Joint spatial, polarization, and temporal estimation based on multiple sparse Bayesian learning in GNSS multipath environments

Global Navigation Satellite System (GNSS) suffers from the multipath signals reflected by various objects in the vicinity of receivers. A severe multipath environment may enormously hamper the tracking performance, resulting in meter-level pseudorange error. To solve the parameter estimation (angle, polarization, and time delay) problem and enhance multipath mitigation, particularly in the presence of high spatial and temporal correlation between the line-of-sight signal and multipath signals, we propose a novel method based on multi-dimensional processing and multiple sparse Bayesian learning (MSBL). Our method establishes a joint spatial, polarization, and temporal GNSS signal model and utilizes sparsity in the spatial and temporal domains, as well as signal characteristics in the polarization domain, to solve the multi-dimensional processing problem. We then derive an MSBL-based joint estimator of angle, polarization, and time delay for each signal, extending an off-grid estimator for angles and time delays to reduce complexity and improve resolution. Simulation results demonstrate that our approach achieves close-to-optimal performance compared to the Cramér-Rao bound and outperforms other methods, particularly under highly spatially or temporally correlated signals and with multiple multipath signals. These results show that our proposed method has excellent robustness and effectiveness in combating multipath.