Robust parameter design for genetic algorithm to bicriteria open-shop scheduling problems

Abstract

The minimised sum of makespan and total tardiness of an open-shop scheduling problem for a set of jobs with non-identical ready times are considered. A genetic algorithm is employed to solve this problem. The fitness function comprises both the modified makespan and total tardiness, thus avoiding the effect of a dominating criterion. A variable weighting approach for the two criteria is utilised to alter the search directions of each generation and accelerate convergence of the algorithm. An experimental design is employed to determine the best combination of parameter levels that can then be adopted in the genetic algorithm. Finally, the proposed genetic algorithm is implemented using the data of case company. The results show that the schedule generated by the proposed genetic algorithm outperforms the company's current FCFS with SPT scheduling approach in terms of makespan and total tardiness as well as number of tardy jobs.