Strong tracking filter modeling for GPS robust navigation

Abstract

Efficient implementation of positioning algorithm plays a crucial role in modern Global Positioning System (GPS). Conventional non-linear Least Squares (LS) method is applied iteratively by Taylor expansion, which doesn't combine different epoch-time for mutual restraint. While with respect to generally used extended Kalman filter (EKF), it requires an accurate system model estimation and exact stochastic Gaussian white noise, resulting in non-uniformly convergence with an unknown bias or model error. To solve this problem, this paper proposes a Strong Tracking Filter (STF) modeling. To achieve robustness about model uncertainty and tracking capability on the mutation status, the STF adjusts the real-time state prediction error covariance matrix and the corresponding gain matrix. It also makes use of Time-varying Fading Factor (TVFF) to deal with the past data, which weakens the stale data to the impact of current filtered value. Simulation shows that the proposed STF is capable of enhancing more than half precision than traditional EKF and LS.