Bounds on Stationary Expectations for Markov Processes

Many performance engineering and operations research modeling formulations lead to Markov models in which the key performance measure is an expectation defined in terms of the stationary distribution of the process. In models of realistic complexity, it is often difficult to compute such expectations in closed form. In this talk, we will discuss a simple class of bounds for such stationary expectations, and describe some of the mathematical subtleties that arise in making rigorous such bounds. We will also discuss how such bounds can be used algorithmically. This work is joint with Denis Saure and Assaf Zeevi.