Targetless Extrinsic Calibration of Multiple LiDARs Based on Pose Graph Optimization

The integration of multiple LiDAR systems enhances the perception capabilities of autonomous vehicles; however, it also introduces challenges in extrinsic calibration. The precise and convenient extrinsic calibration is crucial for the fusion of multiple LiDARs. To address this issue, this study presents a targetless extrinsic calibration method for multi-LiDAR extrinsic parameters based on pose graph optimization. This method eliminates the need for specific scenes, target and precise initial values, making it more user-friendly for multi-LiDAR. Firstly, LiDAR odometry is used to estimate the position of each LiDAR, and hand-eye calibration is employed to recover the initial extrinsic between LiDARs. Subsequently, multiple submaps are constructed based on the pose sequence estimated by the LiDAR odometry, and the relative transformations between pairs of LiDARs are estimated through submap-submap matching. Finally, the relative transformations between LiDARs to be calibrated are used as constraint factors to build a pose graph model and optimize the extrinsic results between the main LiDAR and the LiDAR to be calibrated. The experiments conducted on our platform validate the robustness, accuracy, and practicality of the proposed multi-LiDAR calibration method.