Review on ensemble meta-heuristics and reinforcement learning for manufacturing scheduling problems

With the development of Artificial Intelligence, Internet of Things and Big Data, intelligent manufacturing has become a new and popular trend in manufacturing industries. Manufacturing scheduling is one of the most critical components in intelligent manufacturing systems. It aims to optimize some specific objectives, e.g., production cost, customer satisfaction and energy efficiency, by making optimal decisions of processing routes, machine assignment, operation sequence, etc. Due to manufacturing scheduling problems featured with large scale, strong coupling and real-time optimization requirements, it is a huge challenge to effectively cope with them. As the extensive and successful applications of artificial intelligence in manufacturing areas, meta-heuristics and reinforcement learning methods achieve great breakthroughs in addressing manufacturing scheduling problems. It is noted that a hybridization of meta-heuristic and reinforcement learning algorithms has been recently proposed to solve such complicated problems. Firstly, this work summarizes the designs of meta-heuristics and reinforcement learning methods for dealing with manufacturing scheduling problems, respectively. Secondly, we review the hybridization of meta-heuristics and reinforcement learning methods in solving manufacturing scheduling problems, where the essential roles of reinforcement learning for meta-heuristics are analyzed and discussed from the views of ensemble methods, optimization criteria, scheduling models, performance evaluation metrics and stopping conditions. Finally, we conclude this work and sum up future research directions regarding the hybridization methods in handling manufacturing scheduling problems.