A fuzzy logic system for dynamic job shop scheduling

We study a dynamic job shop scheduling problem to minimize total weighted tardiness. Since it is a strongly NP-hard problem, the typical and efficient solution approach is to use dispatching rules; however, the performance of these simple dispatching rules is usually far from satisfactory. A fuzzy logic system is proposed and designed to dynamically guide the selection of dispatching rules for different problem instances at different stages by learning from fuzzy rules and previous solutions. Several experiments are designed and conducted in different scenarios to evaluate the effectiveness of the fuzzy logic system against the traditional dispatching rules. Preliminary experimental results indicate that the fuzzy logic system outperforms all the dispatching rules tested.