Research On LiDAR SLAM Based On Local Graph Optimization

In recent years,with the development of localization and navigation technology,simultaneous localization and mapping(SLAM)has attracted more and more attention.The SLAM algorithms based on 3D LiDAR sensor now play an important role in the fields of autonomous driving,unmanned vehicle logistics system,Power line inspection,and three-dimensional reconstruction.Based on a portable mobile platform equipped with LiDAR sensor,this paper designs a 3D LiDAR location and mapping system.And main contributions of this paper are summarized as follows:First,the paper introduces the portable mobile platform from aspects of hardware layer,algorithm layer,and the client.And the time synchronization of multiple sensors is achieved from the hardware and software levels.Secondly,this paper studies the method for localization and mapping based on local optimization using LiDAR sensor.The front-end of the method uses feature point registration to obtain the pose between adjacent frames.While the problem of point cloud distortion is mitigated using quaternion spherical interpolation.In the back-end,to balance the accuracy and computational complexity,point cloud registration based on local omnidirectional map and optimization based on local pose graph between key frames are adopted,which both improve the localization accuracy.Then the ICP algorithm is used to complete loop detection between key frames.Following steps above,a localization and mapping system based on LiDAR is implemented in this paper,and experiments are designed to verify the front-end and back-end methods.Finally,this paper explorers the fusion algorithm of 3D LiDAR and IMU.Aiming at solving the problem of inconsistency between LiDAR and IMU sensor coordinate systems,this paper further proposes an off-line calibration method that can be used to initialize LiDAR and IMU.The method can be used to estimate the external parameters between the two sensors and the IMU bias,which are verified based on some experiments carried on the open source data set.