LIDAR-Inertial Real-Time State Estimator with Rod-Shaped and Planar Feature

State estimation and mapping based on Light Detection Ranging (LIDAR) are important for autonomous systems. Point cloud registration is a crucial module affecting the accuracy real-time performance of LIDAR simultaneous localization (SLAM). In this paper, novel point feature selection LIDAR-inertial tightly coupled systems proposed. front-end, carried out after marking rod-shaped planar information which different from existing inertial measurement unit (IMU) integration scheme. This preprocessing method subsequently reduces outliers. IMU pre-integration outputs high-frequency result used to provide initial value solution. scan-to-map module, computationally efficient graph optimization framework applied. Moreover, odometry further constrains states. back-end, sliding-window incorporates loop closure global constraints reduce cumulative error. Combining front-end we propose low drift high positioning system. Furthermore, conducted an exhaustive comparison in open data sequences real-word experiments. The proposed system outperforms much higher than state-of-the-art methods various scenarios. Compared with LIO-SAM, absolute trajectory error (ATE) average RMSE (Root Mean Square Error) study increases by 64.45% M2DGR street dataset (street_01, 04, 07, 10) 24.85% our actual scene datasets. most time-consuming each system, runtime can also be significantly reduced due back-end model.