On a set of optimal policies in continuous time Markovian decision problem (Q2544324)

This paper is concerned with a continuous time parameter Markovian decision problem with finitely many states and actions. An \(\alpha\)-optimal policy, which maximizes the total expected return with discount factor \(\alpha\), is considered. It is shown that the set of \(\alpha\)-optimal policies consists of the policies of which action belong to the specified set. An upper bound of the number of a-optimal regions and a simple procedure to find them are proposed from the view point of a sensitivity analysis. Moreover, the limiting properties of \(\alpha\)-optimal policies and the set of optimal policies in some sense are discussed.