Analysis of the learning results of a reinforcement learning AI that suggests scanning paths in the Selective Laser Thermoregulation method

We develop the Selective Laser Thermoregulation (SLT) method, in which a metal sample is heated by scanning the laser beam point at high-speed using a fiber laser and a galvanometer. The SLT method has difficulty in finding a scanning path that can heat the sample uniformly and at a high temperature. Therefore, we developed a reinforcement learning AI using the Temporal Difference (TD) method, which suggests scanning paths from the coordinates of irradiation points.