Research On Map Merging Based On Multi-Robot Cooperative SLAM

SLAM(Simultaneous Localization and Mapping)is very important in the field of robotics,which can draw a detailed map of the surrounding environment.Especially in an unknown environment,it is of great significance by using robots to build a detailed map which can be used to autonomous navigation of robots,such as welcome robots,cleaning robots,inspection robots,autonomous vehicles,etc.Moreover,it enables people to grasp global maps more quickly in significant situation,such as military wars,mine disasters and deep sea exploration,etc.With the development in the last three decades,many advanced SLAM technologies have been developed,most of which,however,are based on a single robot.Multi-robot cooperative SLAM can reduce the repetitive detection area and increase the speed of mapping.With the extensive application of SLAM technology and the increasing range of detection area,this advantage has drawn more and more attention.Consequently,more and more related researches have been carried out.The main content of this thesis is to solve the problem of map merging in multi-robot cooperative SLAM.At present,there are three main types of maps,including feature map,grid map and topology map,and different maps have different representations.This thesis mainly studied feature map and grid map.And according to different features of the maps,the study applied two different methods to map fusion.First of all,this thesis introduced EKF-SLAM,FastSLAM algorithm and HybridSLAM algorithm combining the advantages of both.Then,an improved HybridSLAM algorithm was proposed in this paper.Based on the improved HybridSLAM algorithm,two feature maps were merged according to the Kalman filter principle.This fusion method was verified by MATLAB simulation.Besides,in order to improve accuracy and scope of laser detection,we upgraded the TurtleBot2,including reserve the Kobuki chassis and replace Kinect by laser range finder,Rplidar A2.And the effect of the upgrade was verified by actually running SLAM.Improved TurtleBot2 provided the robot model for the simulation of map merging.Finally,this article introduced a different fusion method to grid maps.A grid map constructed by a robot running SLAM was taken as a single picture,and each cell represented a pixel.The study used the image stitching technology to find two overlapping areas of the map,and then calculated the relative transformation matrix to map merging.It also was validated in the Gazebo simulator and the ROS platform.Two improved TurtleBot2 models and indoor environments were built in Gazebo,and the controls of SLAM of robots and the fusion of two grid maps were carried out in ROS.