A Genetic Algorithm for Job-Shop Problems with Various Schedule Quality Criteria

Much recent research has investigated the use of genetic algorithms (GAs) in job-shop scheduling. Mostly, this has involved comparison or construction of ingenious reprsentations and operators in the context of finding a schedule which minimises makespan. Many more criteria exist with which to judge schedule quality, however. Often, makespan may be a low priority aspect of schedule quality, We describe a generally-applicable GA approach to job-shop problems and examine its performance on a range of benchmark problems, for each of a wide range of different schedule quality criteria. Performance is compared against a range of standard heuristic rules, and also against a stochastic hillclimbing (SH) method. We find that the GA does best overall across all kinds of schedule quality criterion.