Falling motion control for humanoid robots while walking

Humanoid robots are prone to fall caused by disturbances. If the disturbance is weak, a humanoid robot can avoid falling by using feedback control. If the disturbance is strong, humanoid robots can perform an Ukemi motion: an active shock-reducing motion. This research proposes a technique for selecting an optimal strategy to handle disturbances while walking, and describes a method of generating Ukemi motion. This technique detects disturbances using sensor data while walking steadily. Moreover, an experiment with a real robot was performed where fall detection was computed using discriminant analysis of walking data labelled as fall and non-fall. Furthermore, this study uses the three-dimensional linear inverted pendulum mode (3D-LIPM), which has been utilized in gait generation of humanoid robots, to generate a solution for the center of gravity for the contracting/expanding fall correction movement. The effectiveness of the proposed technique was confirmed by the verification experiments.