Research On Two-dimensional Simultaneous Localization And Mapping Technique Based On Lidar

With the advent of intelligence era,intelligent mobile robot has brought high efficiency and convenience in more and more fields.The ability of auto-navigation is an important symbol of highly intelligent robot.Robot explores the unknown environment to achieve autonomous movement by sensors,which contains a series of operations such as acquisition,processing,analysis,understanding and decision-making to environmental information.The robot utilizes the sensor to collect the scene data and then estimate the position and posture information.SLAM is the process of mapping the environment map by position and posture information.There are three major needs of SLAM technology including high stability,low complexity and high consistency need to improve now.In this paper,the MATLAB simulation experiment is designed to study the SLAM algorithm,the LIDAR experiment of SLAM application technology is implemented,and the LIDAR prototype of SLAM system is designed and implemented.(1)The SLAM models including motion model,observation model and map model are designed.Analyzing the motion process and noise influence of the two-dimensional SLAM problem,the conditional probability transfer equation based on the velocity of motion is used as the motion model.The observation model is established using the transform of polar coordinates to rectangular coordinate.The occupancy grid map model is designed to represent the distribution of obstacles.(2)The filtering principles of KF and EKF are analyzed.The SLAM algorithm procedure using EKF is designed,and the simulation experiment based on EKF-SLAM is carries out.The effects of different parameters such as system noise covariance,measurement noise covariance,number of landmarks and data association thresholds on SLAM results are studied experimentally.The experimental results expound that the system noise covariance,the measurement noise covariance and the data association threshold are the factors that affect the consistency between the SLAM estimation and the real value,and the number of landmarks is the factor that affects the accuracy of slam estimation.The parameter optimization program is provided.(3)The algorithm principle of PF and FastSLAM is analyzed.The data structure of binary tree is designed in FastSLAM calculation process,and the FastSLAM technology based on EKF-SLAM and particle filter is studied with simulation experiment.The experiments verify that the binary tree can effectively reduce time and memory usage,evaluate the positioning accuracy of EKF-SLAM and FastSLAM with different particle numbers using the same dataset,and prove that the particle number can effectively improve the estimation precision in FastSLAM.The location error of EKF-SLAM and FastSLAM using the same data set is evaluated,and the consistency between the FastSLAM estimation and the real value is proved.(4)The research and design of application-based SLAM system is carried out.The Hector SLAM and Cartographer are studied,and the experiment environment of RPlidar A1 LIDAR with ROS platform is built to carry out the experiments on the campus buildings.The experimental parameter table is designed based on the parameter optimization program in EKF-SLAM research and the data structure of binary tree is used to optimize the calculation process.Applicability characteristics of Hector SLAM and Cartographer are obtained by the experiments.The research of hardware environment in SLAM system is carried out,and LIDAR prototype based on MEMS micro-mirror scanning is designed and realized(optics,mechanics,electronics,software)under new technology.There are key technologies such as high precision time information extraction technology and detector dark count filtration technology.