Adaptive Robust UKF Algorithm and Its Application in MIMU/GPS Initial Alignment

In MIMU/GPS integrated alignment, the observation of GNSS signals under occlusion and interference contains a lot of gross errors, and the system dynamics model is inaccurate. Traditional adaptive filtering cannot correct model anomalies when gross errors exist in the observation. In this paper, robust factors and adaptive factors are introduced to make appropriate adjustments to the observation noise covariance matrix and state prediction covariance matrix in the Kalman filtering equation. First, the correct equivalent weight factor is constructed to suppress the problem of gross error in observation; Then the prediction residual method is used to construct the adaptive factor and correct the anomaly of the dynamic model in real time. The prediction residual method overcomes the limitation that the observation is less than the state quantity in the conventional robust estimation algorithm. Finally, the algorithm is verified with the measured data. The experimental results show that this method can effectively suppress and eliminate the influence of gross errors in observations and dynamic model anomalies, and then balance the trust of filter estimates on GNSS observations and dynamic models, which can improve the accuracy and reliability of alignment.