強化学習を用いたヒューマノイドによる持ち上げ運動

摘要

Lift-up is a fundamental behavior for a humanoid robot, which can make various behaviors withmany degrees of freedom (DOF) and by using many sensors. For creating the lift-up action, we have to copewith environmental conditions and robot dynamics of those many DOF and sensors. This paper shows thatby applying Reinforcement Learning, a humanoid robot, I.e., HOAP-1, learns to acquire the lift-up actioneffectivly with the interaction to its own actuators and sensors.