New framework for dynamic scheduling of production systems

Abstract

A concept for dynamic scheduling in manufacturing systems is proposed. The scope of 'dynamic scheduling' treated includes online dynamic change of some scheduling parameters such as rules for part dispatching, machine selection, or routing. IF-THEN-type heuristic operators are utilized to perform this online, real-time rule selection, and offline machine learning is used to obtain more detailed and powerful heuristics than those implemented by human experts or programmers. A learning algorithm has been developed to formulate operators that can treat quantity-type as well as quality-type information. A prototype computer program named Learning Aided Dynamic Scheduler (LADS) has been developed. A simulation study using LADS indicates good results for dynamic scheduling using LADS.