LAUSR

Sensor calibration is a fundamental step for improving the performance in sensor fusion, the aim of which is to spatially and temporally register sensors with respect to each other. This paper presents a high-accuracy autocalibration method to estimate extrinsic parameters between LiDAR and an IMU. LiDAR/IMU calibration is a challenging task since the raw measurements are distorted, biased, noisy, and asynchronous. Our calibration approach adopts continuous-time trajectory estimation wherein the IMU trajectory is modeled by Gaussian process(GP) regression with respect to the independent sampling timestamps. Accordingly, the distorted and delayed LiDAR points sampled at discrete timestamps can be analytically modeled in on-manifold batch optimization. To efficiently and accurately associate laser points with stable environmental objects, the method is carried out in known environments with a point map that is segmented as structured planes and managed by a specially designed octree map. We thoroughly investigated factors relevant to the calibration accuracy and evaluated the performance of the proposed method using both simulated and real-world datasets. The results demonstrate that the accuracy and robustness of our calibration approach are sufficient for most applications.