Fault-Tolerant Supervisory Control of Discrete Event Systems Modeled by Bounded Petri Nets

This paper considers bounded Petri nets with both controllable and uncontrollable transitions, and addresses the synthesis of a fault-tolerant supervisor in a setting where the control specifications are described via arbitrary forbidden markings. When determining the supervisor, we handle uncontrollable transitions by analyzing the reverse net and by obtaining a set of weakly forbidden markings, based on which we determine the maximally permissive control policy. We implement the supervisor by encoding the system state information into two monitor places in a way that allows us to both determine the online control policy efficiently and identify/correct single place faults (i.e., faults that corrupt the number of tokens in a single place of the Petri net, including the monitor places). The overall method need not perform reachability analysis, has low complexity requirements for online computation, and can be generalized to monitor-based control schemes that are tolerant to any number of faults.