GNSS/Map Integration with Batch Factor Graph Optimization for Reliable Trajectory Estimation in Urban Canyon

Global Navigation Satellite System (GNSS) localization in urban environments has long been challenged by multipath effects and interference from non-line-of-sight (NLOS) signals. Many studies introducing additional information (e.g., inertial navigators or light detection and ranging LiDAR) typically rely on a stable and reliable GNSS solution for state initialization or relative positioning. However, these methods require costly and complex system integration. In contrast, integrating map information for matching is a reliable and one of the cheapest solutions to achieve positioning errors of less than ten meters in deep urban environments. However, the accuracy of the initial solution significantly impacts the accuracy of the map-matching algorithm. In addition, vehicles may experience multiple states (e.g., turning or crossing lanes) while driving, leading to map-matching errors. Therefore, fully utilizing the potential of GNSS in urban environments is an urgent problem. This study proposes a method that integrates real-time kinematic (RTK) positioning algorithms and reliable meter-level positioning map-matching results into an advanced factor graph optimization (FGO) framework. Experimental results show that our method improves the average accuracy by 20% compared with the traditional method, demonstrating the significant role of integrated RTK and high-quality a priori solutions for nonlinear system optimization.