Professional Documents
Culture Documents
Lyapunov Stability The DTLTI system (1) is said to be stable in the sense of Lyapunov if
there exists a Lyapunov function V (x) such that
Lyapunov Stability Test Given the DTLTI system (1) find if there exists a matrix P ∈ S n
such that
a) V (x) := xT P x > 0, ∀x 6= Θ
b) V (x(k + 1)) − V (x(k)) < 0, ∀x(k + 1) = Ax(k), x(k) 6= Θ.
Remarks
a) ⇒ V (x) := xT P x > 0, ∀x 6= Θ ⇒ P Â Θ.
b) ⇒ V (x(k + 1)) = x(k + 1)T P x(k + 1) = x(k)T AT P Ax(k).
V (x(k + 1)) − V (x(k)) = x(k)T AT P Ax(k) − x(k)T P x(k),
= x(k)T AT P A − P x(k) < 0, ∀x(k) 6= Θ
¡ ¢
⇒ AT P A − P ≺ Θ.
Lyapunov Stability Test Given the DTLTI system (1) find if there exists a matrix P ∈ S n
such that the LMI
P Â Θ, AT P A − P ≺ Θ,
is feasible.
is feasible.
Robust Stability The DTLTI uncertain system (2) is said to be robustly stable if (2) is
asymptotically stable for all A ∈ A.
Problem The set of all (discrete-time) stable matrices is not a convex set.
Proof By counter-example:
1/2 2
A1 = , A 2 = AT
1, λi (A1 ) = λi (A2 ) = 1/2 < 1, i = 1, 2,
0 1/2
P Â Θ, AT
i P Ai − P ≺ Θ, ∀i = 1, . . . , N,
is feasible.
c) There exists P ∈ Sn such that the LMI
T
P Ai P
 Θ, ∀i = 1, . . . , N,
P Ai P
is feasible.
PN
Proof For any Ā ∈ A there exists a ξ̄ ∈ Ξ such that Ā = ξ̄i Ai . Since
i=1
ξ̄ ∈ Ξ ⇒ ξ̄i ≥ 0, multiply each inequality in the QS condition c) by ξ̄i to obtain
ξ̄i AT
¡ ¢
ξ̄i P i P
¡ ¢ Â Θ, ∀i = 1, . . . , N.
P ξ̄i Ai ξ̄i P
which implies that Ā is stable. The proof is completed by noticing that the above holds for all
A ∈ A.
Remarks
• Quadratic stability ⇒ robust stability ⇒ vertex stability.
• Quadratic stability : robust stability : vertex stability.
so that A ∈ Sn is stable iff f (A) < 1. But as seen before (Background Material - part III),
I A
λmax (AA) < 1 ⇔ AA ≺ 1 ⇔ Â Θ ⇔ stability!
A I
System Interpretation
The (robust) stability of a symmetric system can be always proven with P = I.
due to convexity.
Proof It is always true that (γ 1/2 F − γ −1/2 ∆H)T (γ 1/2 F − γ −1/2 ∆H) º Θ hence
γ −1 H T ∆T ∆H + γF T F º H T ∆T F + F T ∆H.
which provides
H T ∆T F + F T ∆H ¹ γ −1 H T ∆T ∆H + γF T F ¹ γH T H + γF T F.
Proof (robust stability) Using the previous lemma with H = C and F = (P B) T we get
Θ Â AT P + P A + γC T C + γP BB T P,
 AT P + P A + C T ∆T B T P + P B∆C,
= A(∆)T P + P A(∆)
which holds for all ∆ ∈ D. In addition, the fact that P Â Θ guarantees robust stability.