SI-LIO: High-Precision Tightly-Coupled LiDAR-Inertial Odometry Via Single-Iteration Invariant Extended Kalman Filter

Cong Zhang,Jie Zhang,Qingchen Liu,Yuanzhou Liu,Jiahu Qin

DOI: https://doi.org/10.1109/lra.2024.3497757

IF: 5.2

2024-01-01

IEEE Robotics and Automation Letters

Abstract:This work focuses on the accuracy of LiDAR-inertial odometry (LIO). We propose a novel high-precision tightly-coupled LIO method, SI-LIO, based on the invariant extended Kalman filter with a single-iteration estimate update. This method utilizes the Lie exponential map between the matrix Lie group SE $_{4}(3)$ and its Lie algebra such that the linear approximations referring to Taylor expansions are only conducted on the orientation estimation errors in estimation process. Compared to the state-of-the-art filter-based LIO systems, our method effectively reduces the linearization errors, which enhances the system accuracy, notably in scenarios accompanied by larger prediction errors. Besides, the single-iteration update nature endows SI-LIO with a shorter estimation time than FAST-LIO2. Experiments conducted on public datasets demonstrate the higher accuracy of SI-LIO. Specifically, the estimation accuracy of SI-LIO outperforms FAST-LIO2 by approximately 15% when they both employ a single-iteration update. The source code of our LIO system are available at https://github.com/USTC-AIS-Lab/SI-LIO.git .