On Markov policies for minimax decision processes (Q5927554)

The authors consider the stochastic maximization problem with minimum function, firstly with the set of all general policies (called the general problem) and secondly, restricted to the set of all Markov policies (called the Markov problem). The authors also discuss an embedded problem of the stochastic maximization problem and obtain an optimal policy through the invariant embedding approach. In the last part they consider a two-stage stochastic decision process which does not admit any optimal Markov policy.