Hybrid control algorithm for humanoid robots walking based on episodic reinforcement learning

This paper presents a hybrid dynamic control approach to acquire biped walking of humanoid robots focussed on policy gradient episodic reinforcement learning with fuzzy evaluative feedback. The proposed structure of controller involves two feedback loops: conventional computed torque controller and episodic reinforcement learning controller. Reinforcement learning part includes fuzzy information about Zero-Moment Point errors. To demonstrate the effectiveness of our method, simulation tests using middle-size 36 DOFs humanoid robot MEXONE are performed.