Unbounded cost Markov decision processes with limsup and liminf average criteria: new conditions

This paper deals with denumerable discrete-time Markov decision processes with unbounded costs. The criteria to be minimized are both of the limsup and liminf average criteria, instead of only the limsup average criterion widely used in the previous literature. We give another set of conditions under which the existence of an optimal stationary policy for the two average criteria is ensured. The results in this paper are applied to an admission control queueing model. Copyright Springer-Verlag 2005