Abstract
It is nesessary for humanoid robots to stand and walk in the presence of external the disturbances. Humans usually switch the control strategies depending on disturbances: stabilization in the upright position, and falling avoidance using stepping motion. In order to achieve robustness to unknown disturbances, humanoid robots require such a switching of control strategies. For this purpose, it is important to explicitly consider the physical constraint in the control law. In this paper, the author proposes a method to detect the falling and generate the stepping motion to avoid falling, which is based on the maximal output admissible set on the COG-ZMP inverted pendulum model. The desired landing position and time are calculated based on the distance between COG state and the maximal output admissible set. The effectiveness of the proposed method is verified with simulations.
