Presentation Information

[8p-N106-7]Implementing Deep Reinforcement Learning for autonomous laser cavity optimization

〇Yuki Ikeya1,2, Takashi Sekine1, Endo Tsubasa2, Toshio Otsu2, Takaaki Morita1, Yoshinori Tamaoki1, Yoshinori Kato1, Toshiyuki Kawashima1, Yohei Kobayashi1 (1.Hamamatsu Photonics K.K., 2.ISSP, Univ.Tokyo.)

Keywords:

spatial light modulator,machine learning,laser cavity

In order to obtain ideal oscillation characteristics in the development of a laser resonator, a huge amount of parameter search and control technology is required. The introduction of reinforcement learning is considered to be an effective means of solving the problem, but it is unknown whether it can be applied to a real laser system. We have made it possible to dynamically and quickly control the angle and curvature of the mirror with high degrees of freedom by using a spatial light modulator (LCOS-SLM) instead of the resonator mirror. In this presentation, we will report on a control method using DQN, which is one of the deep reinforcement learning methods.