Stability and Criticality Analysis for Integer Linear Programs With Markovian Problem Data

Abstract

This paper presents the stability and criticality analysis of integer linear programs with respect to perturbations in stochastic data given as Markov chains. These perturbations affect the initial distribution, the transition matrix, or the stationary distribution of Markov chains. Stability analysis is concerned with obtaining the set of all perturbations for which a solution remains optimal. This paper gives expressions for stability regions for perturbations in the initial distribution, the transition matrix, the stationary distribution, and the product of elements of the transition matrix and the stationary distribution. Furthermore, criticality measures that describe the sensitivity of the objective function with respect to an element of the problem data are derived. Stability regions that preserve the stochasticity of the problem data are given. Finally, stability regions for perturbations of elements of the transition matrix, given that the problem is not linear in the initial distribution or the transition matrix, are obtained using a small perturbation analysis. The results are applied to sensor placement problems and numerical examples are given.