Motion Design for Humanoids Based on Principal Component Analysis: Application to Human-Inspired Falling Motion Control

This paper proposes a motion design method for humanoids to imitate human-inspired motions by iteratively modifying joint reference trajectories based on the principal component analysis. The objective is to establish a practical method to realize dynamic human motions in humanoids without using complicated equations of motion. In the proposed method, captured motions of a human are firstly projected to a joint angle space of a humanoid by inverse kinematics. These joint angle trajectories are commanded to the servo motors as a function of orientation angles such a pitch angle of the humanoid, not as a function of time. Thus, the reference trajectories are commanded in a faster timing to synchronize the joint angles with the actual orientations, and this allows the humanoid to compensate the error due to the difference between the human and the humanoid in size. Furthermore, a method to iteratively modify the reference trajectories depending on resulting humanoid's motion is proposed. To extract the feature of motion, the principal component analysis is utilized. In the experiments, several motions of a human in fall were analyzed from a viewpoint of impact to the human body, and a forward-rolling motion was selected to be imitated to the humanoid. According to the proposed procedure, the forward-rolling motion was iteratively implemented to the humanoid. The experimental results showed that the humanoid achieved three quarters of the full rotation and that the proposed method could be utilized as a basis to introduce an iterative learning control to completely imitate the human-inspired forward-rolling motion.