A forecast horizon and a stopping rule for general Markov decision processes (Q1104254)

This paper presents conditions for the existence and detection of a forecast horizon for Markov decision processes with Borel state space. The conditions include the assumptions that the control set is finite and that there is a unique optimal stationary policy. The approach and results are closely related to other works in the area, e.g., \textit{K. Hinderer} and \textit{G. Hübner} [Information theory, statistical decision functions, random processes, Trans. 7th Prague Conf., Vol. A, Prague 1974, 245-257 (1977; Zbl 0416.90076)]; \textit{G. Hübner} [Bonn. Math. Schr. 98, 57-65 (1977; Zbl 0384.93041); Z. Angew. Math. Mech. 57, 477-480 (1977; Zbl 0368.90111)].