Research On Indoor Map Construction Algorithm Based On 3D LiDAR

The low-resolution 3D LiDAR sampling is sparse,and the 3D LiDAR sampling in motion is subject to point cloud distortion.Indoor environments are considered degraded environments because there are significantly fewer feature points in such structured scenes indoors compared to the rich environmental features outdoors,especially in long corridor environments.Based on the above two points,the point cloud collected by 3D LiDAR sampling in indoor environment will not accurately reflect the real information of the environment,and will even have a large drift or map building failure.This paper explores the 3D laser SLAM(Simultaneous Localization and Mapping)algorithm based on the existing 3D LiDAR SLAM algorithm for indoor applications,and the main contents are shown as follows:(1)The principles of four currently popular and widely used 3D radar laser SLAM algorithms,LOAM,LEGO-LOAM,HDL＿GRAPH＿SLAM,and SUMA,are briefly introduced.The performance of the four algorithms is tested in the public data set,and the LEGO-LOAM algorithm with the best overall performance is selected as the further research algorithm.(2)To address the inherent problems of 3D LiDAR sampling and the characteristics of indoor environment,the front-end of LEGO-LOAM is used to fuse IMU(Inertial Measurement Unit)data to correct the LiDAR point cloud distortion,and at the same time,the real-time IMU data is fused with 3D LiDAR data through an extended Kalman filter to obtain a more accurate bit pose.In addition,an adaptive covariance matrix is designed,and the system determines the number of degrees of freedom used for matching based on the number of indoor environmental feature points.Finally,the comparison is validated on a publicly available dataset.(3)To address the problem that the loopback detection method used in the LEGOLOAM algorithm can produce scale drift over time or in large scenes,Scan Context global descriptors are introduced in the loopback detection part of the LEGO-LOAM algorithm.We set the time and frame thresholds according to the indoor environment,and use vector nearest neighbor search and similarity score matching strategies to improve the loopback detection efficiency and mitigate the drift problem.Finally,the comparison is validated on a public dataset.(4)An experimental platform for mobile robots is built,and the improved algorithms are constructed in real time maps in real scenes to verify the effectiveness of the improved LEGO-LOAM algorithm.