Improving GNSS Positioning by De-noising Consecutive Correlator Outputs Using Graph Fourier Transform Filtering

Abstract

Navigation and positioning with GNSS are ubiquitous in our daily lives. However, GNSS receivers are vulnerable to irregular incoming signals, so their positioning, navigation, and timing (PNT) performance are usually degraded, especially in urban areas. It is, therefore, essential to enhance the baseband of GNSS receivers to sufficiently adapt to and be robust to volatile environments. Correlator outputs are the elementary products from GNSS baseband signal processing. The navigation and positioning accuracy is highly related to the quality of the correlator outputs, which contain less sophisticated noise sources than the code and carrier errors produced by the traditional tracking loops. In that case, the baseband processor can estimate the actual time of arrival (TOA) more easily. In this work, we directly de-noise the complex correlator outputs in the consecutive time domain to produce more accurate pseudorange measurements in harsh environments. Graph signal processing (GSP) is more extraordinary in alleviating irregular noise power than traditional digital signal processing (DSP). Thus, the GSP is applied to optimize the raw graph signals formed with the correlator outputs varying with the time and code offset. Then, the irregular graph domain is processed with graph Fourier transform (GFT) by exploring the geometry structure of the network of correlator outputs. The proposed GFT filtering method for the consecutive complex correlator outputs is realized in a GNSS software-defined radio (SDR) processing the GPS L1 C/A signals. Then, static GPS intermediate frequency (IF) data are collected in an urban area to test the proposed SDR. The real-world experiments demonstrate that the baseband processing results can be de-noised more efficiently, and the positioning accuracy is improved by 65.3% compared to the traditional algorithm.