Abstract
In the presence of humans, biped robots must avoid falling because falling down can lead to serious accidents. This paper presents a fall avoidance control method for biped robots under unknown disturbances that is based on real-time trajectory planning. First, a biped robot is modeled as an inverted pendulum. The robot generates in real time the trajectories of its center of mass and center of pressure, taking into account the dynamics and kinematics of the robot and the direction and magnitude of disturbances. By using the planning result, the robot returns to its stationary state after taking a few steps without falling. The experimental results show the effectiveness of the proposed method.
