Adaptive control of discrete time weakly controlled Markov and semi-Markov models

Abstract

A Bayesian approach to Markov decision process problem [1] under stochastic uncertainty, when unknown transition probabilities are weakly disturbed with disturbances dependent on a decision strategy only is investigated. Observed decision process is assumed to be stationary in discrete time with finite, countable or measurable phase state is based on separation principle of assessment and optimization problems.