State estimation for quadrupedal using linear inverted pendulum model

Abstract

This paper presents an estimator for quadruped robot to obtain state parameters, which is a challenge issue, on account of a variety of intrinsic sensor noise and force disturbances. Based on this issue, we exploit a Linear Inverted Pendulum Model (LIPM) to estimate the state of the center of mass (CoM), simultaneously considering the external force disturbance. Meanwhile, Extended Kalman Filter (EKF) is put to use through fusing the information from forward kinematics. To validate the feasibility of the proposed method, it is implemented on a quadruped platform and we carry out a series of experiments. The results of data analysis demonstrate the performance of the proposed estimator via comparing with the actual data from the motion capture system and force platform. The average error of velocity estimation is less than 0.04m/s, and the local positon estimation error is less than 6%, which are within the control tolerance.