Enhancing the Effectiveness of Multi-pass Scheduling Through Optimization via Simulation

Abstract

While difficult in their own right, scheduling problems are further complicated by the concurrent flow of various parts, the sharing of different types of resources, and the random occurrence of disruptive events. To deal with such complexity, multi-pass scheduling has been developed. Successful application of multi-pass scheduling, however, largely depends on its ability to quickly and effectively select the best decision-making rule. The objective of the present work is to enhance the performance of multi-pass scheduling through optimization via simulation. To this end, we combine random search and statistical selection to create a potent approach for optimization over a large but finite solution space when the objective function must be evaluated using noisy estimation. The nested partitions method is a global search strategy that is continuously adapted via a partitioning of the feasible solution region. By allocating computing resources to potentially critical design points, the optimal computing budget allocation method, in turn, provides an adaptive sampling mechanism from a feasible region. Through carefully designed experiments, the proposed approach is shown to perform markedly better than naive multi-pass scheduling.