Real-time Positioning And Mapping Of Outdoor Lidar Based On LOAM Improvemen

Simultaneous Localization and Mapping(SLAM)refers to the technology that uses sensors to obtain data information to build a map of the surrounding environment and locate it in real time.It is a key technology for autonomous driving and robot to complete autonomous navigation in unknown environments.Due to the strong anti-interference and long detection range of Li DAR,this paper chooses Li DAR for localization and mapping in outdoor environment.Based on the classical algorithm framework Lidar Odometry and Mapping(LOAM),a real-time SLAM system with high accuracy and strong robustness in outdoor scenes is designed.The specific research content based on Li DAR SLAM in this paper is as follows:The original point cloud is preprocessed,and the point cloud is segmented into ground based on plane fitting method.The ground point cloud was screened out to reduce the influence of some abnormal points and the number of point clouds.Aiming at the disorder characteristics of point cloud,this paper proposed an orderly coding method for segmented non-ground point cloud by using the distance information of point cloud.The point cloud is divided into rings with different numbers according to the distance,and the ring number corresponding to each point cloud is its sequence number.LiDAR odometry based on feature point method is studied.The method of extracting point cloud features based on calculating curvature is not effective in the scene of outdoor geometric structure degradation.The improved Principal Components Analysis(PCA)algorithm is used to extract the feature points of the point cloud.In the PCA selection of adjacent points to fit the plane stage,the adjacent points are dynamically selected according to the distance information.Different number of feature points are selected in different distance intervals to realize adaptive feature extraction algorithm.The extracted feature points were used to construct the distance residual function to realize the data association between feature points.The objective function is solved by nonlinear optimization method,so as to estimate the pose of the front Li DAR odometer.Back-end loop detection and optimization are used to reduce the cumulative error of odometry.The change of relative pose is used to select key frames.The Scan Context method is used to detect the loop back of key frames.The similarity between the current frame and the history frame is calculated to determine whether the closed loop is generated.The accuracy of loop back detection is improved by the Iterative Closest Point(ICP)algorithm and the Normal Distributions Transform(NDT)algorithm.The point clouds are concatenated to construct a global map.The closed-loop constraint factor and odometry factor are added to the factor graph to optimize the global map and trajectory pose.The effectiveness and robustness of the proposed SLAM system are verified on KITTI dataset and MVSECD dataset.The localization accuracy is verified by comparing the pose error of the odometer with several excellent algorithms.In this paper,the odometer translation error is reduced by at least 19%,and the rotation error is reduced by at least 7.1%.This paper analyzes the robustness of SLAM system for localization and mapping in different outdoor environments.The running time of each module of the system was compared and analyzed.The real-time performance of the SLAM system is verified.