Abstract: A new approach for nonlinear filtering systems is presented in this paper to meet the requirements both on accuracy and computing time. This method, called as approximate second-order extended Kalman filter (AS-EKF), is based on the frame of recursive linear minimum variance estimation. Other than extended Kalman filter (EKF) linearizing all nonlinear models, the new approach estimates the expectation value to the second-order of accuracy. We show that this technique is more accurate than EKF, and it also costs the less computing time than unscented Kalman filter (UKF).It can be used in the nonlinear filtering concerned with both accuracy and computing time.

Key words: nonlinear filtering, extended Kalman filter, state estimation