A cost-shaping linear program for average-cost approximate dynamic programming with performance guarantees

We introduce a new algorithm based on linear programming for optimization of average–cost Markov decision processes (MDPs). The algorithm approximates the differential cost function of a perturbed MDP via a linear combination of basis functions. We establish a bound on the performance of the resulting policy that scales gracefully with the number of states without imposing the strong Lyapunov condition required by its counterpart in de Farias and Van Roy (2003). We investigate implications of this result in the context of a queueing control problem.