High-precision positioning in urban environments is still a challenging task for global navigation satellite system (GNSS) as the observations are susceptible to outliers. Receiver autonomous integrity monitoring (RAIM) has been widely used in GNSS positioning to resist the outliers. However, conventional RAIM typically assumes only a single outlier, when multiple outliers are present, the method’s robustness can be adversely affected and a time-consuming iterative processing procedure is often required. To process multiple outliers more robustly and efficiently, we propose a consistency-checking approach based on the concept of random sample consensus (RANSAC). Modifications are made to enhance the RANSAC algorithm, including the use of spatial and temporal continuity of positioning geometry to accelerate RANSAC, and adaptive determination of the inlier-outlier threshold based on observation and parameter uncertainty. The effectiveness and efficiency of the proposed method are validated in GNSS single point positioning (SPP) in both a simulation and a real driving test.
High-precision positioning in urban environments is still a challenging task for global navigation satellite system (GNSS) as the observations are susceptible to outliers. Receiver autonomous integrity monitoring (RAIM) has been widely used in GNSS positioning to resist the outliers. However, conventional RAIM typically assumes only a single outlier, when multiple outliers are present, the method’s robustness can be adversely affected and a time-consuming iterative processing procedure is often required. To process multiple outliers more robustly and efficiently, we propose a consistency-checking approach based on the concept of random sample consensus (RANSAC). Modifications are made to enhance the RANSAC algorithm, including the use of spatial and temporal continuity of positioning geometry to accelerate RANSAC, and adaptive determination of the inlier-outlier threshold based on observation and parameter uncertainty. The effectiveness and efficiency of the proposed method are validated in GNSS single point positioning (SPP) in both a simulation and a real driving test.