GNSS observation-domain fault detection and exclusion with prior information correlation: application to RTK

Abstract In Global Navigation Satellite System (GNSS) high-precision positioning, the identification and handling of anomalous observations are essential to ensure positioning accuracy and reliability. However, existing fault detection and exclusion (FDE) methods in the observation-domain often ignore the correlation between observations and lack constraints in the iterative processing, which easily leads to state divergence. Therefore, the paper proposes an observation-domain FDE method that considers prior information correlations. Monte Carlo simulation experiments demonstrate that the improved method exhibits a stronger fault identification ability under conditions of highly correlated observations. The real-time kinematic positioning (RTK) field experiments for landslide monitoring further validate its effectiveness. Compared to traditional methods, the proposed approach reduces the proportion of epochs affected by multi-fault modes from 56% to 17% and achieves a 100% FDE success rate. It also mitigates solution jumps and divergence, reducing float solution 3D errors from 6.189 m to 0.147 m. In fixed solutions, it achieves centimeter-level accuracy and increases the ambiguity-fixed rate to 93.6%. Additionally, it delivers 100% availability and continuity in both horizontal and vertical directions. It is worth mentioning that the improved method enhances performance without significantly compromising computational efficiency, thereby demonstrating important practical application value.