Abstract: Deep reinforcement learning (DRL) can achieve favorable dynamic performance compared to conventional control methods. However, steady-state errors are often ...