Novel Rolling Horizon Procedures for Large-scale NP-hard Single-machine Scheduling Problem
【摘要】：正A kind of rolling horizon procedures for 1|ri|∑C1 scheduling problem with globally known information is presented in this paper. The problem is solved by using a series of rolling on-line small-scale locally scheduling instead of a off-line large-scale globally scheduling. At every decision time, the sub-problem for locally scheduling is solved based on the rolling window and a portion of the solution is implemented. While the decision time is being put forward, global schedule is realized step by step. Usually it is hard to analyze the global performance of rolling horizon procedures as a heuristic. However, in this paper, a terminal penalty function is appended to the objective function of sub-problems to make the local objective consistent with the global objective. Global performances of rolling horizon procedures are analyzed. Several conclusions would be drawn out about the global performance of the special rolling horizon procedures. It is concluded that the procedural global performance is getting close to the ultimately realized actual cost and getting better and better from one decision time to another. It is assured that the global performance of this kind of procedures is not worse than that of the initial schedule anyway.