The discounted method and equivalence of average criteria for risk-sensitive Markov decision processes on Borel spaces (Q964743)

The work concerns discrete-time Markov decision processes evolving on a Borel space. The system is driven by a risk-averse decision maker with a constant risk sensitivity coefficient \(\lambda > 0\), and the performance of a control policy is measured by the (superior limit) risk-sensitive average cost criterion associated with a nonnegative cost function. Under mild (semi-) continuity and compactness conditions, the following problems are studied via the discounted approach: (i) establishing the existence of optimal stationary policies, and (ii) determination of conditions under which the equality of the optimal value functions associated with the inferior limit and superior limit average criteria can be ensured. The approach of the paper relies on standard dynamic programming ideas and on a simple analytical derivation of a Tauberian relation.