Mapping And Path Planning Based On Lidar Feature Extraction

In order to realize the leap of China’s light industry from a large manufacturing country to a strong manufacturing country,it is an important initiative to realize the intelligence of light industry equipment.As a kind of intelligent equipment,autonomous robots are gradually used in light industry.In the field of autonomous robot research,the primary solution is the problem of positioning and navigation.The robot must have a high-precision map to achieve accurate positioning and navigation,and the high-precision map is based on the precise positioning of the robot,so the two complement each other,and it has become a hot issue in this field.In terms of map construction,feature map has the advantages of simplicity,visualization and small storage capacity compared with the grid map.However,due to its difficulty in the path planning,the contemporary mainstream localization,mapping and path planning algorithms are based on the grid map.To make full use of the advantages of the feature map,it is of great significance to study the algorithms of localization,map construction and path planning based on the feature map.By taking autonomous robots with lidar as the research object,the method of lidar feature extraction are discussed in this thesis,and the algorithms of localization,mapping and path planning for robots based on the extracted features are proposed.The specific research results are as following:(1)After analyzing the data characteristics of the point set obtained by lidar scanning,a new algorithm is proposed to split the point sets rapidly and accurately by calculating the inclination difference of the line formed by two adjacent points.On this basis,two-point fitting is used instead of the least-square method to fit the segmented point set into line segments,which further improves the calculation efficiency.Aiming at the problem of over-segmentation in the process of segmentation,the characteristics of the sub-point set after segmentation are analyzed,and the judgment principle and implementation steps of sub-point set merging are obtained.Properties and presentations of corner feature and line segment feature are defined,and the accurate classification and extraction of corresponding features are realized.(2)The matching algorithm based on corner and line segment features is realized by using feature attributes to search corresponding points,which improves the speed and accuracy of the matching algorithm.Aiming at the matching error caused by the difference of the length of the line feature,the corresponding point of the line segment feature matching is modified from the middle point of the segment to the projection of the endpoint of the corresponding line segment,which improves the matching accuracy.Aiming at the problem that some extracted features may have large errors,the weight coefficients of extracted features are added to the objective evaluation function to improve the accuracy of transformation parameters after matching.Compared with the mainstream iterative closest point(ICP)algorithm and correlative scan matching(CSM)algorithm,the experimental results show that the average matching error of the proposed algorithm is reduced by 47.23%and 24.39%respectively,and the computational efficiency is improved by 84.75%and 29.04%respectively.(3)Thruough the method of graph optimization,the pose and feature position of the robot after matching are optimized,which solves the problem of large cumulative error when using the matching results directly,and realizes the global localization of the robot and the construction of feature map more accurately.In order to improve the accuracy of features,clustering is used to merge the same features in the submap.Meanwhile,to solve the problem of feature expansion of line segments in the map,this thesis analyzes the position relationship between line segment features,and puts forward the method of feature expansion and merging of different types of line segments.(4)In the global path planning,according to the position,direction and length of line segment features,a searching and optimization algorithm based on feature map is proposed,which realizes the fast global path planning and makes full use of the advantage of high computational efficiency of feature map.This thesis analyzes the problem of the robot searching direction selection at turning points and corners,and puts forward the solution and implementation steps.Aiming at the problem that the robot may fail to search the path at the end of the obstacle segment,the algorithm is improved by using the method of circumventing the end point of an obstacle.In order to solve the problem that the search path may not be the shortest path,the variable parameter optimization algorithm is used to optimize the searched path nodes to obtain the optimal path.The simulation results show that compared with A*algorithm,the proposed algorithm can reduce the path length,the number of path nodes and the computation time by 11.05%,57.89%and 96.94%,respectively.(5)In local path planning,the obstacle detection method is improved,so that the feature map can be applied to the dynamic window approach(DWA).By using the idea of variable weight coefficient to improve the objective function,the problem that DWA algorithm is sensitive to the global parameters is solved.Finally,the searching and optimization algorithm is fused with the improved DWA algorithm.the global path planning is used to derive possible shortest path first,then each small path which uses the node as the target position is optimized by using local path planning,meanwhile,in order to solve the problem that the moving speed of robot slowdown noticeably in the middle path nodes,this thesis improves the calculation method of direction function to achieve the optimal path finally.It is verified that under the same conditions,compared with the fusion algorithm of A*and DWA,the proposed fusion algorithm can reduce the computational cost of time by 74.92%at most and improve the smoothness of the robot’s moving path by 83.33%at most.In summary,a complete set of autonomous SLAM and path planning algorithm based on the feature map is implemented in this thesis,which includes feature extraction,matching and positioning,mapping and path planning based on lidar.The experimental results show that the algorithm is effective and can fully reflect the advantages of high computational efficiency of the feature map.