Abstract: Low estimation precision and high loss-of-lock threshold are the two important drawbacks of the extended Kalman filter (EKF) which is the widely used GPS frequency estimation algorithm in high dynamic circumstances, caused by linearizing all nonlinear models. To resolve the problems of EKF, the unscented Kalman filter (UKF) which is a new kind of linear filter is introduced to estimate frequency. Instead of the linearization steps required by the EKF, a series of Sigma points and unscented transform (UT) are used to predict and update the state vector and covariance. Simulations indicate that the estimation precision of UKF is greatly improved compared with that of EKF, that the loss-of-lock threshold is about 1dB lower than that of EKF, and that the estimation performance is improved.

Key words: global positioning system (GPS), high dynamic, extended Kalman filter, unscented Kalman filter