Center of mass estimator for humanoids and its application in modelling error compensation, fall detection and prevention

Abstract

We introduce a center of mass estimator, and its application in full body inverse dynamics control, fall detection and fall prevention for humanoid robots. We use the Linear Inverted Pendulum Model dynamics with an offset to predict the center of mass motion. This offset can be interpreted as a modelling error on the center of mass position, or an external force exerted on the center of mass of the robot, or a combination of both. The center of mass estimator is implemented on our Atlas humanoid robot. During the DARPA Robotics Challenge Finals, it successfully saved our robot from falling on two occasions, and made us the only competitive team without a fall or human physical intervention among all humanoid teams.