Abstract
Examination of the job-shop scheduling literature uncovers a striking trend. As methods for the deterministic job-shop problem have gradually improved over the years, they have come to rely on neighbourhoods for selecting moves that are more and more constrained. We document this phenomenon by focusing on the approach of Nowicki and Smutnicki (Management Science, 1996, 42(6), 797-813), noted for proposing and implementing the most restrictive neighbourhood in the literature. The Nowicki and Smutnicki (NS) method which exploits its neighbourhood by a tabu search strategy, is widely recognised as the most effective procedure for obtaining high quality solutions in a relatively short time. Accordingly, we analyse the contribution of the method's neighbourhood structure to its overall effectiveness. Our findings show, surprisingly, that the NS neighbourhood causes the method's choice of an initialisation procedure to have an important influence on the best solution the method is able to find. By contrast, the method's choice of a strategy to generate a critical path has a negligible influence. Empirical testing further discloses that over 99.7% of the moves chosen from this neighborhood (by the NS rules) are disimproving-regardless of the initial solution procedure or the critical path generation procedure employed. We discuss implications of these findings for developing new and more effective job-shop algorithms.
Original language | English |
---|---|
Pages (from-to) | 457-480 |
Number of pages | 24 |
Journal | Journal of Heuristics |
Volume | 6 |
Issue number | 4 |
DOIs | |
Publication status | Published - 1 Sept 2000 |
Funding
This work has been sponsored by the Royal Society of Edinburgh and the Carneigie Trust for the Universities of Scotland while the first author was a visiting researcher at the University of Colorado, Boulder. The authors would like to thank Dr. Nowicki for providing us with his code from which we learnt a great deal, Professor Frank Werner and Dr. Thomas Tautenhahn for helping us greatly with our implementation of insert and the assistance provided by Professor Fred Glover in the preparation of this work which has greatly enhanced the quality of the paper.
ASJC Scopus subject areas
- Artificial Intelligence
- Computational Theory and Mathematics
- Control and Systems Engineering
- Theoretical Computer Science