Consideration on robotic giant-swing motion generated by reinforcement learning