Analysis of Position and State Estimation of Quadruped Robot Dog Based on Invariant Extended Kalman Filter

Abstract: Compared with the state estimation of quadruped robots based on external sensors such as camera and lidar, body can provide high-frequency stable odometer estimation. By analyzing methods legged robot sensor, invariant extended Kalman filter (IEKF) sensor is determined to conduct analysis robot. Through various path tracking experiments in simulation real environment, influence travel speed, distance different steering angles position results was analyzed, IEKF model optimized by compensating angular velocity. Experiments show that within set speed range, after adding velocity compensation, accuracy error dog well controlled less than 1%.