STABILITY OF SWITCHED SYSTEMS

Daniel Liberzon

Coordinated Science Laboratory and

Dept. of Electrical & Computer Eng.,
Univ. of Illinois at Urbana-Champaign

1
 SWITCHED vs. HYBRID SYSTEMS
Switched system:

• is a family of systems
• is a switching signal

Switching:
• state-dependent or time-dependent
• autonomous or controlled

Details of discrete behavior are “abstracted away”

Hybrid systems give rise to classes of switching signals

Properties of the continuous state: stability

2
 STABILITY ISSUE

unstable

Asymptotic stability of each subsystem is

not sufficient for stability

3
 TWO BASIC PROBLEMS

• Stability for arbitrary switching

• Stability for constrained switching

4
 TWO BASIC PROBLEMS

• Stability for arbitrary switching

• Stability for constrained switching

5
GUAS and COMMON LYAPUNOV FUNCTIONS

GUAS:

GUES:

is GUAS if (and only if) s.t.

where is positive definite

quadratic is GUES
6
 COMMUTING STABLE MATRICES => GUES
P  {1, 2}, A1 A2  A2 A1

 1  2  1  2 …
s1 t1 s2 t2 t

A2 tk A1s k … A2 t1 A1s1
x(t )  e e e e x(0)
A2 (tk  ...  t1) A1 ( s k  ...  s1)
e e x(0)  0

quadratic common Lyap fcn [Narendra & Balakrishnan ’94]:

A1T P1  P1 A1  I

A2T P2  P2 A2  P1
7
 COMMUTATION RELATIONS and STABILITY
Lie algebra:

Lie bracket:

GUES and quadratic common Lyap fcn guaranteed for:

• nilpotent Lie algebras (suff. high-order Lie brackets are 0)

e.g.

• solvable Lie algebras (triangular up to coord. transf.)

• solvable + compact (purely imaginary eigenvalues)

Further extension based only on Lie algebra is not possible

[Agrachev & L ’01] 8
 SWITCHED NONLINEAR SYSTEMS

• Commuting systems

=> GUAS [Mancilla-Aguilar,

Shim et al., Vu & L]

• Linearization (Lyapunov’s indirect method)

• Global results beyond commuting case – ???

[Unsolved Problems in Math. Systems and Control Theory]
9
 SPECIAL CASE
globally asymptotically stable

Want to show: is GUAS

Will show: differential inclusion

is GAS
10
 OPTIMAL CONTROL APPROACH
Associated control system:

where

(original switched system )

Worst-case control law [Pyatnitskiy, Rapoport, Boscain, Margaliot]:

fix and small enough

11
 MAXIMUM PRINCIPLE

(along optimal trajectory)

Optimal control:

is linear in
(unless )

at most 1 switch

GAS 12
SYSTEMS with SPECIAL STRUCTURE

• Triangular systems

• Feedback systems

• passivity conditions

• small-gain conditions

• 2-D systems

14
 TRIANGULAR SYSTEMS
For linear systems, triangular form GUES

exponentially fast

exp fast

0
quadratic common Lyap fcn diagonal

For nonlinear systems, not true in general

Need to know (ISS)
[Angeli & L ’00]
15
 FEEDBACK SYSTEMS: ABSOLUTE STABILITY

controllable

Circle criterion: quadratic common Lyapunov function

is strictly positive real (SPR):

For this reduces to SPR (passivity)

Popov criterion not suitable: depends on

16
FEEDBACK SYSTEMS: SMALL-GAIN THEOREM
controllable

Small-gain theorem:
quadratic common Lyapunov function

17
 TWO-DIMENSIONAL SYSTEMS
Necessary and sufficient conditions for GUES
known since 1970s

worst-case
switching

quadratic common Lyap fcn <=>

convex combinations of Hurwitz

18
 WEAK LYAPUNOV FUNCTION

Barbashin-Krasovskii-LaSalle theorem:

is GAS if s.t.

• (weak Lyapunov function)

• is not identically zero along any nonzero solution
(observability with respect to )

Example:

=> GAS
observable
19
COMMON WEAK LYAPUNOV FUNCTION

Theorem: is GAS if

• .

• observable for each

• s.t. there are infinitely many

switching intervals of length

To extend this to nonlinear switched systems and

nonquadratic common weak Lyapunov functions,
we need a suitable nonlinear observability notion
20
 NONLINEAR VERSION
Theorem: is GAS if s.t.

• Each system

is small-time norm-observable:

pos. def. incr. :

• s.t. there are infinitely many

switching intervals of length
21
 TWO BASIC PROBLEMS

• Stability for arbitrary switching

• Stability for constrained switching

22
 MULTIPLE LYAPUNOV FUNCTIONS
x  f1 ( x) , x  f 2 ( x)  GAS
V1 , V2  respective Lyapunov functions

V (t ) (t )

x  f (x)
=>
is GAS

t
 1  2  1  2
Useful for analysis of state-dependent switching
23
MULTIPLE LYAPUNOV FUNCTIONS

decreasing sequence
V (t ) (t )

=> GAS

decreasing sequence

t
 1  2  1  2
[DeCarlo, Branicky]
24
 DWELL TIME
The switching times t1 , t 2 , ... satisfy ti 1  ti   D
x  f1 ( x) , x  f 2 ( x)  GES
dwell time
V1 , V2  respective Lyapunov functions

25
 DWELL TIME
The switching times t1 , t 2 , ... satisfy ti 1  ti   D
x  f1 ( x) , x  f 2 ( x)  GES
V1
a1 | x |  V1 ( x)  b1 | x | ,
2 2
f1 ( x)   1V1 ( x)
x
V2
a2 | x | 2  V2 ( x)  b2 | x | 2 , f 2 ( x)   2V2 ( x)
x

Need: V1 (t 2 )  V1 (t0 )

 1  2  1
t
t0 t1 t2
26
 DWELL TIME
The switching times t1 , t 2 , ... satisfy ti 1  ti   D
x  f1 ( x) , x  f 2 ( x)  GES
V1
a1 | x |  V1 ( x)  b1 | x | ,
2 2
f1 ( x)   1V1 ( x)
x
V2
a2 | x | 2  V2 ( x)  b2 | x | 2 , f 2 ( x)   2V2 ( x)
x

Need: V1 (t 2 )  V1 (t0 )
must be 1

b1 b2 (1  2 ) D
V1 (t 2 )  e V1 (t 0 )
a2 a
1

27
 AVERAGE DWELL TIME
T t
N (T , t )  N 0 
 AD

# of switches on (t , T ) average dwell time

N 0  0  no switching: cannot switch if T  t   AD
N 0  1  dwell time: cannot switch twice if T  t   AD

28
 AVERAGE DWELL TIME
T t
N (T , t )  N 0 
 AD

Theorem: [Hespanha]

1 (| x | )  V p ( x)   2 (| x | ) x  f (x)
Vp
f ( x)   V p ( x) => is GAS
x p
log 
V p ( x )   Vq ( x ),  p, q if  AD 


GAS is uniform over in this class

Useful for analysis of hysteresis-based switching logics

29
 MULTIPLE WEAK LYAPUNOV FUNCTIONS
Theorem: is GAS if

• .

• observable for each

• s.t. there are infinitely many

– milder than a.d.t.
switching intervals of length
• For every pair of switching times
s.t.
have

Extends to nonlinear switched systems as before

30
 APPLICATION: FEEDBACK SYSTEMS
observable
positive real

Weak Lyapunov functions:

Theorem: switched system is GAS if

• s.t. infinitely many switching intervals of length
• For every pair of switching times at
which we have
(e.g., switch on levels of equal “potential energy”)
31
RELATED TOPICS NOT COVERED

• Computational aspects (LMIs, Tempo & L)

• Formal methods (work with Mitra & Lynch)

• Stochastic stability (Chatterjee & L)

• Switched systems with external signals

• Applications to switching control design

32
 REFERENCES
Lie-algebras and nonlinear switched systems: [Margaliot & L ’04]
Nonlinear observability, LaSalle: [Hespanha, L, Angeli & Sontag ’03]
(http://decision.csl.uiuc.edu/~liberzon)

33