Linear Control Systems Lecture # 5 Stability and Matlab – p. 1/2 Internal Stability Internal stability deals with the boundedness and asymptotic behavior (as t → ∞) of the solutions of ẋ = Ax The solution of ẋ = Ax is given by x(t) = eAt x(0) Definition: Thesystem ẋ = Ax is stable if eAt ≤ γ, ∀ t ≥ 0, γ > 0 and asymptotically (or exponentially) stable if At e ≤ γe−λt , ∀ t ≥ 0, γ > 0, λ > 0 – p. 2/2 If A is diagonalizable eAt = n X eλi t vi wiT i=1 The behavior depends on Re [λi ] λ t Re [λi ] < 0 ⇒ e i ≤ e−λt , λ > 0 λ t Re [λi ] = 0 ⇒ e i = 1 Re [λi ] > 0 ⇒ eλi t is unbounded – p. 3/2 In general eAt = mi r X X i=1 k=1 Wik tk−1 (k − 1)! eλi t Re [λi ] > 0 ⇒ tk−1 eλi t is unbounded k−1 λ t Re [λi ] < 0 ⇒ t e i ≤ γe−λt , γ > 0, λ > 0 Re [λi ] = 0 and k = 1 ⇒ eλi t is bounded Re [λi ] = 0 and k ≥ 2 ⇒ tk−1 eλi t is unbounded – p. 4/2 When will the dimension of the Jordan block Ji be higher than one? (A − λi I)vi = 0 Let qi be the algebraic multiplicity of λi . If qi = nullity(A − λi I) = n − rank(A − λi I) then there are qi linearly independent eigenvectors associated with the eigenvalue λi If this is true for every eigenvalue that has multiplicity higher than one, then we can find n linearly independent eigenvectors and A is diagonalizable – p. 5/2 If nullity(A − λi I) = n − rank(A − λi I) < qi for any eigenvalue with multiplicity higher than one, then A is not diagonalizable nullity(A − λi I) is equivalent to the geometric multiplicity of the eigenvalue λi . We need qi − nullity(A − λi I) number of generalized eigenvectors – p. 6/2 Theorem: The system ẋ = Ax is Stable if and only if Re [λi ] ≤ 0, for i = 1, 2, . . . , n and for every eigenvalue with Re [λi ] = 0 and algebraic multiplicity qi ≥ 2, rank(A − λi I) = n − qi Asymptotically (or exponentially) stable if Re [λi ] < 0, for i = 1, 2, . . . , n – p. 7/2 Example: Study the stability of ẋ = Ax, where −1 0 3 1 0 −2 1 0 A= 0 0 −3 1 0 0 0 −3 Eigenvalues : − 1, −2, −3, −3 The system is asymptotically stable – p. 8/2 Example: Study the stability of ẋ −1 0 0 1 A= 0 0 0 0 Eigenvalues : = Ax, where 3 1 1 0 0 0 0 0 − 1, 1, 0, 0 The system is unstable – p. 9/2 Example: Consider the series and parallel connections of two identical systems, each represented by " # " # h i 0 1 0 ẋ = x+ u, y = 1 0 x −1 0 1 or H(s) = As = 0 −1 0 1 1 0 0 0 0 0 0 −1 0 0 1 0 1 s2 + 1 , A = p Eigenvalues : 0 −1 0 0 ± j, ±j 1 0 0 0 0 0 0 −1 0 0 1 0 – p. 10/2 Series Case: λ1 = j, q1 = 2, n − q1 = 2 As − λ1 I = −j 1 0 0 −1 −j 0 0 0 0 −j 1 1 0 −1 −j → rank(As − λ1 I) = 3 0 1 0 0 0 −j 0 0 0 0 0 1 1 0 0 −j The system is unstable – p. 11/2 Parallel Case: λ1 = j, q1 = 2, n − q1 = 2 Ap − λ1 I = −j 1 0 0 −1 −j 0 0 0 0 −j 1 0 0 −1 −j → rank(Ap − λ1 I) = 2 0 1 0 0 0 −j 0 0 0 0 0 1 0 0 0 −j The system is stable – p. 12/2 What is the effect of state transformations on stability? A → x = P z → P −1 AP Av = λv P −1 Av = λP −1 v (P −1 AP )(P −1 v) = λP −1 v A and P −1 AP have the same eigenvalues. If vi is an eigenvector of A, then P −1 vi is an eigenvector of P −1 AP rank P −1 AP − λI = rank P −1 (A − λI)P = rank (A − λI) Internal stability is invariant under state transformations – p. 13/2 Input-Output Stability Definition: The linear system ŷ(s) = H(s)û(s) is Bounded-Input Bounded-Output (BIBO) stable if for every bounded input u(t), the output y(t) is bounded Equivalently, for every ku > 0 there is ky > 0 such that ku(t)k ≤ ku , ∀ t ≥ 0 ⇒ ky(t)k ≤ ky , ∀ t ≥ 0 By taking the inverse Laplace transform y(t) = Z t H(t − τ )u(τ ) dτ 0 where H(t) = L−1 {H(s)} is the impulse response matrix – p. 14/2 Theorem: The system ŷ(s) = H(s)û(s) is BIBO stable if and only if Z ∞ kH(t)k dt < ∞ 0 Proof of sufficiency: ky(t)k = ≤ ≤ ≤ Z t H(t − τ )u(τ ) dτ 0 Z t kH(t − τ )k ku(τ )k dτ 0 Z ∞ kH(t − τ )k dτ ku 0 Z ∞ def kH(σ)k dσ ku = ky 0 – p. 15/2 Proof on Necessity: Use a contradiction argument (in the SISO case). Given ku > 0, suppose there is ky > 0 such that |u(t)| ≤ ku , ∀ t ≥ 0 ⇒ |y(t)| ≤ ky , ∀ t ≥ 0 R∞ but 0 |h(t)| dt is not finite There is t1 (dependent on ky /ku ) such that Z t1 |h(t1 − τ )| dτ > 0 ky ku – p. 16/2 Let ku , u(t) = 0 −k u when h(t1 − t) > 0 when h(t1 − t) = 0 when h(t1 − t) < 0 |u(t)| ≤ ku , for 0 ≤ t ≤ t1 Z t1 Z t1 y(t1 ) = h(t1 −τ )u(τ ) dτ = ku |h(t1 −τ )| dτ > ky 0 0 Contradcition – p. 17/2 Example: Time delay element y(t) = u(t − T ) H(s) = e−sT h(t) = L−1 {H(s)} = δ(t − T ) |u(t)| ≤ ku ⇒ |y(t)| ≤ ku Or Z ∞ δ(t − T ) dt = 1 0 – p. 18/2 When H(s) = C(sI − A)−1 B + D the elements hij (s) of H(s) are proper rational functions of s nij (s) hij (s) = dij (s) where nij (s) are dij (s) are polynomials with deg(nij ) ≤ deg(dij ) – p. 19/2 Since −1 (sI − A) = 1 det(sI − A) Adjoint(sI − A) the poles of hij (s) are roots of det(sI − A) = 0 that is, eigenvalues of A Not all eigenvalues of A will appear as poles of some elements of H(s) because some eigenvalues could be cancelled – p. 20/2 Given a strictly proper rational function H(s), let h(t) = L−1 {H(s)}. When will Z ∞ |h(t)| dt < ∞ 0 H(s) can be expressed as the sum of terms of the form K (s − p)α L−1 K (s − p)α =K tα−1 (α − 1)! ept Theorem: H(s) is BIBO stable if and only if all poles of every element of H(s) have negative real parts – p. 21/2 What is the relationship between asymptotic and BIBO stability? The system ẋ = Ax is asymptotically stable if all the eigenvalues of A have negative real parts. The system H(s) = C(sI − A)−1 B + D is BIBO stable if all poles of all elements of H(s) have negative real parts The poles of H(s) are eigenvalues of A Asymptotic stability ⇒ BIBO stability What about the opposite implication? Some eigenvalues of A may not appear as poles of H(s). If such eigenvalues have nonnegative real parts, then we could have a situation where the system is BIBO stable but not asymptotically stable – p. 22/2 Example: Suppose A is diagonalizable eAt = n X eλi t vi wiT i=1 n At X =L e = −1 (sI − A) i=1 −1 H(s) = C(sI − A) B+D = n X i=1 1 s − λi 1 s − λi vi wiT Cvi wiT B + D If Cvi = 0 or wiT B = 0, the eigenvalue λi cancels out of H(s) – p. 23/2 {A, B, C, D} → x = P z → {Λ, P −1 B, CP, D} ẋ = Ax + Bu → ż = Λz + (P −1 B := B̃)u y = Cx + Du → y = (CP := C̃)z + Du Λ= λ1 ... w1T .. −1 =: . , P = [v1 , · · · , vn ], P T λn wn CP = [Cv1 , · · · , cvn ] =: [c̃1 , · · · , c̃n ] w1T B b̃1 .. .. −1 P B = . =: . TB wn b̃n – p. 24/2 Uncontrollable Mode and Unobservable Mode −1 H(s) = C(sI − A) B= n X i=1 1 s − λi c̃i b̃i – p. 25/2 Example: Suppose we want to stabilize an unstable system described by Gp (s) = 1 s−1 Consider a cascade compensator Gc (s) = so that Gp (s)Gc (s) = s−1 s+1 1 s−1 · s−1 s+1 = 1 s+1 The system is BIBO stable. Is it asymptotically stable? – p. 26/2 Find a state model of the system u - s−1 s+1 Gp (s) = Gc (s) = 1 s−1 s−1 s+1 v - 1 s−1 y - → ẋ1 = x1 + v, y = x1 → ẋ2 = −x2 + u, v = −2x2 + u ẋ1 = x1 − 2x2 + u " # " # 1 −2 1 ẋ = x+ u 0 −1 1 – p. 27/2 A= " 1 −2 0 −1 # Eigenvalues : 1, −1 The system is not asymptotically stable. The eigenvalue +1 is called a hidden mode. Unstable hidden modes are not acceptable because they can be excited by initial conditions or disturbances For example x1 (0) = α, x2 (0) = 0, u(t) ≡ 0 ⇒ x2 (t) ≡ 0 ⇒ x1 (t) = αet – p. 28/2