Research On Laser-Controlled Locking System Based On Deep Reinforcement Learning Algorithm

In order to improve the overall performance of all-solid-state single-frequency lasers,we design an artificial intelligence control system based on deep reinforcement learning algorithm to complete the locking control of frequency and light field phase of all-solid-state single-frequency lasers.The system is based on the Microcontroller Unit(MCU)operation,through the MCU’s analog-to-digital conversion system,to achieve the acquisition of error signals and feedback control signal output.The feedback control signal acts on the laser resonator through piezoelectric ceramics and other devices to realize the automatic locking of the frequency and light field phase of the all-solid-state single-frequency laser.By designing the laser locking control system based on servo system and BP neural network,the miniaturization and adaptive function of the control system are realized.Furthermore,the deep reinforcement learning algorithm is applied to the processing of error signals and the generation of feedback signals,and based on the all-solid-state single-frequency laser,a framework structure of deep reinforcement learning that is compatible with it is constructed,and the program logic that can run correctly in the MCU is written.The error signal in the laser system collected by the MCU is used as the input of the neural network,and a new error signal is generated after the signal feedback of the neural network adjusts the output to control the laser system,forming a feedback control loop.The neural network is trained by continuously reinforcing learning and updating until it converges.When the neural network converges,the deep neural network of the control system can adaptively output the optimal feedback signal under different error signal conditions,that is,the lock control of the laser can be realized.