Markov Decision Processes with Incomplete Information and Semiuniform Feller Transition Probabilities

This paper deals with control of partially observable discrete-time stochastic systems. It introduces and studies Markov Decision Processes Incomplete Information semiuniform Feller transition probabilities. The important feature these models is that their classic reduction to Completely Observable belief states preserves continuity Under mild assumptions on cost functions, optimal policies exist, optimality equations hold, value iterations converge values for models. In particular, Partially the results this imply new generalize several known sufficient conditions observation probabilities weak states, existence policies, validity defining convergence values.