Automatic design of scheduling policies for dynamic flexible job shop scheduling via surrogate-assisted cooperative co-evolution genetic programming

Abstract

At present, a lot of references use discrete event simulation to evaluate the fitness of evolved rules, but which simulation configuration can achieve better evolutionary rules in a limited time has not been fully studied. This study proposes three types of hyper-heuristic methods for coevolution of the machine assignment rules (MAR) and job sequencing rules (JSR) to solve the DFJSP, including the cooperative coevolution genetic programming with two sub-populations (CCGP), the genetic programming with two sub-trees (TTGP) and the genetic expression programming with two sub-chromosomes (GEP). After careful parameter tuning, a surrogate simulation model is used to evaluate the fitness of evolved scheduling policies (SP). Computational simulations and comparisons demonstrate that the proposed surrogate-assisted CCGP method (CCGP-SM) shows competitive performance with other evolutionary approaches using the same computation time. Furthermore, the learning process of the proposed methods demonstrates that the surrogate-assisted GP methods help accelerating the evolutionary process and improving the quality of the evolved SPs without a significant increase in the length of SP. In addition, the evolved SPs generated by the CCGP-SM show superior performance as compared with existing rules in the literature. These results demonstrate the effectiveness and robustness of the proposed method.