Classical and logic based regulator design and its complexity for partially observed automata

The problem of steering the state of a partially observed finite automaton (i.e. a partially observed deterministic finite input-state output system) M to a desired terminal state is considered. The authors provide necessary and sufficient dynamic programming conditions for the controllability of M. These conditions are stated in terms of backward recurrence equations involving the state estimate sets generated by either classical dynamical observers or logical dynamical observers.<<ETX>>