ABSTRACT
This chapter describes how we make humanoid robots robustly perform skillful full-body motor tasks under complex environments and limited resources. Starting from the foundation of full-body humanoid dynamics, we introduce dynamics-based
controllers, then discuss how to make the algorithms realistic and practical. As the key motor function of humanoids, we mainly focus on dynamic bipedal balancing. Basically we utilize wide knowledge from naive robotics and control theory, then introduce some neuroscientific clues to improve the control performance on real robots. This is not at all a theory-oriented study, but a realistic experimental approach. Sufficient experimental results on a life-size hydraulic humanoid robot are attached.
