Constrained Whole Body Motion Planning in Task Configuration and Time

Abstract

Our goal in this research was to develop an efficient motion planning algorithm for a humanoid to enable it to perform complex tasks. To perform the tasks, the humanoid sometimes must be able to generate whole body motion trajectory under the end effector constraints as well as collisions; self-collision and unexpected collision with obstacles in the environment. Also, more complex constraints such as joint limit, joint speed limit would make planning whole body motion more difficult. This constrained redundancy resolution makes the motion planning for humanoid difficult. Although the previous motion planning algorithms are still feasible to find solution, they are typically planning in the high dimensional joint configuration space then finally require high computing. This paper proposes that planning on the specified whole body task space such as pelvis height, chest orientation is able to find feasible solution while satisfying constraints and collision free. We also expect this approach that can be applied on various humanoids so we demonstrated on Atlas and Valkyrie robot. In order to demonstrate the feasibility of the proposed approach, we applied our motion planning algorithm humanoid performing tasks that are proposed by DARPA challenge such as opening a door, closing valve, cutting wall, pulling emergency switch.