Adaptive fusion filtering algorithm and its application for INS/GPS integrated navigation system

A new adaptive fusion filtering (AFF) algorithm based on interactive multiple models (IMM) is put forward to solve problems of bad robustness and low accuracy, existing in extended Kalman filter (EKF) when the system model includes uncertainties. In the IMM-AFF algorithm, the system structure is described by two models, and a Sage-Husa filter corresponding to the one and a strong tracking filter (STF) corresponding to another work in parallel independently. The state estimation of system is the weighted fusion of the two filters by using model probabilities, so that the merits of Sage-Husa filter and STF are combined and their demerits are overcome through AFF. Consequently, the proposed IMM-AFF algorithm shows robustness against model uncertainties and high state estimation accuracy. This fusion filter is applied in an INS/GPS integrated navigation system. Furthermore, simulation results under various error environments show that IMM-AFF algorithm is superior to EKF in estimation accuracy and robustness, especially positioning accuracy.