Multiple Lyapunov Functions and Other Analysis Tools for Switched and Hybrid Systems Michael Branicky Dept. of Electrical Eng. and Computer Sc. Case Western Reserve University Multiple Lyapunov Functions and Other Analysis Tools for Switched and Hybrid Systems – p. 1/16 Contents • Systems under consideration • Lyapunov-like functions • Stability using multiple Lyapunov functions • Limit cycles in continuous switched systems Multiple Lyapunov Functions and Other Analysis Tools for Switched and Hybrid Systems – p. 2/16 Switched systems Systems of the form ẋ(t) = fi (x(t)) where x ∈ Rn is the state vector, t ∈ R+ ∪ {0} is time and i ∈ Q = {1, 2, . . . , N } is an index. • fi is globally Lipschitz continuous for all i • i is chosen such that the overall system is non-Zeno Multiple Lyapunov Functions and Other Analysis Tools for Switched and Hybrid Systems – p. 3/16 Discrete-time switched systems Systems of the form xk+1 = fi (xk ) where k ∈ Z+ ∪ {0} is the discrete-time sample number and i ∈ {1, 2, . . . , N } and everything else is as before. • fi is globally Lipschitz continuous for all i Multiple Lyapunov Functions and Other Analysis Tools for Switched and Hybrid Systems – p. 4/16 Continuous switched systems The system ẋ(t) = fi (x(t)) is continuous at the switching times. That is, if at times tj , j = 1, 2, 3, . . . the vector field fij−1 switches to fij , then fij−1 (x(tj−1 )) = fij (x(tj )) Multiple Lyapunov Functions and Other Analysis Tools for Switched and Hybrid Systems – p. 5/16 Hybrid systems Systems of the form ẋ(t) = f (x(t), q(t)) q(t) = ν(x(t), q(t− )) where x ∈ Rn is the state vector, t ∈ R+ is time and q ∈ Q = {1, 2, . . . , N } is the discrete state. We can also imagine an external input u ∈ Rm , either switching or continuous; this would make the system a controlled hybrid system. Multiple Lyapunov Functions and Other Analysis Tools for Switched and Hybrid Systems – p. 6/16 Switching sequences S = x0 ; {(ij , tj )}N j=0 where x0 ∈ Rn is the initial state and (ij , tj ) are pairs specifying ẋ(t) = fij (x(t)) for tj ≤ t < tj+1 In the following, S denotes the set of all switching sequences associated with the system. Multiple Lyapunov Functions and Other Analysis Tools for Switched and Hybrid Systems – p. 7/16 Switching sequences • Projections of switching sequences: π1 (S) = x0 ; {(ij )}N j=0 , • π2 (S) = x0 ; {(tj )}N j=0 Endpoints of time intervals where system i is "switched in": S|i = x0 ; {(i, tj )}, j ∈ {k ∈ Z+ ∪ {0}|fi is active} Multiple Lyapunov Functions and Other Analysis Tools for Switched and Hybrid Systems – p. 8/16 Switching sequences • Interval completion of a strictly increasing Nj sequence T = {tj }j=0 : I(T ) = Nj [ [t2j , t2j+1 ] j=0 • Even sequence of T : E(T ) = t0 , t2 , t4 , . . . Multiple Lyapunov Functions and Other Analysis Tools for Switched and Hybrid Systems – p. 9/16 Lyapunov-like functions Definition 2.2 Given a strictly increasing time sequence T , a continuous pos. def. function V (x(t)) with V (0) = 0 and continuous partial derivatives is Lyapunov-like for a trajectory x(t) over T if • V̇ (x(t)) ≤ 0 for ∀t ∈ I(T ), x 6= 0 • V (x(t)) is monotonically non-increasing on E(T ) Multiple Lyapunov Functions and Other Analysis Tools for Switched and Hybrid Systems – p. 10/16 Lyapunov-like stability for switched systems Theorem 2.3 Suppose we have candidate Lyapunov fcts. Vi for corresponding vector fields ẋ = fi (x), fi (0) = 0 for i ∈ Q. If, for all S ∈ S and all i ∈ Q, Vi is Lyapunov-like for xS (t) (the trajectory associated with the switching seq. S) over S|i, then the system is stable in the sense of Lyapunov. Multiple Lyapunov Functions and Other Analysis Tools for Switched and Hybrid Systems – p. 11/16 Lyapunov-like stability for switched systems • The theorem holds for time-varying fi as well • It does not hold for infinite numbers of vector fields • For restricted classes of vector fields, one can use this to design (feedback laws and) switching sequences that guarantee stability Multiple Lyapunov Functions and Other Analysis Tools for Switched and Hybrid Systems – p. 12/16 A switched system requiring MLF Multiple Lyapunov Functions and Other Analysis Tools for Switched and Hybrid Systems – p. 13/16 Lyapunov-like stability for "smoothly" switched systems Theorem 2.7 Suppose we have a continuous candidate Lyapunov-like function Vλ : Rn × K → R+ ∪ {0} for corresponding vector fields ẋ = f (x, λ), f (0, λ) = 0, λ ∈ K (a compact set). If, for all S ∈ S and all λ ∈ K, Vλ satisfies Vλj+1 (x(tj+1 )) < Vλj (x(tj )) ∀tj , tj+1 ∈ T for xS (t) over S|λ, then the system is stable in the sense of Lyapunov. Multiple Lyapunov Functions and Other Analysis Tools for Switched and Hybrid Systems – p. 14/16 Limit cycles in continuous switched systems Theorem 4.1 Suppose D is a simply connected domain in R2 and f : R2 → R2 is a Lipschitz cont. ∂f1 ∂f2 + vector field such that ∇f (x) = ∂x ∂x2 , which 1 exists almost everywhere, is not zero almost everywhere over any subregion in D. Then D contains no closed trajectories of x = f (x). Multiple Lyapunov Functions and Other Analysis Tools for Switched and Hybrid Systems – p. 15/16 Related literature • Johansson and Rantzer, "Computation of piecewise quadratic Lyapunov functions for hybrid systems," IEEE TAC 1998 • Pogromsky, Jirstrand and Spangeus, "On stability and passivity of a class of hybrid systems," CDC 1998 • Mignone, Ferrari-Trecate and Morari, "Stability and stabilization of piecewise affine and hybrid systems: an LMI approach," CDC 2000 • Rubensson and Lennartson, "Stability and robustness of hybrid systems using discrete-time Lyapunov Multiple Lyapunov Functions and Other Analysis Tools for Switched and Hybrid Systems – p. 16/16