An improved LIO-SAM lidar inertial odometer with dynamic point filtering

Abstract LIO-SAM is an extensively utilized SLAM (Simultaneous Localization and Mapping) algorithm, featuring an integrated lidar-inertial odometry framework for high-precision, real-time trajectory estimation and mapping in mobile robots. In this paper, we integrate the lidar Iris loop detection algorithm into the LIO-SAM framework to solve the poor loop detection of LIO-SAM itself and the poor loop detection effect of Scan Context resulting from varying lidar directions passing through the same location. Additionally, an enhanced dynamic point filtering algorithm relying on ERASOR is developed to mitigate the presence of dynamic points in the point cloud map created by SLAM. Finally, validation with the KITTI dataset confirms the effectiveness of these algorithms, laying a robust foundation for further advancements.