A Parameter Adaptive Genetic Algorithm With Restarts for Flexible Job Shop Scheduling Problems with Worker Flexibility

In Production Scheduling (PS), the Flexible Job Shop Problem with Worker Flexibility (FJSSP-W) demands to determine a (near) optimal sequence of jobs while assigning their corresponding operations to admissible machine-worker combinations. The paper proposes a Genetic Algorithm (GA) variant designed for the FJSSP-W. It maintains a population of feasible candidate solutions throughout the search process and derives its parameter configuration from problem-specific attributes, e.g., the flexibility of the production environment, and the variety of the processing times. Further, the GA adapts its mutation and recombination parameters and employs a restarting scheme with population growth to avoid premature convergence. Experimental results are compared to state-of-the-art FJSSP-W solvers. The results demonstrate the GA's scaling potential in terms of problem complexity. The GA keeps its performance, while other approaches decline when exposed to problems of higher complexity. The GA appears to be beneficial for problems with a high amount of assignment options per operation and low variety in processing times. The results confirm the suitability of problem-specific parameter settings and the restarting scheme for a wide range of PS problems, and FJSSP-W, particularly.