Extended Kalman Filter with Adaptive Measurement Noise Characteristics for Position Estimation of an Autonomous Vehicle

This paper proposes the position estimation method of an autonomous vehicle on flat terrain, which based on playback navigation algorithm. The proposed method is sensor fusion using the extended Kalman filter (EKF) for state estimation from the low-cost global positioning system (GPS) receiver and incremental encoder. The singular value decomposition (SVD) is applied to evaluate the adaptive measurement noise covariance in the EKF. This improves the accuracy of estimation to correspond to the errors involved along various portions of the trajectory, instead of using fixed values. The result showed that the proposed method can improve an accuracy of position estimation of autonomous vehicle on flat terrain.