A prediction based iterative decomposition algorithm for scheduling large-scale job shops

In this paper, we present a prediction based iterative decomposition algorithm for solving large-scale job shop scheduling problems using the rolling horizon scheme and the prediction mechanism, in which the original large-scale scheduling problem is iteratively decomposed into several sub-problems. In the proposed algorithm, based on the job-clustering method, we construct the Global Scheduling Characteristics Prediction Model (GSCPM) to obtain the scheduling characteristics values, including the information of the bottleneck jobs and the predicted value of the global scheduling objective. Then, we adopt the above scheduling characteristics values to guide and coordinate the process of the problem decomposition and the sub-problem solving. Furthermore, we propose an adaptive genetic algorithm to solve each sub-problem. Numerical computational results show that the proposed algorithm is effective for large-scale scheduling problems.