Robust GNSS Estimation using Factor Graphs, Modified Gaussian Mixture Model and a Transformed Domain Method

Xin Zhang, Min Xu, Zhenjun Zhang

Abstract:	Jamming, Non-line-of-sight (NLOS) signals such as multipath, and other sources causing outliers during the measurement process lead to multi-modal behavior of the measurement system and hence non-linearity. Strong nonlinearity undermines convergence and accuracy of a GNSS estimator. In this work, a smoother is proposed for the classic GNSS positioning problem, by extending the legacy factor graph representation and graph-based nonlinear optimization. Two state-of-the-art methods are combined to ensure a robust estimation of GNSS antenna phase center. In the baseband stage, digital intermediate frequency (IF) samples are processed with robust cost functions; both frequency domain and time domain approaches are evaluated. In the measurement stage, an adaptive version of max-Gaussian mixture model, batch covariance estimation (BCE), was adopted to accurately estimate the multimodal behaviors of the measurement. Experimental results with real-world IF samples show that by combining these two classes of methods, GNSS position fixes are restored decently under heavy jamming conditions. Further insights are provided about the pros and cons of the two robust functions in baseband processing, and the pros and cons of different measurement modeling techniques compared with the adopted BCE approach.
Published in:	Proceedings of the 2020 International Technical Meeting of The Institute of Navigation January 21 - 24, 2020 Hyatt Regency Mission Bay San Diego, California
Pages:	745 - 749
Cite this article:	Zhang, Xin, Xu, Min, Zhang, Zhenjun, "Robust GNSS Estimation using Factor Graphs, Modified Gaussian Mixture Model and a Transformed Domain Method," Proceedings of the 2020 International Technical Meeting of The Institute of Navigation, San Diego, California, January 2020, pp. 745-749. https://doi.org/10.33012/2020.17175
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In