Falling motion control is necessary because humanoid robots are vulnerable to falling. This issue has been the subject of several previous studies; the contributions of this paper are a motion selection method and a method for generating fall-avoidance motions and active shock reducing motions. We propose a method using the predicted ZMP to evaluate the risk of falling. Based on the predicted ZMP, the humanoid robot can detect the fall and decide which motion to take based on this index. If the predicted ZMP departs from support polygon, then the falling risk is high, so the robot should use fall-avoidance motions or reduce the damage by active shock-reducing motions. An active shock-reducing motion must be able to respond to varied initial conditions to reduce impact damage to the robot. We propose a shock-reducing motion using a dynamical 3D-symmetrization method by which the COG trajectory can be generated online and the vertical velocity can be specified. We present an experiment using a dynamics simulation, which verifies the method of shock-reducing motion generation. In another experiment using a real robot, we test our method of using the predicted ZMP to select motions. These experiments confirm the effectiveness of the techniques.
展开▼
