On the separation of estimation and control in discrete-event systems

The discussion of this paper concerns both centralized and decentralized control of logical discrete-event systems. Of interest are the maximal information sets of the centralized or decentralized supervisors and the potential for control and estimation policy independence. We show that there exists a form of a centralized supervisor's maximal information sets that is independent of the supervisor's control policy. We also show that this method of separation is not generally applicable in decentralized settings. These results are consistent with the literature in stochastic control; however, supervisory control potentially presents a more simple framework in which to explore these concepts.