Localization for car-like vehicles using loosely coupled dual-GNSS integration

Abstract This paper explores an approach to global localization for ground autonomous systems using factor graph optimization to integrate a camera, an IMU, and a dual-GNSS setup. While autonomous driving typically relies on high-cost sensors, we propose an accessible alternative solution that utilizes standard-precision GNSS receivers. A dual-GNSS configuration facilitates this shift by providing absolute positioning and reliable heading information even at rest, a robust approach already proven in maritime and UAV applications. Challenges include GNSS signal degradation and high IMU noise, raising the question of whether effective pose estimation could be achieved with loosely coupled sensor fusion. We demonstrate that, when GNSS receiver noise is sufficiently low, a second GNSS receiver enhances localization accuracy within a practical baseline and improves orientation accuracy and initialization robustness in the proposed localization framework.