Rule acquisition for rolling horizon heuristics in single machine dynamic scheduling

This study deals with a method for rule acquisition in single machine dynamic scheduling as well as scheduling with machines in parallel and proposes a rolling horizon procedure in which the obtained rules are used as knowledge for dispatching of jobs. In a rolling horizon procedure, a multi-period scheduling problem is solved and only the first period's solution is applied, then the multi-period model is updated one after another. Rules to be used at each period describe a schedule status and provide an adequate dispatching decision. We use inductive learning for rule acquisition and generate training cases by computational simulation involving the process of rolling schedules under dynamic environment. The applicability of the proposed method is presented by computational experiments. We also demonstrate the effectiveness of the obtained rules by applying them to a single machine dynamic scheduling.