Motion separating based whole body motion planning for humanoid robots using a gradient descent method

Abstract

This paper introduces whole body motion planning for humanoid robots that is fundamentally important for functioning of humanoid robots within the living space of humans. Humanoid robots have multiple degrees of freedom (DOF). Because of this multi-DOF structure, whole body motion planning of humanoid robots is difficult and complex. In this paper, we separated whole body motion of humanoid robots into several partial motions. Separating whole body motion into several partial motions, we can overcome the difficulty and complexity of whole body motion planning for humanoid robots with high degrees of freedom. Each separated partial motion has only 6 DOF: X, Y, Z, Pitch, Yaw, and Roll. Based on these low-degrees of freedom, we can utilize efficient motion planning algorithm. As an algorithm of partial motion planning, the gradient-descent method was used. The correctness and efficiency of our algorithm was established by conducting simulations and experiments on the humanoid robot simulator and physical humanoid robot using the whole body motion planning algorithm developed in this research.