IVHC-DLIO: Incremental Voxels based Hybrid Coupled Direct LiDAR-Inertial Odometry

This paper proposes an incremental voxels based hybrid coupled direct LiDAR-inertial odometry (IVHC-DLIO) algorithm. The hybrid coupling approach combines the advantages of loosely coupled and tightly coupled methods. On one hand, the incremental voxel-based mapping technique is employed, resembling a loosely coupled approach. The IVox-GICP solver leverages the easy maintenance and efficient k- NN search capability of incremental voxels to achieve incremental GICP point cloud registration for scan-to-map alignment. On the other hand, the factor graph is used to achieve the tightly coupled state optimization and the optimized state is then utilized for subsequent point cloud motion compensation and state propagation, thereby enhancing mapping accuracy. The proposed method is evaluated on various public datasets, demonstrating excellent performance in terms of robustness, localization accuracy, and runtime.