Indoor Laser-based Simulataneous Localization And 3D Mapping

Lidar-based simultaneous localization and 3D mapping is the key technology in the field of unmanned driving,and is also an important research directions in the autonomous navigation of mobile robots.The main work of this article is as follows:1.A method for simultaneous localization and 3D mapping using motors and 2D laser radar is proposed.Firstly,the method of lidar data acquisition is studied,and a set of platform that is driven by the motor to rotate the two-dimensional lidar back and forth is designed.Secondly,the method of solving the three-dimensional point cloud coordinates of the motor and the two-dimensional laser radar is introduced,and the steps of packaging the radar,the motor,and the solution of solving the coordinates into packages under ROS are given.2.In this paper,a variety of point cloud feature extraction and registration algorithms are studied.A curvature-based point cloud feature extraction method is proposed and the point that may bring errors to the subsequent registration algorithm is removed.Aiming at the disadvantages of traditional point cloud registration using ICP algorithm and its variant algorithm,a method based on triangle congruence relation to find the corresponding feature point pair was proposed.For the first time,the data structure of dictionary tree was applied to the corresponding feature point search strategy.The pose transformation of lidar is solved by constructing a nonlinear least squares problem.3.For the point cloud registration between frames and frames will cause the error to accumulate over time.This paper fully studied the optimization algorithm and proposed a global optimization algorithm based on the voxel grid idea.Then,the filter theory used in the traditional back-end optimization algorithm and the commonly used nonlinear optimization are introduced.The back-end optimization of our SLAM system is performed based on the general-graph optimization algorithm,and the comparison of the results before and after the pose optimization is given.4.Introduce the data transmission of our SLAM system on the ROS platform and use the Robo Masters platform to experiment in both long corridor and spiral staircase environments.The point clouds of the point cloud feature extraction node and the motion trajectories of laser odometry node under RVIZ are given,as well as the final results of the localization and mapping of the experimental results.All calculations in this system are completed on the laptop in real time.The experimental results show that the proposed Li DAR-based simultaneous localization and 3D mapping system can accurately estimate the trajectory of the lidar and construct a high-quality 3D point cloud in real time.The relative accuracy of the test in the indoor environment is about 2%.The results obtained in this dissertation can be widely applied in the field of autonomous vehicles and autonomous navigation for mobile robots.It has great significance for the popularization of mobile robots in China.