Evaluation of ordering strategies for constraint satisfaction reactive scheduling

Abstract

Reactive scheduling is essential in any scheduling system to incrementally reconcile the discrepancies between the generative schedule and current status of the factory. Typical events requiring the reactive scheduling process include the delayed delivery of materials, machine breakdown, and failure to meet quality control standards. To efficiently cope with these unexpected events, a reactive scheduling procedure is proposed based on the constraint satisfaction approach, and applied to the reactive adjustment of hot-rolling schedules at steelworks. Various domain-specific strategies are devised as variable and value-ordering heuristics to guide the search directions in the reactive scheduling process, and implemented in a scheduling expert system. From the experiment conducted on the prototype system developed, we find that the schedule performance is mainly determined by the variable ordering strategy based on the criticality of scheduling activities, and it can be further improved by adopting the appropriate value-ordering strategy based on the least constraining resource heuristic.