A High-Precision LiDAR-Inertial Odometry via Invariant Extended Kalman Filtering and Efficient Surfel Mapping

Houzhan Zhang, Rong Xiao, Jiaxin Li, Chuangye Yan, Huajin Tang

Abstract: Simultaneous localization and mapping (SLAM) via light detection and ranging (LiDAR)-inertial odometry is a crucial technology in many automated applications. However, constructing a consistent state estimator with an efficient mapping method still remains a challenge for LiDAR-inertial odometry (LIO) systems. In this article, we propose a tightly coupled LIO system via invariant extended Kalman filter (InEKF) and efficient surfel mapping. First, based on the InEKF theory, we build a consistent state estimator for a tightly coupled LIO system. Second, we propose a novel LIO system by combining the InEKF state estimator with a surfel-based map, named SuIn-LIO, which not only enables the accuracy of state estimation and mapping but also enables real-time registration of a new LiDAR scan. Extensive experiments on different public benchmark datasets demonstrate that SuIn-LIO can achieve comparable performance with other state-of-the-art methods in accuracy and efficiency. To benefit of the community, our implementation will be open-sourced on Github.