Rolling Horizon Procedure for Large-scale Job-shop Scheduling Problems

Abstract

A rolling horizon procedure (RHP) is developed to address a kind of large-scale job-shop scheduling problems with release times in this paper. At each decision time, jobs arriving within a future horizon are included into a predictive window, a rolling window is identified based on the operation precedence relations of jobs in the predictive window, a sub- problem is constructed based on the rolling window, and the sub-problem is solved by use of the shifting bottleneck procedure (SBP). A partial solution obtained from the sub- problem is preserved to be the existing schedule, and the left is included into the next predictive window to be rescheduled at the next decision time. The existing schedule will finally be integrated into the global solution for the job-shop problem. In such a manner RHP can decompose a large-scale job-shop scheduling problem into a series of single-machine scheduling problems with restricted sizes. The computational results show that such RHPs are effective and outperform the priority dispatching rules.