Study on LiDAR inertial tightly coupled SLAM method for outdoor environment

Abstract Aiming at the present situation of laser simultaneous localization and mapping (SLAM) algorithms in outdoor environments, such as insufficient point cloud processing, back-end loop and poor mapping effect, a laser IMU tightly coupled localization and mapping scheme is proposed. The point cloud distortion compensation is completed through IMU pre-integration, pathchwork ground segmentation is introduced, and the point cloud with low recognition is eliminated by non-ground point clustering. The optimal pose is obtained by combining with inter-frame matching, which improves the stability of feature association and the real-time performance of the system. A map optimization scheme based on sliding window idea is designed to build a point cloud map by integrating odometer, IMU pre-integration and loop information. The Kitti data set was used to verify the odometer accuracy and the superiority of loopback and mapping effect of the proposed scheme. The feasibility of the proposed scheme was evaluated in different outdoor environments, and the superiority of the proposed scheme was further verified by multidimensional comparison with LeGO-LOAM algorithm at qualitative and quantitative levels. The experimental results show that the SLAM algorithm proposed in this paper has high robustness and mapping accuracy. Compared with the LiDAR-inertial odometry-SAM algorithm, the root mean square error of the proposed algorithm on Kitti00, 07 and 09 sequences is reduced by 20%, 49.73% and 10.62%, respectively. Compared with the LeGO-LOAM algorithm, the error of the proposed algorithm in the Kitti 00 sequence containing the looping scene is reduced by 69.26 %, and the error in the Kitti 10 sequence without the looping scene is reduced by 19.65%. In the real outdoor environment, the elevation deviation of the z -axis and the plane error of the xy -axis are controlled within 7 cm, and the pitch angle, yaw angle and roll angle are within 1 degree.