Robust Extrinsic Calibration for Arbitrarily Configured Dual 3D Lidars using a Single Planar Board

Many robotic and mobile vehicles have widely used multiple lidars to perceive surrounding circumstances. In such systems, a precise extrinsic calibration is essential to seamlessly utilize all data from different sensors. Most of the studies on the multi-lidar calibration use a target with known dimensions, need additional sensors, and rely on tailored initialization steps by users. Furthermore, because some of their studies assume that two lidars are located on parallel planes, those methods are hard to be generalized for rotated and complex sensor configurations. In this paper, we propose a dual lidar extrinsic calibration algorithm that is accurate and robust to arbitrary configurations using a single planar board with reflective tapes. By sensing three different postures of the target, the proposed algorithm geometrically calculates the transformation of dual lidars. Due to no need for accurate dimensions of the target and user-intervention for the target, the proposed method is more practically feasible than the previous methods. To evaluate the proposed method, we set three different configurations and obtain a dataset using two 16-channel 3D lidar sensors. Experiments with real data are demonstrated that the proposed method is accurate and robust.