Reinforcement learning for process control with application in semiconductor manufacturing

Process control is widely discussed in the manufacturing process, especially semiconductor manufacturing. Due to unavoidable disturbances manufacturing, different process controllers are proposed realize variation reduction. Since Reinforcement Learning (RL) has shown great advantages learning actions from interactions with a dynamic system, we introduce RL methods for and propose new controller called RL-based controller. Considering fact that most existing run-to-run (R2R) mainly rely on linear model assumption input–output relationship, first discuss theoretical properties of based assumption. Then performance traditional R2R (e.g., Exponentially Weighted Moving Average (EWMA), double EWMA, adaptive general harmonic rule controllers) compared processes. Furthermore, find have potential deal other complicated nonlinear The intensive numerical studies validate controllers.