Stability Analysis of Discrete-Time Linear Complementarity Systems

TL;DR

This paper introduces a novel stability analysis framework for Discrete-Time Linear Complementarity Systems (DLCS), including new conditions, an exact verification algorithm, and applications to Linear Complementarity Systems (LCS).

Contribution

It provides the first exact algorithm for DLCS stability verification and derives new Lyapunov stability conditions without requiring solution uniqueness.

Findings

Feasibility of copositive programs for stability conditions

Development of an exact cutting plane algorithm

Numerical examples demonstrating the approach

Abstract

A Discrete-Time Linear Complementarity System (DLCS) is a dynamical system in discrete time whose state evolution is governed by linear dynamics in states and algebraic variables that solve a Linear Complementarity Problem (LCP). The DLCS is the hybrid dynamical system that is the discrete-time counterpart of the well-known Linear Complementarity System (LCS). We derive sufficient conditions for Lyapunov stability of a DLCS when using a quadratic Lyapunov function that depends only on the state variables and a quadratic Lyapunov function that depends both on the state and the algebraic variables. The sufficient conditions require checking the feasibility of a copositive program over nonconvex cones. Our results only assume that the LCP is solvable and do not require the solutions to be unique. We devise a novel, exact cutting plane algorithm for the verification of stability and the computation of the Lyapunov functions. To the best of our knowledge, our algorithm is the first exact approach for stability verification of DLCS. A number of numerical examples are presented to illustrate the approach. Though our main object of study in this paper is the DLCS, the proposed algorithm can be readily applied to the stability verification of LCS. In this context, we show the equivalence between the stability of a LCS and the DLCS, resulting from a time-stepping procedure applied to the LCS for all sufficiently small time steps.