Research On Indoor Multi-robot SLAM Based On 3D Lidar Point Cloud

The mobile robot is equipped with sensors to observe unknown environmental information and create a map.At the same time,determines the position of the robot in the map,that is,simultaneous localization and mapping(SLAM).SLAM is the basis algorithm for mobile robots to perform navigation,path planning,and autonomous exploration.It has always been a key research content in the field of mobile robots.With the development in recent years,the single robot SLAM algorithm has gradually matured.However,in the face of large-scale environments,single-robot SLAM has problems with low work efficiency,poor stability,and low map accuracy.In summary,the single robot SLAM is not the best choice for performing tasks in a large-scale environmentCompared with single robot SLAM,multi-robot SLAM could cooperate to build maps by sharing environmental information and fusing different sensors information.Multi-robot SLAM has the advantages of high efficiency,high fault tolerance,high map accuracy.It is more suitable for a variety of complex tasks and more researchers had focused on multi-robot SLAM.At the same time,multi-robot SLAM needs to solve the problem of the relative pose estimation and local map fusion of different robots.Therefore,it is of great significance to study multi-robot SLAM.This thesis studies the relative pose estimation,map association,global map optimization of multi-robot SLAM algorithm.Firstly,analyzing the commonly used three-dimensional environment map description methods,and compared the advantages and disadvantages of point cloud maps,three-dimensional grid maps,semantic maps,using OctoMap to describe environmental information.And we also performed the study of the classic algorithms in 3D SLAM included:LOAM algorithm and Cartographer algorithm.By analyzing the basic principles of the two algorithms,compared the similarities and differences between the two algorithms,the Cartographer algorithm is selected as the basis for the research of the multi-robot SLAM algorithm.Secondly,the characteristics of OctoMap and indoor environment were analyzed,and a method of extracting OctoMap feature points based on sliding windows was proposed.According to the rate of change of the duty ratio(ratio of the number of occupied stereo grids to the number of idle stereo grids)in the sliding window,combined with the area segmentation-based plane segmentation algorithm,the distance from the point to the plane was calculated to complete the OctoMap feature point extraction.Based on the retrospective method to search the largest common sub-graph,we achieved OctoMap fusion associate the feature points.Thirdly,analyzing the structure of the multi-robot system and combine the performance of the robot in this subject.Based on the hybrid multi-robot system,the robots build the maps in a distributed manner and collaborated in the centralized processing.Based on the idea of graph optimization SLAM,the multi-robot SLAM algorithm is divided into front-end and back-end.The front-end calculated constraints through sequential association of submaps and closed-loop detection to construct a multi-robot pose map.Based on the idea of bundle adjustment,the back-end establishes a global error function,solves iteratively,optimizes the trajectory of the robot,and creates a globally consistent OctoMap.Finally,the experimental platform was set up.The Xiaoqiang XQ-4 Pro robot was equipped with Rfans-32 3D lidar,which sensed environmental information.This robot was based on the ROS framework,which could save lidar information and robot motion information to create sub-map sequences.The validity of the proposed multi-robot SLAM algorithm was verified based on the created sub-map sequences and public data.