Research And Implementation Of Indoor SLAM System Based On Cartographer

In recent years,with the rise of autonomous robots represented by home service robots,the rapid development of AR/VR technology and the autopilot technology,the Simultaneous Localization and Mapping(SLAM)technology have attracted wide attention.In this paper,the Cartographer indoor SLAM algorithm of Google is studied.In order to improve the accuracy of its mapping and positioning,the algorithm is improved.First,a three wheel universal mobile robot platform was built to carry out field SLAM test indoors.The robot uses the raspberry pie 3B as the hardware development platform to realize the bottom control and drive each sensor.The robot uses the robot operating system(ROS)as the software development platform,and realizes the data transmission between the sensors through the message transfer mechanism and the coordinate system transformation(TF)system provided by the ROS.Second,a lidar data compensation method is proposed to solve the problem that the distance between lidar and glass cannot be detected due to the defect of lidar.By studying the data acquisition mechanism of lidar,according to the data of the lost part of the lidar,the ultrasonic sensor data is used for reference,and then the least square method is used to fit the data,and the compensation of the lost data is completed.According to the field test results,this method can effectively fill the distance data between lidar and glass,and improve the success rate and accuracy of SLAM.Third,detailed analysis the mapping and positioning principles of the mainstream Gmapping SLAM and Hector SLAM algorithm.Then we use the same rosbag dataset to simulate locate and map on ROS system.According to the results of the final construction,we compare and analyze the advantages and disadvantages of these two SLAM algorithms and Cartographer,and choose Cartographer as the main research and implementation algorithm in this paper.Fourth,in this paper,a method of improving the grid map building with the feature extraction of lidar scanning is proposed,which effectively reduces the influence of the fluctuation of the scanning point data on the map building.Because of the turbulence caused by the error of the lidar itself and the improper operation of human beings,it is easy to cause the fluctuation of the scanning point,which leads to the existence of a large number of error points in the edge of the image building.In this paper,according to the analyse of the front-end algorithm occupy grid map construction and the local scan matching of the Cartographer,extracts the feature of the line segment in the scanning frame and performs the fitting.After that,the scanning point is moved to the straight line after fitting.After field testing,this method can effectively improve the clarity and sharpness of edges in map building.After that,the back-end graph optimization theory of Cartographer and the branch and bound algorithm to accelerate the solution of nonlinear least squares are also studied,and the mobile robot platform is used to realize the simultaneous construction and location of the field.The experimental results show that the improved method of lidar data compensation and the feature extraction of the lidar scan frame can effectively reduce the range error of the lidar and improve the accuracy of the Cartographer SLAM algorithm,and control the location error below 0.24m.