Automatic Control Of Sagnac Optical Interferometry System Based On Reinforcement Learning

In recent years,with the continuous development and improvement of machine learning and reinforcement learning,people are trying to apply this technology to more and more scientific and technological fields,and have had a significant impact on these fields,especially in the field of robotics,which has achieved remarkable results,which is a technological revolution.Machine learning and reinforcement learning also have amazing application prospects in the field of optics,and are a hot research topic.It is the dream of many experimental physicists to realize the tuning of optical path and components using robots.This kind of tuning has specific quality indicators,so it is suitable for machine learning and reinforcement learning technology.In this paper,we apply reinforcement learning to Sagnac optical interference system.Using programmable pm100 d optical power meter,Picomotor piezoelectric mirror frame kit and other equipment,we study and realize the automatic control of Sagnac optical interference system based on reinforcement learning greedy strategy and SPGD algorithm.Compared with manual debugging of Saganc optical interference system,the automatic control of Sagnac optical interference system based on reinforcement learning is close to the level of manual debugging,and can replace human in some experiments.At the same time,it can realize remote control of Sagnac optical interference system,so this system can also be applied to experimental platforms that cannot be adjusted manually,such as closed vacuum or extremely complex optical platforms.The main contents of thesis are as follows:(1)The Sagnac optical interference system is built experimentally.The generation principle of Sagnac interference phenomenon is analyzed in detail theoretically.In experiment,the most classical Sagnac optical interference system structure is used to build the experimental light path;(2)Research and analyze how to use Python to design the serial communication module between the computer and each device,write the data acquisition module and control module,realize the real-time extraction of the status data of Pico motor and optical power meter,and control the Sagnac optical interference system;(3)The extracted state data of Pico motor and power meter are analyzed and processed by using two algorithms of reinforcement learning greedy strategy(epsilon greedy)and SPGD,and the automatic control of Sagnac optical interference system based on reinforcement learning is realized.In recent years,with the continuous development and improvement of machine learning and reinforcement learning,people have tried to apply this technology to more and more fields of science and technology,especially in the fields of computers and physical robots,and have achieved remarkable results,and has had a major impact on these fields,it is a technological revolution.Machine learning and reinforcement learning also have amazing applications in the field of optics and are a hot research topic.