Real-time outlier detection of satellite orbit and clock products using reverse error estimation

Jiande Huang,Xingxing Li,Xin Li,Jiaqi Wu,Keke Zhang,Yongqiang Yuan,Wei Zhang
DOI: https://doi.org/10.1007/s10291-024-01753-0
2024-10-08
GPS Solutions
Abstract:Real-time precise orbit and clock products are necessary for Global Navigation Satellite System (GNSS) precise positioning to achieve a centimeter-level accuracy. Besides the accuracy of orbit and clock products, the demands for their reliability in safety–critical applications are on the rise. In real-time process, it is challenging to monitor the instantaneous accuracy of precise orbit and clock products due to the absence of reference with equivalent timeliness and higher accuracy. This makes it difficult to conduct outlier detection on real-time orbit and clock products to exclude satellites with abnormal product accuracy. In this study, we develop a reverse error estimation model for monitoring real-time orbit and clock products, using the pseudorange and carrier phase observations from globally distributed stations. Based on the estimated product errors, we generate the outlier detection thresholds for each satellite. The quality information of real-time products can be calculated using the detection thresholds and instantaneous product errors, thereby improving the stability and accuracy of positioning. Real-time products from Centre National d'Etudes Spatiales, as well as one-year observations of 130 stations from the IGS Multi-GNSS experiment network, are processed to evaluate the outlier detection method. The validation results indicate that the centimeter-level time variation and step-change in product errors can be detected, which means that the satellites with abnormal product accuracy can be identified correctly. Considering the quality information, observations from the satellites with abnormal product accuracy are excluded in precise positioning. During the period of product abnormality, the positioning accuracy can be improved from 11.3 cm to 6.6 cm for single system, and from 6.7–8.0 cm to 5.1–5.4 cm for multiple-systems. The results indicate that the exclusion can improve the accuracy of single-system positioning by 40% and multi-system positioning by 20–30%, respectively.
remote sensing
What problem does this paper attempt to address?