This paper explores the application of reinforcement learning (RL) methods for process control in semiconductor manufacturing, presenting a new RL-based controller that shows significant advantages over traditional run-to-run controllers in dealing with both linear and nonlinear processes.