GNSS Positioning Based on 3D LiDAR Perception and Virtual Grid-Based Differential Correction in Urban Environments

Accurate global navigation satellite system (GNSS) positioning in urban canyon environments is severely challenged by non-line-of-sight (NLOS) errors caused by signal blockage and multipath propagation. This paper proposes a novel GNSS localization framework that integrates three-dimensional (3D) light detection and ranging (LiDAR) perception with virtual grid-based differential correction to improve positioning performance for both multi-sensor receivers (core nodes) and surrounding low-cost GNSS devices (base nodes). First, an NLOS identification method is developed by combining LiDAR-based ray tracing with pseudorange innovation clustering, enabling reliable detection of mid- and high-elevation NLOS satellites. Then, a satellite selection strategy based on dilution of precision (DOP) optimization is proposed to balance NLOS error compensation and satellite geometry preservation. For cooperative localization, a dynamic virtual grid is constructed around the core node, and pseudorange observations are simulated to generate spatially correlated reference points, allowing base nodes to perform single-differenced positioning relative to the core node. Experimental results on the UrbanNav dataset demonstrate that the proposed framework significantly improves GNSS positioning accuracy and stability for the core node in complex urban scenarios. Meanwhile, it enhances the localization performance of nearby low-cost terminals through real-time environmental information sharing.