Enhance Pose Accuracy of GNSS/INS Integration by Fusing LiDAR Structure Features Based on Continuous-Time State Representation

Accurate and reliable reference poses are a crucial foundation for numerous scientific and engineering applications. This study introduces a two-stage reference pose generation method to produce a more reliable trajectory for large-scale outdoor environments. The first stage is a tightly coupled GNSS/INS integration to generate initial poses, which consists of a forward Extended Kalman filter (EKF) and a backward Rauch–Tung–Streibel smoother (RTSS). In the second stage, raw IMU measurements, line/plane features extracted from LiDAR point clouds, and GNSS RTK solutions, including position and velocity, are fed to an optimization procedure based on continuous-time pose representation to refine the initial poses yielded in the first stage. Two vehicle experiments are conducted to validate the effectiveness of our proposed algorithm. The results demonstrate that our proposed method can significantly reduce position errors in both GNSS partly blocked and challenging environments, with improvements (71.4%, 72.4%, 69.8%) and (67.3%, 68.8%, 63.6%) in the east, north, and up directions in comparison with traditional tightly coupled GNSS/INS integration. Additionally, enhancements in attitude estimation are observed, particularly for the yaw angle.