Modular synthesis of efficient schedules in a timed discrete event plant

Abstract

Treats optimal scheduling in large timed discrete event systems as a supervisory control problem. Scheduling tasks in a steel plant are treated as a realistically sized case study. A sequence of tasks must be completed in as soon as possible, while satisfying all the constraints in the model. These different constraints are specified via different components in a modular plant representation. Components can be represented as timed Petri nets, leading to a graph of interacting modules. The acyclic nature of the graph consisting of the most critical components is exploited in order to find a heuristic but fast way of searching through the very large set of feasible orderings.