Robust Multisensor Localization Using Sparse Landmarks for Autonomous Driving

The advent of autonomous driving technology highlights the crucial need for reliable and real-time localization, which is essential for ensuring vehicle safety and efficient driving. Techniques relying solely on traditional sensors like GNSS and IMU often present significant challenges in local-ization accuracy. Moreover, LiDAR-based point cloud local-ization methods demand extensive storage and computational resources. To address these issues, this paper introduces a map-based localization system for autonomous vehicles that integrates LiDAR, GNSS, and IMU data within an Extended Kalman Filter framework together with sparse map-based features, offering a more efficient and robust solution for vehicle positioning and navigation. Real-world experiments and comparison results demonstrate the efficiency of our proposed method in urban and suburban environments.