Some approximations for stochastic games with unbounded reward and average payoff

Abstract

We consider approximation problems arising in stochastic games with unbounded cost and expected average cost criterion. We focus on finite state approximation of stochastic games with a countable state space. In addition to the convergence of the value, we obtain (i) the convergence of the policies, (ii) the robustness of policies. The growing interest in stochastic games with unbounded cost in recent years was partly driven by applications of stochastic games in telecommunications systems in general, and in queueing systems in particular.