The Method of Data Fault Detection of the Integrated Navigation System Based on AR Measurement Modling

Abstract

Aiming to improve the reliability of the SINS/GPS integrated navigation system, considering that the traditional residual chi-square detection method is not efficient in detecting small mutations and slow mutations. An information fault detection method for integrated navigation system based on AR measurement modeling is proposed. This method establishes an AR model of the measured data under no-fault conditions, and combines the Kalman filter model to obtain the measured predicted value for residual calculation. It avoids that the traditional residual chi-square detection method introduces observation pollution data, which leads to the failure of detection statistics tracking, is insensitive to slowly changing failures, and reduces the impact of alarm loss rate on data reliability. The simulation experiment results show that the application of this method in the SINS/GPS integrated navigation system can effectively improve the fault tolerance of the integrated navigation system against slow changes or even small-scale faults, which verifies the effectiveness of the algorithm.