Objectives With the development of the global navigation satellite system (GNSS), the number of visible satellites has increased, and the constellation configuration has been improved. While improving user observation information, multi-system GNSS positioning increases the risk of multiple gross errors, posing a threat to the integrity of the system and restricting the application of GNSS in complex environments. A common method, the receiver autonomous integrity monitoring (RAIM) to ensure the reliability of positioning based on single error, but the reliability of positioning would decline in the case of multiple gross errors. The problem of poor detection and recognition, based on the correlation analysis method of the post-test residual vector, the residual vector is affected by multiple gross errors, showing that the correlation with the observed feature vector of gross errors is weakened. The phenomenon makes gross error detection distortion. The aim of the study is to improving the accuracy of multiple gross errors detection by a new RAIM algorithm based on the density center of observed dataset.
Methods This paper proposes a correlation analysis method based on the density center of the observation data set to realize the detection and identification of multiple gross errors. Firstly, we construct the observation dataset through the QR calibration method. Then, we estimate the density center by the improved Mean Shift model. Finally, we test the correlation distance between the observation feature points and the density center for detection and recognition of multiple gross errors was compared.
Result The gross errors are simulated by the factual observation data, and the correlation distance of density center to gross error satellite and normal satellite, In the case of a single gross error, two gross errors, and three gross errors, the average difference correlation distance are 1.122 m and 1.516 m, 1.021 m and 1.266 m, 1.177 m and 1.588 m respectively. Compared the correlation distance of QR test vector to gross error satellite and normal satellite, the average difference correlation distance are 0.639 m and 1.142 m, 0.497 m and 0.510 m, 0.108 m and 0.198 m respectively.
Conclusions The new RAIM algorithm overcomes the problem of gross error detection distortion caused by the reduced correlation between the calibration vector and the observation vector in the presence of multiple gross errors, which can effectively improve the reliability of multi-GNSS positioning.