Industrial scheduling with Monte Carlo tree search and machine learning

Abstract

Although scheduling has been studied for decades, no general approach for solving industrial scheduling problems has yet been found. The variety of production systems often requires unique modifications of scheduling algorithms to meet industrial requirements. To address this challenge, an algorithm is proposed that combines Monte Carlo tree search and machine learning. It tries to eliminate the need for problem-specific knowledge while improving the production system’s performance. The algorithm has been investigated using a reentrant flow-shop problem. The results show that a combination of random-based search algorithms and machine learning is a promising way to handle complex industrial scheduling problems.