Supervisor synthesis and throughput optimization of partially-controllable manufacturing systems

Abstract

One of the challenges in the design of supervisors with optimal throughput for manufacturing systems is the presence of behavior outside the control of the supervisor. Uncontrollable behavior is typically encountered in the presence of (user) inputs, external disturbances, and exceptional behavior. This paper introduces an approach for the modeling and synthesis of a throughput-optimal supervisor for manufacturing systems with partially-controllable behavior on two abstraction levels. Extended finite automata are used to model the high abstraction level in terms of system activities, where uncontrollability is modeled by the presence of uncontrollable activities. In the lower abstraction level, activities are modeled as directed acyclic graphs that define the constituent actions and dependencies between them. System feedback from the lower abstraction level, including timing, is captured using variables in the extended finite automata of the higher abstraction level. For throughput optimization, game-theoretic methods are employed on the state space of the synthesized supervisor to determine a guarantee to the lower-bound system performance. This result is also used in a new method to automatically compute a throughput-optimal controller that is robust to the uncontrollable behavior.