Commutation Relations and Stability of Switched Systems: Daniel Liberzon

COMMUTATION RELATIONS and
STABILITY of SWITCHED SYSTEMS
Daniel Liberzon
Coordinated Science Laboratory and

Dept. of Electrical & Computer Eng.,
Univ. of Illinois at Urbana-Champaign
COMMUTING STABLE MATRICES => GUES
(commuting Hurwitz matrices)
For subsystems – similarly

COMMUTING STABLE MATRICES => GUES
Alternative proof:
quadratic common Lyapunov function

[Narendra–Balakrishnan ’94]
..
.
is a common Lyapunov function

NILPOTENT LIE ALGEBRA => GUES
Lie algebra:
Lie bracket:
Nilpotent means sufficiently high-order Lie brackets are 0

For example:
(2nd-order nilpotent)
Recall: in commuting case
In 2nd-order nilpotent case
Hence still GUES [Gurvits ’95]

SOLVABLE LIE ALGEBRA => GUES
Larger class containing all nilpotent Lie algebras
Suff. high-order brackets with certain structure are 0
Lie’s Theorem: is solvable triangular form
Example:
exponentially fast
exp fast
quadratic common Lyap fcn diagonal

[Kutepov ’82, L–Hespanha–Morse ’99]
MORE GENERAL LIE ALGEBRAS
Levi decomposition:
radical (max solvable ideal)
• is compact (purely imaginary eigenvalues) GUES,

quadratic common Lyap fcn
• is not compact not enough info in Lie algebra:
There exists one set of stable generators for which

gives rise to a GUES switched system, and another
which gives an unstable one
[Agrachev–L ’01]
SUMMARY: LINEAR CASE
Lie algebra w.r.t.
Assuming GES of all modes, GUES is guaranteed for:
• commuting subsystems:
• 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)
Quadratic common Lyapunov function exists in all these cases

Further extension based only on Lie algebra is not possible
SWITCHED NONLINEAR SYSTEMS
Lie bracket of nonlinear vector fields:
Reduces to earlier notion for linear vector fields

(modulo the sign)
SWITCHED NONLINEAR SYSTEMS
• Commuting systems
GUAS
Can prove by trajectory analysis [Mancilla-Aguilar ’00]

or common Lyapunov function [Shim et al. ’98, Vu–L ’05]
• Linearization (Lyapunov’s indirect method)
• Global results beyond commuting case – ?

[Unsolved Problems in Math. Systems & Control Theory ’04]
SPECIAL CASE
globally asymptotically stable
Want to show: is GUAS
Will show: differential inclusion
is GAS
OPTIMAL CONTROL APPROACH
Associated control system:
where
(original switched system )
Worst-case control law [Pyatnitskiy, Rapoport, Boscain, Margaliot]:

fix and small enough
MAXIMUM PRINCIPLE
(along optimal trajectory)

Optimal control:
is linear in
(unless )
at most 1 switch
GAS
GENERAL CASE
Want: polynomial of degree
(proof – by induction on )
bang-bang with switches
GAS
THEOREM
Suppose:
• GAS, backward complete, analytic

• s.t.
and
Then differential inclusion is GAS,

and switched system is GUAS [Margaliot–L ’06]
Further work in [Sharon–Margaliot ’07]

REMARKS on LIE-ALGEBRAIC CRITERIA
• Checkable
• conditions
In terms of the original
• data
Independent of representation
• Not robust to small perturbations
In any neighborhood of any pair of matrices

there exists a pair of matrices generating the entire
Lie algebra [Agrachev–L ’01]
How to measure closeness to a “nice” Lie algebra?
ŁOJASIEWICZ INEQUALITY
f : R n ! R real analytic function

Z := f x 2 R n : f ( x) = 0 g – zero set of f
Then for everyK ½ R n compact, 9 C; ® > 0 s.t.
C jf ( x) j ¸ d ist ( x; Z ) ® 8x 2 K
Meaning: if jf ( x) j · " then d ist ( x; Z ) · ±:= ( C ") 1 =®
When f is a maximum of finitely many polynomials, explicit

bounds on the exponent ® can be obtained
ŁOJASIEWICZ APPLIED TO OUR SET – UP
f A 1 ; A 2 ; : : : ; A N g – finite set of Hurwitz matrices
Suppose k[A i ; A j ]k F · " 8 i; j
p
where kAk F := t rAA T is Frobenius norm
Let f ( A 1 ; A 2 ; : : : ; A N ) := m a x i;j k[A i ; A j ]k2F
This is max of N ( N ¡ 1 ) =2 polynomials of degree 4 in
coefficients of A i (or can use sum to get a single polynomial)
Let ± be distance from ( A 1 ; : : : ; A N ) to nearest N-tuple
(B1; : : : ; BN )
of pairwise commuting matrices
(Frobenius norm of difference between stacked matrices )
Łojasiewicz inequality gives C f ( A 1 ; A 2 ; : : : ; A N ) ¸ ±®
i.e., A i = B i + ¢ i with k¢ i k F · ( C " 2 ) 1 =®
Higher-order commutators (near nilpotent / solvable) – similar
PERTURBATION ANALYSIS
A i = B i + ¢ i ; k¢ i kF · ( C " 2 ) 1 =®
B i commute (or generate nilpotent or solvable Lie algebra)
) 9 common Lyapunov function V ( x) = x T P x :
P B i + B iT P = ¡ Q i < 0 8 i
For original matrices: P A i + A Ti P = ¡ Q i + P ¢ i + ¢ TP
i
¸ m in ( Q i )
which is still negative definite if k¢ k
i F < 2 ¸ m a x( P )
¸ (Q )
) stability of x_ = A ¾x is ensured if ( C " 2 ) 1 =® < 2 ¸m in ( Pi )
m ax
For commuting and solvable cases, specific known constructions of P
can be used to estimate the right-hand side [Baryshnikov–L, CDC ’13]
C ; ® depend on N (# of matrices) and on compact set where they live
d e g f = 4 ) ® can be explicitly estimated
estimating C is a bit more difficult [Ji–Kollar–Shiffman]
DISCRETE TIME: BOUNDS on COMMUTATORS
– Schur stable
GUES:
Let
GUES
Idea of proof: take , , consider
contraction induction small

basis

Commutation Relations and Stability of Switched Systems: Daniel Liberzon

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

Commutation Relations and Stability of Switched Systems: Daniel Liberzon

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Commutation Relations and Stability of Switched Systems: Daniel Liberzon

Uploaded by

Copyright:

Available Formats

COMMUTATION RELATIONS and

STABILITY of SWITCHED SYSTEMS

Coordinated Science Laboratory and

(commuting Hurwitz matrices)

For subsystems – similarly

quadratic common Lyapunov function

is a common Lyapunov function

Nilpotent means sufficiently high-order Lie brackets are 0

In 2nd-order nilpotent case

Hence still GUES [Gurvits ’95]

quadratic common Lyap fcn diagonal

radical (max solvable ideal)

• is compact (purely imaginary eigenvalues) GUES,

There exists one set of stable generators for which

Assuming GES of all modes, GUES is guaranteed for:

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

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

• solvable + compact (purely imaginary eigenvalues)

Quadratic common Lyapunov function exists in all these cases

Lie bracket of nonlinear vector fields:

Reduces to earlier notion for linear vector fields

Can prove by trajectory analysis [Mancilla-Aguilar ’00]

• Linearization (Lyapunov’s indirect method)

• Global results beyond commuting case – ?

Want to show: is GUAS

Will show: differential inclusion

(original switched system )

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

(along optimal trajectory)

Want: polynomial of degree

bang-bang with switches

• GAS, backward complete, analytic

Then differential inclusion is GAS,

Further work in [Sharon–Margaliot ’07]

• Not robust to small perturbations

In any neighborhood of any pair of matrices

f : R n ! R real analytic function

Meaning: if jf ( x) j · " then d ist ( x; Z ) · ±:= ( C ") 1 =®

When f is a maximum of finitely many polynomials, explicit

Idea of proof: take , , consider

contraction induction small

You might also like