On common linear/quadratic Lyapunov functions for switched linear systems

Similar documents
On the game-theoretic value of a linear transformation on a symmetric cone

ON POSITIVE SEMIDEFINITE PRESERVING STEIN TRANSFORMATION

Some characterizations of cone preserving Z-transformations

On the game-theoretic value of a linear transformation relative to a self-dual cone

M. Seetharama Gowda. J. Tao. G. Ravindran

Key words. Complementarity set, Lyapunov rank, Bishop-Phelps cone, Irreducible cone

Row and Column Representatives in Qualitative Analysis of Arbitrary Switching Positive Systems

Weakly homogeneous variational inequalities and solvability of nonlinear equations over cones

Z-transformations on proper and symmetric cones

Complementarity properties of Peirce-diagonalizable linear transformations on Euclidean Jordan algebras

7.1 Linear Systems Stability Consider the Continuous-Time (CT) Linear Time-Invariant (LTI) system

Analysis of Control Systems on Symmetric Cones

Weakly homogeneous variational inequalities and solvability of nonlinear equations over cones

Control System Analysis on Symmetric Cones

On the stability of switched positive linear systems

Permutation invariant proper polyhedral cones and their Lyapunov rank

Positive and doubly stochastic maps, and majorization in Euclidean Jordan algebras

The eigenvalue map and spectral sets

Switched systems: stability

M. Seetharama Gowda a, J. Tao b & Roman Sznajder c a Department of Mathematics and Statistics, University of Maryland

Complementarity Problems And Positive Definite Matrices

Hybrid Systems Course Lyapunov stability

An alternative proof of the Barker, Berman, Plemmons (BBP) result on diagonal stability and extensions - Corrected Version

Weakly homogeneous variational inequalities and solvability of nonlinear equations over cones

Interval Matrix Systems and Invariance of Non-Symmetrical Contractive Sets

Strict diagonal dominance and a Geršgorin type theorem in Euclidean

Modeling & Control of Hybrid Systems Chapter 4 Stability

Gramians based model reduction for hybrid switched systems

On the stability of switched positive linear systems

Preliminary results on the stability of switched positive linear systems

MATH 423 Linear Algebra II Lecture 33: Diagonalization of normal operators.

Hybrid Control and Switched Systems. Lecture #11 Stability of switched system: Arbitrary switching

A Characterization of the Hurwitz Stability of Metzler Matrices

Lecture Note 5: Semidefinite Programming for Stability Analysis

Optimality, Duality, Complementarity for Constrained Optimization

Matrix Mathematics. Theory, Facts, and Formulas with Application to Linear Systems Theory. Dennis S. Bernstein

LINEAR ALGEBRA BOOT CAMP WEEK 4: THE SPECTRAL THEOREM

L 2 -induced Gains of Switched Systems and Classes of Switching Signals

Some inequalities involving determinants, eigenvalues, and Schur complements in Euclidean Jordan algebras

Stability Criteria for Switched and Hybrid Systems

Real Symmetric Matrices and Semidefinite Programming

Lyapunov stability ORDINARY DIFFERENTIAL EQUATIONS

Polynomial complementarity problems

Appendix A Solving Linear Matrix Inequality (LMI) Problems

Lecture 1. 1 Conic programming. MA 796S: Convex Optimization and Interior Point Methods October 8, Consider the conic program. min.

The local structure of affine systems

Chapter III. Stability of Linear Systems

Nonlinear Control. Nonlinear Control Lecture # 8 Time Varying and Perturbed Systems

Quadratic and Copositive Lyapunov Functions and the Stability of Positive Switched Linear Systems

SCHUR IDEALS AND HOMOMORPHISMS OF THE SEMIDEFINITE CONE

Lecture 7: Positive Semidefinite Matrices

On common quadratic Lyapunov functions for stable discrete-time LTI systems

STABILITY OF PLANAR NONLINEAR SWITCHED SYSTEMS

1. Find the solution of the following uncontrolled linear system. 2 α 1 1

Stability lectures. Stability of Linear Systems. Stability of Linear Systems. Stability of Continuous Systems. EECE 571M/491M, Spring 2008 Lecture 5

ENERGY DECAY ESTIMATES FOR LIENARD S EQUATION WITH QUADRATIC VISCOUS FEEDBACK

Interior Point Methods: Second-Order Cone Programming and Semidefinite Programming

Lyapunov Stability Analysis: Open Loop

The Q Method for Symmetric Cone Programmin

May 9, 2014 MATH 408 MIDTERM EXAM OUTLINE. Sample Questions

Semidefinite Programming

Nonlinear Control. Nonlinear Control Lecture # 3 Stability of Equilibrium Points

MCE/EEC 647/747: Robot Dynamics and Control. Lecture 8: Basic Lyapunov Stability Theory

On the simultaneous diagonal stability of a pair of positive linear systems

1 Strict local optimality in unconstrained optimization

EE363 homework 7 solutions

Solution of Additional Exercises for Chapter 4

The Trust Region Subproblem with Non-Intersecting Linear Constraints

On Stability of Linear Complementarity Systems

Analysis and design of switched normal systems

Semidefinite Programming Basics and Applications

Daniel Liberzon. Abstract. 1. Introduction. TuC11.6. Proceedings of the European Control Conference 2009 Budapest, Hungary, August 23 26, 2009

Fischer-Burmeister Complementarity Function on Euclidean Jordan Algebras

Chapter 1: Systems of Linear Equations and Matrices

Some New Results on Lyapunov-Type Diagonal Stability

Hybrid Systems - Lecture n. 3 Lyapunov stability

Linear Algebra, part 2 Eigenvalues, eigenvectors and least squares solutions

On Dwell Time Minimization for Switched Delay Systems: Free-Weighting Matrices Method

Topics in Applied Linear Algebra - Part II

Lecture: Introduction to LP, SDP and SOCP

Introduction to Nonlinear Control Lecture # 3 Time-Varying and Perturbed Systems

Geometric Mapping Properties of Semipositive Matrices

On Linear Copositive Lyapunov Functions and the Stability of Switched Positive Linear Systems

Lyapunov Stability Theory

A Globally Stabilizing Receding Horizon Controller for Neutrally Stable Linear Systems with Input Constraints 1

On the Kalman-Yacubovich-Popov lemma and common Lyapunov solutions for matrices with regular inertia

Convex Optimization. (EE227A: UC Berkeley) Lecture 6. Suvrit Sra. (Conic optimization) 07 Feb, 2013

Properties of Solution Set of Tensor Complementarity Problem

20-Apr-2009, 15:21. The conference website can be found at

Topic # /31 Feedback Control Systems. Analysis of Nonlinear Systems Lyapunov Stability Analysis

LIE TRANSFORM CONSTRUCTION OF EXPONENTIAL NORMAL FORM OBSERVERS. Suba Thomas Harry G. Kwatny Bor-Chin Chang

Stability and Stabilizability of Switched Linear Systems: A Short Survey of Recent Results

A Delay-dependent Condition for the Exponential Stability of Switched Linear Systems with Time-varying Delay

Jordan normal form notes (version date: 11/21/07)

Nonlinear Programming Models

March 8, 2010 MATH 408 FINAL EXAM SAMPLE

Lecture: Examples of LP, SOCP and SDP

A note on simultaneous block diagonally stable matrices

HW3 - Due 02/06. Each answer must be mathematically justified. Don t forget your name. 1 2, A = 2 2

15. Conic optimization

Transcription:

On common linear/quadratic Lyapunov functions for switched linear systems M. Seetharama Gowda Department of Mathematics and Statistics University of Maryland, Baltimore County Baltimore, Maryland, USA gowda@math.umbc.edu, http://www.math.umbc.edu/ gowda On common linear/quadratic Lyapunov functions for switched linear systems p.1/30

Joint work with Melania Moldovan Based on the forthcoming article in Nonlinear analysis and variational problems, Eds: P. Pardalos, Th.M. Rassias, and A.A. Khan, Springer 2009, Volume dedicated to the memory of George Isac Objective is to study switched linear systems via complementarity and duality ideas. On common linear/quadratic Lyapunov functions for switched linear systems p.2/30

Outline Continuous/discrete/positive linear systems Complementarity ideas Z-matrices and transformations Switched systems Duality ideas Positive switched systems An extension of Mason-Shorten result Miscellaneous On common linear/quadratic Lyapunov functions for switched linear systems p.3/30

Continuous Linear System Given A R n n, consider ẋ + Ax = 0. Asymptotic stability: x(t) 0 as t. Lyapunov s Theorem (1893): The following are equivalent: (1) A is positive stable. (2) There exists a P 0 with A T P + PA 0. (3) The system ẋ + Ax = 0 is asymptotically stable. On common linear/quadratic Lyapunov functions for switched linear systems p.4/30

A is positive stable: Re λ > 0 for all eigenvalues of A. Q(x) := x T Px is a quadratic Lyapunov function. When A is positive stable, Q decreases along any trajectory of the system. On common linear/quadratic Lyapunov functions for switched linear systems p.5/30

Discrete Linear System x(k + 1) = Ax(k); k = 1, 2,.... Asymptotic stability: x(k) 0 as k. Stein s Theorem (1952): The following are equivalent: (1) A is Schur stable. (2) There exists a P 0 with P A T PA 0. (3) The system x(k + 1) = Ax(k) is asymptotically stable. On common linear/quadratic Lyapunov functions for switched linear systems p.6/30

A Schur stable: λ < 1 for all eigenvalues of A. Q(x) := x T Px is a quadratic Lyapunov function. When A is Schur stable, Q decreases along any trajectory of the system. On common linear/quadratic Lyapunov functions for switched linear systems p.7/30

Positive systems All trajectories of ẋ + Ax = 0 are constrained to R+. n This happens if and only if A is a Z-matrix: All off-diagonal entries are nonpositive. If there is d > 0 such that A T d > 0, then Q(x) = x T d is a linear Lyapunov function. In this case, the system is asymptotically stable. On common linear/quadratic Lyapunov functions for switched linear systems p.8/30

Linear complementarity problems A R n n and q R n. LCP(A,q): Find x R n such that x 0, Ax + q 0, and x T (Ax + q) = 0. A is a Q-matrix if LCP(A,q) has a solution for all q. A is a P-matrix if LCP(A,q) has a unique solution for all q or equivalently, x Ax 0 x = 0. On common linear/quadratic Lyapunov functions for switched linear systems p.9/30

Semidefinite LCPs L : S n S n linear, Q S n. SDLCP(L,Q): Find X S n such that X 0, L(X) + Q 0, and X,L(X) + Q = 0. L has the Q-property if SDLCP(L,Q) has a solution for all Q S n. L has the P-property if X and L(X) commute and X L(X) 0 implies X = 0. On common linear/quadratic Lyapunov functions for switched linear systems p.10/30

LCPs and positive systems Let A be a Z-matrix. Then the following are equivalent: There exists d > 0 such that A T d > 0. A is a Q-matrix. A is a P-matrix. On common linear/quadratic Lyapunov functions for switched linear systems p.11/30

SDLCPs and linear systems A R n n, let L A (X) := AX + XA T the Lyapunov transformation Gowda-Song (2000): The following are equivalent: There exists a P 0 such that L A (P) 0. A is positive stable. L A has the Q-property L A has the P-property. On common linear/quadratic Lyapunov functions for switched linear systems p.12/30

SDLCPs and discrete linear systems For A R n n, let S A (X) := X AXA T the Stein transformation. Gowda-Parthasarathy (2000) : The following are equivalent: There exists a P 0 such that S A (P) 0. A is Schur stable (equivalently, S A is positive stable). S A has the Q-property. S A has the P-property. Do L A and S A have some sort of Z-property? On common linear/quadratic Lyapunov functions for switched linear systems p.13/30

The Z-property V Finite dimensional real Hilbert space. K proper cone in V : K is a closed convex cone with K ( K) = {0}. A linear L : V V has the Z-property if x K, y K, x,y = 0 L(x),y 0 where K := {y V : y,x 0 x K}. L A and S A have the Z-property on the semidefinite cone. On common linear/quadratic Lyapunov functions for switched linear systems p.14/30

Properties equivalent to the Z-property Schneider-Vidyasagar (1970) exp( tl)(k) K for all t 0. ẋ + L(x) = 0, x(0) K x(t) K for all t 0. Stern (1981): For a Z-transformation, the following are equivalent: There exists d int(k) such that L(d) int(k). L 1 exists and L 1 (K) K. L is positive stable. On common linear/quadratic Lyapunov functions for switched linear systems p.15/30

Gowda-Tao (2009): The above are further equivalent to: L has the Q-property with respect to K. Open problem: Does L have the P-property? On common linear/quadratic Lyapunov functions for switched linear systems p.16/30

Switched linear systems {A 1,A 2,...,A m } R n n. ẋ + A σ x = 0, σ {1, 2,...,m} is a switched linear system. Switching Signal σ : [0, ) {1, 2,...,m} is piecewise constant. Survey article by R. Shorten et al in SIAM Review, 2007, Monograph: Switching systems by D. Liberzon. On common linear/quadratic Lyapunov functions for switched linear systems p.17/30

The (uniform exponential) asymptotic stability: There exists β > 0 such that x(t) Me βt x(0) for all t, all solutions, and all signals. On common linear/quadratic Lyapunov functions for switched linear systems p.18/30

Sufficient condition for asymptotic stability: P 0 and A T i P + PA i 0 for all i = 1, 2,...,m. Then V (x) := x T Px is a Common Quadratic Lyapunov Function. Question: When do we have a CQLF? Known conditions: Commutativity, simultaneous triangularization, Lie algebraic conditions, etc. Similar results/ideas for discrete switched systems. On common linear/quadratic Lyapunov functions for switched linear systems p.19/30

Kamenetskiy and Pyatnitskiy (1987): There exists a CQLF if and only if m i=1 (A iy i + Y i A T i ) = 0, Y i 0, i Y i = 0 i. On common linear/quadratic Lyapunov functions for switched linear systems p.20/30

L i : H H linear for i = 1, 2,...,m. L 1 is positive stable and has the Z-property on proper cone K. Moldovan-Gowda (2009) The following are equivalent: There exists d K such that L i (d) K for all i = 1,...,m. m 1 LT i (y i) = 0, y i K i y i = 0 i. Proof based on duality/theorem of alternative: On common linear/quadratic Lyapunov functions for switched linear systems p.21/30

Let K 1 H 1 and K 2 H 2. L : H 1 H 2 linear. Then exactly one of the following is consistent: L(x) (K 2 ), x (K 1 ). L T (y) K1, 0 y K 2. On common linear/quadratic Lyapunov functions for switched linear systems p.22/30

Positive switched systems ẋ + A σ x = 0, σ {1, 2,...,m} where each A i is a Z-matrix. Q(x) = x T d is a common linear Lyapunov function if d > 0 and A T i d > 0 for all i. In this case, the switched system is asymptotically stable. Question: When do we have such a d? On common linear/quadratic Lyapunov functions for switched linear systems p.23/30

Complementarity ideas again x y = min{x,y} Vertical linear complementarity problem: x (A T 1 x + q 1 ) (A T mx + q m ) = 0. If q i < 0, then x 0 and A T i x q i > 0 for all i; such an x produces a vector d > 0 such that A T i d > 0 for all i. On common linear/quadratic Lyapunov functions for switched linear systems p.24/30

Gowda-Sznajder (1994) VLCP has a unique solution for all q 1,q 2,,q m if and only if all column representatives of A = {A 1,A 2,...,A m } are P-matrices. On common linear/quadratic Lyapunov functions for switched linear systems p.25/30

Mason-Shorten (2007) A 1 and A 2 are positive stable Z-matrices. The following are equivalent: There exists d > 0 in R n such that A T 1 d > 0 and AT 2 d > 0. Every column rep. of {A 1,A 2 } is positive stable. Every column rep. of {A 1,A 2 } has positive determinant. On common linear/quadratic Lyapunov functions for switched linear systems p.26/30

Song-Gowda-Ravindran (2003) A is a compact set of positive stable Z-matrices: There exists d > 0 such that A T d > 0 for all A A. Every column rep. of A is a P-matrix. On common linear/quadratic Lyapunov functions for switched linear systems p.27/30

A = {A 1,A 2,...,A m } pos. stable Z-matrices A # := set of all column representatives of A Â := { m A i D i : D i is nonnegative and diagonal, m D i = I } 1 1 On common linear/quadratic Lyapunov functions for switched linear systems p.28/30

Moldovan- Gowda (2009) The following are equivalent: (1) There exists d > 0 such that A T i i = 1, 2,...,m. (2) Every matrix in  is a P-matrix. (3) Every matrix in A # is a P-matrix. d > 0 for all (4) Every matrix in A # has positive determinant. (5) Every matrix in  has positive determinant. (6) Every VLCP corresponding to A T has a unique solution. On common linear/quadratic Lyapunov functions for switched linear systems p.29/30

Let A be a compact set of positive stable Z-matrices in R n n. Then the following are equivalent: (1) There exists a d > 0 such that A T d > 0 for all A A. (2) Every column representative of A is a P-matrix. (3) Every column representative of A has positive determinant. On common linear/quadratic Lyapunov functions for switched linear systems p.30/30