A metaheuristic based on simulation for stochastic Job-shop optimization

This paper deals with stochastic Job-shop with random processing times where the objective is to find schedules robust enough in order to minimize the total completion time of all the operations. The problem is handled by use of a multi-start metaheuristic and a simulation software. At each iteration of the metaheuristic the best deterministic schedule is tested using the SIMAN simulation language; at this given schedule is then adjoined the average simulated makespan. The metaheuristic is then searching for another different schedule in the deterministic space which could be worse than the previous one. The proposed approach is applied to a small instance for demonstration. A set of instances for job-shop problems is then adapted for stochastic use in order to validate this work. The results, both in term of computational time and of quality, show the relevance of this study.