Combination of dispatching rules and prediction for solving multi-objective scheduling problems

Multi-objective scheduling problems are generally difficult to solve in a limited amount of time. Using dispatching rules is able to find solutions for a relatively acceptable amount of time, but with quality far from the optimal one. The method presented in this paper is based on the same principle as other dispatching rule methods, but improved by combining the method with predicted data. Predicted information of a future schedule provides better overview of the current state of the solving problem, possible impacts of each decision, and consequently leads to better performance of classical dispatching rules. Calculation of predicted data on the other hand requires only a fraction of time, as a result of quadratic complexity of the method for prediction. The proposed method is implemented and deployed in several make-to-order manufacturing sites.