Optimization of control reconfiguration strategies

A mega-controller architecture is introduced for reconfiguration of discrete event system controllers with dynamic observation sets. The mega-controller uses several reconfiguration strategies to revise the control policies based on the availability of the observation means. It is known that the mega-controller could have multiple choices of reconfiguration strategies in a given state and therefore it should decide between these choices. The current paper provides an optimization model for this decision process. By converting the finite automata model of the reconfigured controller (or mega-controller) to a Markov decision process, the minimum cost reconfiguration policy at every state of the mega-controller is determined. The optimal policy is obtained by minimizing the cost over all policies. Computation of this optimization problem is further reduced for a special class of systems by the applying two lemmas proposed in this paper