Sensor Failure Tolerant Supervisory Control

This paper discusses problems related to partial observation supervisory controllers with possibly faulty sensors using the framework of discrete-event systems. At initialization all sensors are operational such that the sensors observe occurrences of events and transmit those observations to the controller, but, when a sensor fails, it ceases to send signals to the controller. A new version of observability is introduced that is part of the necessary and sufficient conditions for controller existence under the assumption of faulty sensors. A polynomial-time construction is given that can be used to test for and then synthesize a non-blocking controller with faulty sensors using standard supervisory control methods.