強化学習による2足ロボットの矢状面内での歩行運動獲得

抄録

Recently, a reinforcement learning approach has been proposed as one of the control methods for legged robots. This paper describes walking motion in the sagittal plane of a biped robot with 6-DoFs using reinforcement learning. In this research, we prepared two types of rewards to generate gaits. One is a normal reward based on the robot’s state at each sampling cycle. The other is a walking reward, which is based on the contact state of the feet with the ground. By introducing these rewards, the agent was trained, and as a result, the biped robot realized forward walking for 10 seconds in the sagittal plane.