Enhanced Autonomous Vehicle Localization via A Priori Maps

This paper investigates a cost-effective approach to autonomous vehicle localization by leveraging lightweight, <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$a priori$</tex> discrete landmark maps as a fallback method during GNSS outages. Specifically, we fuse odometry, inertial, and discrete landmarks from the nuScenes dataset within a Particle Filter, explicitly modeling both map and sensor uncertainties. Compared to high-maintenance High definition (HD) maps and computation-heavy Simultaneous Localization and Mapping (SLAM), this technique significantly constrains vehicle motion and reduces drift, even though the underlying maps are coarser. A key contribution lies in examining the role of landmark availability and distribution: while adding more landmarks generally improves accuracy, it also increases computation time and can destabilize the filter if the density becomes too high. Thus, a balance must be struck between richer landmark information and practical runtime constraints. Overall, the findings highlight that readily available discrete landmark maps, when fused intelligently, provide scalable and maintainable localization support-particularly in environments where GNSS signals are frequently unreliable.