Lidar sensors are an essential sensor for autonomous driving and advanced driver assistance systems. This paper proposes an accurate method for an online and unsupervised estimation of the yaw-mounting error of lidar sensors. A precise estimation is essential, when the extracted information of the lidar, like objects or obstacles, are fused with other sensors or interpreted in a road context, like lane assignment. The approach is based on a grid map, which is a discretized representation of the environment around the ego vehicle and enables an efficient processing of the large number of lidar detections. The proposed algorithm extracts features out of grid maps of two measurement cycles and reconstructs the ego movement. The comparison with the vehicle odometry enables the reconstruction of the lidar misalignment. By combining a large number of measurement cycles, a robust and precise estimation of the lidar misalignment is obtained. Features from the image domain could be directly applied, enabling an efficient and real-time calibration in a productive system. The system identifies itself valid driving sections and performs a continuous recalibration. Simulation and real experiments show a promising accuracy with an orientation error below 0.05° for roof-top lidars and 0.1° for low-cost lidars.
Online and Unsupervised Lidar Calibration based on Grid Maps
2020-09-20
404064 byte
Conference paper
Electronic Resource
English
IMU-based Online Multi-lidar Calibration
IEEE | 2024
|Lidar to camera calibration for generating high definition maps
European Patent Office | 2021
|Lidar to camera calibration for generating high definition maps
European Patent Office | 2020
|Online Orientation Prior For Dynamic Grid-Maps
IEEE | 2021
|