One major hurdle for the deployment of autonomous vehicles in rural environments is achieving accurate localization in areas with tree-canopied roads or outdated point cloud maps. The presence of limited visibility and high variability renders standalone sensor localization unreliable in such situations. To tackle these issues, this paper presents a sensor fusion-based localization framework that integrates data from GNSS, LiDAR, INS, and vehicle odometry. The proposed approach uses a loosely-coupled Extended Kalman Filter for sensor fusion and a weighted gate approach for accurate state estimations. Compared to a state-of-the-art technique, the proposed method achieves a reduction of around 71% in maximum lateral deviations. This method successfully enables a safe and reliable localization in challenging scenarios that are frequently found in the rural and inter-urban sectors.
Sensor Fusion-Based Localization Framework for Autonomous Vehicles in Rural Forested Environments
2023-09-24
1762185 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch