Session B2a: Advancements in Navigation Algorithms

Multipath Mitigation in Direct Position Estimation: Sparse Estimation and Machine Learning Assistance
Yuze Duan, Zuping Tang, Jiaolong Wei, Jie Sun, Huazhong University of Science and Technology
Location: Beacon B

Direct Position Estimation (DPE) has proven to be a robust technique for Global Navigation Satellite Systems (GNSS), delivering superior performance by fully leveraging the restricted relationships among visible satellites. However, multipath remains a significant source of error in constrained environments, resulting in biased measurements and reduced position accuracy. The performance advantage of DPE over traditional two-step positioning often diminishes, as most satellite signals suffer from multipath and non-line-of-sight (NLOS) reception. Research has demonstrated that, in urban environments, signal visibility can be determined using methods based on the receiver's position and environmental characteristics, allowing for corrections to both multipath effects and NLOS errors. This suggests that multipath errors can be modeled as a function of the receiver's position, which aligns well with the basis principles of DPE. Building on this concept, this paper formulates the DPE method under a multipath signal model and introduces a novel algorithm based on sparse estimation theory. The proposed DPE framework integrates a reweighted-$\ell_1$ regularization term to promote the sparsity of multipath errors. This design facilitates the incorporation of a priori information from machine learning-based multipath detection, optimizing the construction of the weighting matrix with regularization. Finally, we provide a concise analysis of the impact of NLOS on DPE performance, deriving the corresponding Cramer-Rao bounds for both prior knowledge-aided and unaided scenarios, offering simplified expressions. Experimental validation using real GNSS data demonstrates the effectiveness of the proposed method, showing superior performance in urban environments compared to methods that utilize satellite carrier-to-noise ratios and elevation angles for weighting.