Residual Clustering Analysis-Based Fault Detection and Exclusion for GNSS/INS Tightly Coupled Integration

Abstract

In tightly coupled integrated navigation involving the inertial navigation system (INS) and the global navigation satellite system (GNSS), the fault detection and exclusion (FDE) method can effectively ensure navigation accuracy by detecting faults in the residual domain. However, in challenging scenarios, the FDE method will degrade rapidly due to error accumulation, measurement faults, and receiver clock bias instability. To address this issue, a novel FDE method based on residual clustering analysis is proposed. By constructing the Euclidean distance matrix (EDM) of measurement residuals, the cluster density of measurement residuals is calculated, eliminating the common bias caused by inaccurate estimation of receiver clock or unmodeled states. Furthermore, to enhance the robustness of the proposed method, the fault detection threshold is dynamically modified according to cluster density and position error covariance, and the degradation of the FDE caused by position error is suppressed. Experimental demonstrations confirm the effectiveness and advantages of the proposed method. Compared with conventional FDE methods, simulation results show that the mean error of horizontal position decreased by 82.25%–83.49% in challenging environments. In field tests, the mean error of horizontal position was reduced by 29.61%–47.87% and 17.66%–64.94%, respectively.