MapCalib: Toward Large-Scale Automatic Calibration for Roadside LiDAR Using High-Definition Map

Abstract. Roadside LiDAR sensors are critical components in automated driving systems and mobile mapping systems. These sensors, typically deployed along roadsides to provide continuous data for perception tasks, require precise calibration to ensure the safety and performance of intelligent connected vehicles. However, large-scale deployment presents new challenges, including low sensor overlap and variability in sensor types, complicating the calibration process. To address these challenges, this study introduces Map- Calib, an innovative method for the automatic calibration of roadside LiDAR systems using High-Definition (HD) map. MapCalib improves calibration efficiency by eliminating the need for specific calibration targets, which simplifies the process and increases safety when managing large-scale roadside LiDAR installations. The method begins with the development of a virtual map projector, which establishes a mapping from the HD map to the LiDAR data, minimizing representation disparities. Next, a Semantic Universal Spatial Context (SUSC) descriptor is proposed to efficiently localize the LiDAR sensor positions within the HD map. Finally, through feature retrieval and iterative optimization, the method calibrates sensor parameters, such as orientation and position. The proposed calibration framework is validated through simulated, public, and self-collected datasets, demonstrating its ability to automatically calibrate multiple LiDAR sensors with high accuracy. Compared to existing calibration methods, MapCalib achieves a notable improvement of 39.9% in Relative Rotation Error (RRE) and 39.3% in Relative Translation Error (RTE).