State estimation of a heavy-duty hexapod robot with passive compliant ankles based on the leg kinematics and IMU data fusion

Abstract

To control autonomous walking of a legged robot, it is essential to obtain the instantaneous velocity and posture of the robot. This paper presents a full body state estimation algorithm for a hexapod robot to estimate the velocity and posture of the trunk body without geometric knowledge of the environment. The velocity and posture estimates of the hexapod robot with passive compliant ankles are further processed by a data fusion method that is proposed based on the extended Kalman filter (EKF) technique, utilizing the leg kinematics model of the robot and the readouts from an on–board inertial measurement unit (IMU). The absolute footholds of the hexapod robot are estimated together with the velocity and posture of the trunk body, with consideration of intermittent ground contacts. Experiments have been conducted on both flat and uneven terrains, and the results have confirmed the effectiveness of the proposed approach.