Evolutionary Multitask Optimisation for Dynamic Job Shop Scheduling Using Niched Genetic Programming
Dynamic job shop scheduling (DJSS) problems are combinatorial optimisation problems where dynamic events occur during processing that prevents scheduling algorithms from being able to predict the optimal solutions in advance. DJSS problems have been studied extensively due to the difficulty of the problem and their applicability to real-world scenarios. This paper deals with a DJSS problem with dynamic job arrivals and machine breakdowns. A standard genetic programming (GP) approach that evolves dispatching rules, which is effective for DJSS problems with dynamic job arrivals, have difficulty generalising over problem instances with different machine breakdown scenarios. This paper proposes a niched GP approach that incorporates multitasking to simultaneously evolve multiple rules that can effectively cope with different machine breakdown scenarios. The results show that the niched GP approach can evolve rules for the different machine breakdown scenarios faster than the combined computation times of the benchmark GP approach and significantly outperform the benchmark GP’s evolved rules. The analysis shows that the specialist rules effective for DJSS problem instances with zero machine breakdown have different behaviours to the rules effective for DJSS problem instances with machine breakdown and the generalist rules, but there is also large variance in the behaviours of the zero machine breakdown specialist rules.
- 7.Yin, W.J., Liu, M., Wu, C.: Learning single-machine scheduling heuristics subject to machine breakdowns with genetic programming. In: Proceedings of IEEE Congress on Evolutionary Computation (CEC 2003), pp. 1050–1055 (2003)Google Scholar
- 8.Park, J., Mei, Y., Nguyen, S., Chen, G., Zhang, M.: Investigating the generality of genetic programming based hyper-heuristic approach to dynamic job shop scheduling with machine breakdown. In: Wagner, M., Li, X., Hendtlass, T. (eds.) ACALCI 2017. LNCS (LNAI), vol. 10142, pp. 301–313. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-51691-2_26CrossRefGoogle Scholar