Biped robot state estimation using compliant inverted pendulum model

Abstract

This study proposes a biped robot state estimation framework based on a compliant inverted pendulum model and a robust state estimator. A proper model that can express the key physical characteristics while considering limited computing power should be defined for the biped robot state estimation. A biped robot’s limited structural stiffness and relatively long legs compared with the cross section of the body lead to undesired flexibility. However, the models used in previous research are either not suitable for state estimation or too simple to express the essential characteristics of the biped robot. A compliant inverted pendulum model is adopted herein to enhance the estimation accuracy. This model is made by adding a virtual spring and a damper to the conventional inverted pendulum. The additional elements represent the mechanical deformation and the undesired flexible movement. Adopting this model makes it possible to reflect the important characteristics of the biped robot while taking advantage of the merits of the single-mass model. In addition, a robust state estimator that we previously proposed is adopted to compensate for the estimation error caused by the modeling error. Using these two factors, the improved COM-kinematics estimate is obtained with respect to the existing simple-model-based biped state estimators.