An improved nonlinear filter based on adaptive fading factor applied in alignment of SINS

This paper investigates the non-linear initial alignment for strapdown inertial navigation system (SINS) with main focus on improving the robustness of alignment filter. Conventional Kalman filter (KF) assumes that the statistic characteristics of the system noise are known in advance and keep unchanged during the filtering process. However, it is difficult to predict the noise characteristics in practice, which may cause the degradation in filter performance. In view of this problem, improved fading unscented Kalman filter (UKF) is proposed. The square of the Mahalanobis distance of the innovation vector, which is found to be chi-square distributed, is used as the judging index. Hypothesis test is performed to test the filter state. If the null hypothesis should be rejected, it means that the abnormal noises exist in the system model, and the fading factors should be introduced to rescale the covariance of the innovation vector. The multiple fading factors are calculated by forcing the estimated value of innovation sequence covariance to be equal to its nominal value. Simulation and experiment results show that, the new alignment algorithm performs better in terms of robustness and convergence in the condition of complex measurement noise.