A tightly coupled integration of GNSS/IMU/LiDAR with parameterized semantic line and plane features to improve pose accuracy in complex environments

Continuous and accurate positioning is one of the critical requirements for established and emerging unmanned systems. Although the GNSS/IMU integration has become a widely-used navigation system, its performance is heavily dominated by GNSS. The dramatical accumulated error of IMU in GNSS outage and wrong updates results by GNSS outliers will influence the reliability of the integration system. In this work, we use light detection and ranging (LiDAR) to enhance the performance of the existing GNSS/IMU integration, where the raw measurements of three sensors are tightly integrated. The raw measurements of LiDAR are abstracted as parametric line and plane features. Two experiments are conducted to assess the proposed algorithm, and the results show that the addition of LiDAR significantly upgrades pose accuracy. In GNSS-challenge scenarios, LiDAR weakens the influence of GNSS outliers and improves the position accuracy by 77.6%, 67.4%, and 63.2% in the right, forward, and up directions, respectively.