RANSAC-Based Fault Detection and Exclusion Algorithm for Single-Difference Tightly Coupled GNSS/INS Integration

There is an urgent need for high-accuracy and high-reliability navigation and positioning in life safety fields such as intelligent transportation and automotive driving, especially in complex urban environments. Although, compared with the GNSS and loosely coupled integration, a tightly coupled GNSS/INS integration can improve the positioning reliability by using raw observations, it still suffers from external challenging environments such as the multipath effect. Therefore, the fault detection algorithm is a premise and guarantee to realize quality control of GNSS/INS integration. Inspired by the application of the random sample consensus (RANSAC) algorithm in GNSS fault detection, this paper proposes a RANSAC-based fault detection and exclusion algorithm for single-difference tightly coupled GNSS/INS integration. Here, a between-receiver single-difference (BRSD) model was designed to prevent the consumption of GNSS observations and reduce the waste of effective parameters, and the global proportion statistics of faults were introduced into the typical RANSAC algorithm to further ensure detection reliability. In this study, the effect of the main parameters on the proposed detection algorithm was analyzed and verified by artificial cycle slips. Multiple filed tests, including typical urban scenarios, were conducted to verify the feasibility and effectiveness of the proposed method. The comprehensive test results show that the north and east positioning accuracy in terms of cumulative distribution function (CDF, CDF = 95%) are improved by 45% and 42% over the tightly coupled mode without the proposed detection method.