Key Technology Analysis And Evaluation Of Laser SLAM System Based On Cartographer

The vehicle-mounted mobile measurement system obtains the high-precision position and attitude(POS)through GNSS/INS,but the GNSS signal often loses lock in city streets,the POS error solved by the INS alone is accumulated over time.POS accuracy is seriously affected or even unusable,and other sensors are required to assist in obtaining high-precision POS.LiDAR is preferred for its rich observational information,high measurement accuracy,and excellent all-weather capability.The laser SLAM algorithm has been developed to integrate LiDAR and INS to solve POS problems.In addition to the underlying SLAM algorithm,the complete laser SLAM system also involves technologies such as multi-sensor integration,data acquisition and visualization(system upper layers).The robot operating system(ROS)includes libraries such as navigation packages,coordinate systems,and 3D visualization tools,speeding up the development of the upper level of the system.In September 2016,Google’s open source Cartographer algorithm platform provided reference for the implementation of the underlying and upper layers of the laser SLAM system.At present,official documents,online blogs,etc.are relatively one-sided and not systematic about the introduction of the platform,so it is difficult to use the advantage of the platform for rapid development.Based on the above characteristics,this paper studies the construction of a complete laser SLAM system based on the Cartographer,analyzes and evaluates key technologies involved,provides a reference for the rational use of the Cartographer platform,and provides references for the integration of vehicle-mounted mobile measurement systems in the later period.The research involves the following:1)Study the key technologies of the complete laser SLAM system and the Cartographer source code logic.Analyze the advantages of the Cartographer platform in terms of algorithm and efficiency,including point cloud data preprocessing,laser SLAM front-end,back-end graph optimization,and loopback detection.2)Analyze the advantages and disadvantages of the non-linear optimization library Ceres Solver and G2O.The Cartographer optimizes the graph based on the Ceres Solver.This article replaces the the Cartographer source code and based on the G2O and Process and analyze the same data and calculate time consumption.For Pose-Graph,Ceres Solver is relatively efficient with respect to G2O,but G2O defines the types of vertices and edges commonly used in graph optimization,G2O is more convenient to implement.3)Implement tactical inertial STIM 300 driver based on ROS to obtain sensor data,and provide general sensor drive design flow to provide reference for sensors on ROS platform;4)Build 2D and 3D laser SLAM systems,including hardware platform and software platform based on the Cartographer,through experiments to analyze the impact of laser SLAM system key technology:back-end optimization,loop detection and calibration error on post-processing point cloud,and based on bearing columns The body and target ball design scheme evaluates the accuracy of the system.After calculation,the indoor accuracy of the 2D SLAM is 5-6 cm,and the outdoor accuracy of the 3D SLAM is 10-15 cm.In addition,in order to use the Cartographer algorithm platform flexibly and develop with the platform,the involved parameters are analyzed through theory and experiments.