Planning Whole-body Humanoid Locomotion, Reaching, and Manipulation

In this chapter we address the planning problem of whole-body motions by humanoid robots. The approach presented benefits from two cutting edge recent advancements in robotics: powerful probabilistic geometric and kinematic motion planning and advanced dynamic motion control for humanoids. First, we introduce a two-stage approach that combines these two techniques for collision-free simultaneous locomotion and upper-body task. Then a whole-body motion generation method is presented for reaching, including steps based on generalized inverse kinematics. The third example is planning of whole-body manipulation of a large object by "pivoting", by making use of the precedent results. Finally, an integrated experiment is shown in which the humanoid robot interacts with its environment through perception. The humanoid robot platform HRP-2 is used as the platform to validate the results.