Characterization of feasible controls for Petri nets with unobservable transitions

Supervisory control of discrete event systems modeled by Petri nets involves enforcing a set of constraints on the state, or marking, of the plant Petri net. Unobservable transitions within the plant may force the control designer to alter these original constraints to account for the inability of the supervisor to act when one of these transitions fires. A method for characterizing the constraints and the associated controllers which can be realized in the face of unobservable transitions is presented in this paper. While these are preliminary results, the characterization can be used by the designer to determine which linear constraints can be implemented without change, which constraints need to be transformed, and how those constraints should be transformed