ACTUATOR FAILURE IN DECENTRALIZED SUPERVISORY CONTROL SYSTEMS

We consider the problem of actuator failure in supervisory control of decentralized discrete event systems. A subset of events that are disabled by the decentralized supervisor may in actuality be enabled as a result of interference in the signal from the supervisor to the actuator. Extending the method of Thorsley and Teneketzis (2006) to decentralized supervisors, we determine conditions that ensure that a given set of actuator failures can never allow the system to complete an undesirable sequence of events. We use the technique of fictitious play to find a member-by-member optimal control specification.