Motion planning for a dynamically-coupled hyper-dynamic manipulator by reinforcement learning

This paper proposes a new motion planning method for a hyper-dynamic robot which aims at realizing the motion control skills exhibited by professional golfers by a dexterous robot. The robot has similar distribution of actuators´ capability to that of human, which is compact but requires dynamically-coupled driving for high speed motion. For such a robot with strong coupling and nonlinearity, to realize a motion including multi-boundary conditions, motion planning is very difficult. As a new method for the motion planning, the motion planning method using reinforcement learning is proposed in this paper. Simulation results show the effectiveness of the proposed motion planning method.