Extrinsic Lidar Calibration in a Workshop Environment

Abstract

Abstract As the automotive industry moves toward an increasingly automated reality and thus is getting more reliant on sensor driven features, the need of accurate calibration methods is getting increasingly critical. To verify reliable performance during the life span of a vehicle, methods to recalibrate sensors during service are developed to accommodate for the limited equipment being present in workshop environments. Proof of concepts for extrinsic calibration of a lidar sensor are developed and evaluated on this premise to investigate how calibration methods can be adapted to work with limited resources. Firstly, a target-based calibration algorithm is modeled to compute the extrinsic parameters of the sensor by computing the rigid transformation between the known position of calibration targets and the estimated position of the same targets seen by the lidar sensor. Secondly an alternative calibration algorithm is implemented that utilizes the internal IMU unit as well as the lidar sensor. The 3D points seen by the lidar sensor are accumulated with some estimated extrinsic parameters over a trajectory. The quality of the accumulated point cloud is then evaluated by comparing how the environment is seen by different scans over time. The algorithm optimizes over different extrinsic parameters using a grid search to find the estimated parameters of the lidar sensor. The target-based method proved accurate when correct target position was found but is highly sensitive to poor target localization. Lidar-to-IMU calibration shows promising results in simulation with a high accuracy, few requirements on the environment and no additional equipment needed.