Abstract
Dynamics based control methods for humanoid robots have been focused in order to overcome difficulties in generating dynamic motions of complex humanoid robots. In recent years, a visualization method for a "knack" of dynamic motion was proposed as a criteria that would provide better understanding of dynamic motions. However the definition of knack seems to be difficult to apply robotic motor learning method because the definition employs energy state to specify a region existing the knack in a motion. In this research, we focus on motor command in order to specify the knack region. In this paper, we propose a method to figure out the knack region in motor command space and clarify the advantages of the proposed method by using dynamic simulation of a humanoid robot.
