Research On 2D Map Building And Localization Of Mobile Robot Based On Lidar

In recent years,map building and localization technology has been widely used in indoor positioning,resource exploration,home entertainment,and automatic driving.This thesis aims to establish a map construction and positioning system using lidar and mobile robots.For map building,this thesis establishes pose graph through scan matching and closed loop detection,and completes the optimal estimation of pose graph through back-end optimization,and then according to the optimal the pose is estimated to complete the establishment of the grid map.For global localization,this thesis introduces a global positioning algorithm based on particle filter,proposes a global localization based on feature matching.Comparing the two localization schemes this thesis proposes global fusion localization based on particle filtering and feature matching.In order to solve the problem of localization failure,this thesis realizes by monitoring the sensor measurement probability.The main work and innovation are as follows:Firstly,this thesis introduces the establishment of graph-based Simultaneous Localization and Mapping(Graph-SLAM)model.According to the model,the realization of three modules of scan matching,closed loop detection and back-end optimization in pose graph is introduced.Based on the pose and measurement information of each node in the pose map,the establishment of the grid map is introduced.Secondly,after introducing the framework and derivation of particle filter,this thesis introduces the motion model and the measurement model,introduces the global localization algorithm based on particle filters.the simulation test is carried out on the algorithm,and the change of positioning error during the motion is discussed.Then,this thesis proposes a global localization based on feature matching,using line segment detection algorithm to extract the line segment features of the grid map and establish a map feature database,using the split-and-merge algorithm to extract the line features of the lidar scan and establish the feature template,finding the candidate pose set by feature association,calculating the optimal estimated pose by the defined matching fit.Compared with the global positioning based on particle filters and feature matching,this thesis proposes global fusion localization based on particle filters and feature matching.the estimated pose based on feature matching localization is used as the initial pose of the localization algorithm based on particle filters.The problem of solving the positioning failure is achieved by monitoring the sensor measurement probability,And through different test cases,the three algorithms are simulated and compared.Finally,this thesis introduces the implementation of map building and global localization on the Robot Operating System.The map building provides grid map input data for global positioning.The global positioning is divided into based on particle filters localization algorithm,based on feature matching localization algorithm and fusion localization algorithm,and the localization error of three algorithms is compared by localization test.