GNSS multipath correction based on LightGBM

As a primary error source of GNSS precise positioning in complex environments, multipath error can adversely limit the positioning accuracy. This paper proposes a phase multipath correction method based on the light gradient boosting machine (LightGBM). The method adopts satellite elevation angle, azimuth angle, and carrier-to-noise ratio (CNR) as input features, uses the static post-processed single-difference phase residual as the target value, and then be applied to subsequent multipath error correction after training. To validate the effectiveness of the proposed algorithm, experiments were conducted in a multipath-strong environment. The results demonstrate that compared with traditional positioning method that ignores multipath error, after applying LightGBM-based multipath correction, the positioning accuracy is improved by 55.8%, 47.0%, and 44.1% in the north (N), east (E), and up (U) directions, respectively, and the ambiguity fixing rate is increased from 29.4% to 68.4%. It is concluded that the nonlinear correlation between multipath errors and the selected data features can be effectively captured by the LightGBM model, which provides a feasible solution for high-precision GNSS positioning applications in strong multipath environments.