A Machine Learning Approach To A Lateral Continuous Force Estimation For A Walking Biped Robot

In this paper, a regressor for determining a lateral external continuous force applied upon a walking biped robot is investigated and verified by a numerical simulation. A pre-defined walking gait of a biped robot is constructed by the Tchebyshev method. And a continuous force-action classifier is generated. It determines whether the lateral external force is a continuous force or not. A regressor which estimates a lateral external continuous force acted upon a walking biped robot is constructed by SVR (Support Vector Regressor). The regressor is verified by a numerical simulation. We assumed that only lateral force is applied upon the COG (Center of Gravity) of the walking biped robot.