Abstract
The conventional residual chi-square detection method has a low detection rate for slow-changing faults in the integrated navigation system of the Inertial Navigation System/Global Navigation Satellite System (INS/GNSS). To tackle this issue, we propose an improved Levenberg-Marquardt-Gauss-Newton (LM-GN) algorithm to aid the integrated navigation system in detecting slow-changing faults. The algorithm is designed to tackle this issue. When constructing the detection statistics, the measurement data is assessed independently for failures. The paper proposes a dual-threshold detection method. It then divides the measurement data into three parts for classification processing. When the detection data falls within the two thresholds, the trust region technology is used. It improves the iteration speed and accuracy of the LM algorithm and improves the sensitivity of fault detection. An experiment is conducted to validate the proposed method's effectiveness. It tests fault detection in the INS/GNSS integrated navigation system. The experimental results demonstrate that the improved LM-GN algorithm-assisted dual-threshold fault detection reduces the system's missed fault alarm rate by 46.7%. It also improves speed accuracy by 43.9% compared to the traditional dual-threshold residual chi-square detection method. Additionally, the positional accuracy has increased by 16.9%.