An Adaptive Overlap-based ICP Algorithm for Multi-LiDAR Calibration in Low-overlap Situations

LiDAR has been widely used as a core sensor in autonomous driving and mobile robotics. Due to the high price of spinning LiDARs, the industry always uses a combination of multiple calibrated solid-state LiDARs. However, there is usually only a small overlap FoV between LiDARs on account of their mounting positions. In such a low-overlap situation, conventional calibration methods always bring catastrophic results. To solve this problem, this paper proposed an adaptive overlap-based ICP algorithm, which can calibrate both 2D and 3D LiDARs. The algorithm adaptively extracts the overlapping point clouds by identifying the correct matching results and constructs an optimized ICP alignment. By controlling the radius range of the point cloud search, our method allows for quick extrinsic calibration of two LiDARs with low-overlap FoV. Experiments show that our algorithm achieves a more robust and accurate low-overlap registration than other methods, and also has speed advantages in general situations.