 41:11-27, 2005.
[17] R.N. Shorten and K. S. Narendra. On the stability and existence of
common Lyapunov functions for stable linear switching systems. Conf.
on Decision and Control, 37th CDC, :3723-3724, Tampa, 1998.
[18] M.A. Wicks, P.A. Peleties and R.A. DeCarlo. Switched controller syn-
thesis for the quadratic stabilization of a pair of unstable linear systems.
European Journal of Control, 4:140-147, 1998.
[19] M.S. Branicky. Multiple Lyapunov Functions and Other Analysis Tools
for Switched and Hybrid Systems. IEEE Trans. on Automatic Control,
43(4), 1998.
[20] X. Liu, X. Shen and Y. Zhang. Stability analysis of a class of hybrid
dynamic systems. Dynamics of Continuous, Discrete and Impulsive Sys-
tems, 8:359-373, 2001.
[21] G. Millerioux and J. Daafouz. Unknown input observers for switched
linear discrete time systems. American Control Conference, proceedings
PR
6(30):5802-5805, july 2004.
Biographies
Elyes Maherzi received the diploma of Master degree
from the “Université de Montpellier II”, France, in 2003.
Currently, he is a PHD student at the “Institut National
des Sciences Appliqués et de Technologie”, Tunisia. His
research interests include the robust control and switched
systems.
16 E. Maherzi et al.
F
2008. He is currently the head of Electrical Engineering
department at the “Ecole Supérieure de Technologie et
d’Informatique, Tunis”, Tunisia. His research interests
include the uncertain systems and multimodels.
OO
engineering from ENSEEIHT, Toulouse, France, in 1967.
He also received the “Docteur ingénieur” and the “Doc-
torat d’Etat” degrees both from the University Paul
Sabatier, Toulouse, in 1968 and 1970, respectively. He
is currently at the LAAS-CNRS as ”Directeur de recher-
che”. He is the author of more than one hundred research
papers and many books (Pergamon Press, North Holland,
Hermes). His research interests include large scale inter-
connected systems, optimal control, and robust control
theory applied to electrical energies network, spatial and aeronautics.