Stray light suppression design of optomechanical systems enabled by deep reinforcement learning

Stray light suppression is a vital component in the development of optomechanical systems, but its complexity and the uncertainty surrounding scattered light require intricate mathematical calculations and a large number of simulation iterations, along with much expertise and time. Consequently, it is time-consuming and challenging to investigate the stray light suppression in optomechanical systems. To validate the feasibility of using reinforcement learning for stray light suppression, this paper adopts a model-based deep reinforcement learning method within a Monte Carlo ray-tracing environment to devise suppression strategies. The experimental results indicate that the model-based deep reinforcement learning method can provide effective stray light suppression measures for various optical system configurations, leading to significant improvements in suppression efficiency.