Opportunities for Reinforcement Learning in Industrial Automation

Abstract

With the development of technology and the increasingly fierce competitive environment, the manufacturing system becomes more and more complex and requires higher coordination. Manufacturing companies are facing the challenge of highly non-linear and flexible business requirements, while the current degree of automation in the smart era has been unable to meet the needs of the environment. In the next stage, the leading manufacturing enterprises should focus on the application of digitalization and even intelligence in the automated manufacturing industry. Additionally, the successful implementation of intelligent applications in the industry and the realization of their ideal value are also the part that researchers must consider. The development of artificial intelligence (AI) and machine learning (ML) has shown that there is great potential to change the manufacturing field through advanced intelligent system tools, especially through reinforcement learning (QL). Therefore, the focuses of this article have been shown as follows: (1) Review the practical application of recent reinforcement learning in various fields, especially in the industrial field. (2) Analyze the method of reinforcement learning applied in industrial applications and its unique performance. (3) Identify the challenges and opportunities for further application of reinforcement learning in automated manufacturing, and discuss the future development of reinforcement learning to better meet the needs of intelligent manufacturing. In order to achieve these goals, this article uses a large number of literature reviews to elaborate a hierarchical analysis of the extensive practice of reinforcement learning in industrial applications. Through these analyses, we try to find the applicability of reinforcement learning in automated manufacturing systems. In addition, we take industrial process systems, human-machine assistance supervision and control, process monitoring, prevention and post-processing, and finally full material process management into consideration, and hope to achieve the desired characteristics in the process and control of the automated manufacturing industry.