Rapid whole-body motion generation using regression of the torso posture of a humanoid robot

Abstract

In daily-life environments, humanoid robots are expected to perform household tasks instead of humans. Humanoid robots can move over a large workspace but generating their motions is still very time consuming. We aim to improve the realtimeness of whole-body motion generation for humanoid robots. In this paper, we introduce a rapid whole-body controller which uses regression of the torso posture. By estimating the torso posture, we reduce the calculation cost of inverse kinematics, as we solve it separately for the legs and arms. We apply the proposed method to the humanoid robot HRP-4 for the task of reaching low-height positions which requires hunkering down. This motion takes 3 seconds.