Optimal control of Markov processes with incomplete state information