A hierarchical framework for realizing dynamically-stable motions of humanoid robot in obstacle-cluttered environments

We address in this paper a hierarchical framework for planning and simulating dynamic motions of humanoid robots in cluttered environments. The robot is aimed to realize a dynamic multi-step motion via a series of support (contact or grasp) configurations. In this framework, a global CoM (Center of Mass) trajectory is firstly generated which ensures the balance during the motion; the whole-body collision-free motion planning is then carried out locally for each transition phase between support configurations; finally the generated trajectories (CoM, end-effectors, joints) serve as control references for realizing the dynamic motion. This framework has been tested in a car-ingress scenario.