Reinforcement learning approach to learning human experience in tuning cavity filters

Abstract

Owing to the rapid development of the communication industry, various kinds of radio frequency components are in great demand and put into mass production. Among them, passive devices such as microwave cavity filters, duplexers and combiners have experienced fast and unexpected upgrades. However, the tuning process of these products, which is always manually operated, still seems hard to be automatically replaced or improved because of the difficulties in extracting human experience. In this study, we make deep investigations into some previous automatic cavity filter tuning solutions, especially the ones using intelligent algorithms. In addition, we propose the method of intelligent tuning based on the reinforcement learning algorithm which dynamically extracts the human strategies during the tuning process. The experimental results prove the powerful performance of reinforcement learning in mastering human skills.