Abstract: In the control applications based on reinforcement learning, profound professional knowledge and engineering experience are required to set the reward function manually. Therefore, to reduce ...