Targetless Extrinsic Calibration of LiDAR–IMU System Using Raw GNSS Observations for Vehicle Applications

Abstract

Nowadays, the low-cost MEMS-based Inertial Measurement Units (IMU) and LiDARs are commercially available in vehicle systems and are usually integrated to achieve robust ego-motion estimation due to their excellent complementary properties. In this case, accurate inter-sensor spatial transformation, i.e. extrinsic parameters, is a fundamental prerequisite for the combined application of LiDAR and IMU. However, current LiDAR-IMU calibration methods usually rely on specially-designed artificial targets or facilities, which greatly limits the flexibility and usability of calibration. Fortunately, modern intelligent vehicles are generally equipped with portable global navigation satellite system (GNSS) devices, which are able to provide precise positioning services and could be used to improve the performance of extrinsic calibration. To this end, we innovatively introduce raw GNSS observations to enhance in-vehicle LiDAR-IMU calibration performance without predefined target support. Based on the centralized Extended Kalman Filter (EKF), the estimator tightly fuses the double-differenced pseudorange and carrier phase observations, IMU data, and extracted plane features to obtain the optimal extrinsic parameters. The results of simulated tests and real-world experiments indicate that the introduction of GNSS can significantly improve the calibration performance. The GNSS-aided LiDAR-IMU calibration results also show better odometry accuracy and mapping consistency than those without GNSS aiding.