Learning-aided dynamic scheduling and its application to routing problem

Abstract

Learning-aided dynamic scheduling is proposed for production line scheduling. In this concept, the scheduling rules are dynamically switched during real operations to reflect changes in the production line status, given requirements and constraints. This switching is governed by some knowledge which is automatically acquired by machine learning during the iteration of simulations of the specific production line. The machine learning is carried out in the form of generation of a binary decision tree, and a new algorithm is developed for this objective. Simulation studies on its application to a routing problem have been performed, and the effectiveness of the concept was verified.