Reinforcement Learning Optimizes Laser Cutting Parameters for Optical Films

A new algorithm, Reinforcement Learning for Laser Cutting (RL2C), has been developed to optimize cutting parameters for optical films, addressing the limitations of traditional trial-and-error methods. The algorithm, presented in a paper on arXiv (2608.10549), uses Q-learning with an epsilon-greedy policy to dynamically adjust parameters such as focal length and laser power beam according to the specific properties of each film type. This approach significantly reduces taper size and film wastage, and requires fewer steps and less time to find optimal parameters compared to conventional methods. The RL2C algorithm also incorporates a dynamic environment space adaptability mechanism, enabling it to adapt to new states encountered during learning across multiple batches of experiments. The research was announced as a new paper on arXiv, a preprint server, and is expected to contribute to more efficient and accurate laser cutting processes in the manufacturing of optical films.