Research On Mapping And Localization Algorithm Based On Lidar In The Dynamic Environment

With the rapid development of intelligent technology,the autonomous ability of robots continues to enhance.Simultaneous Localization and Mapping(SLAM)is the key technology to realize the autonomous movement of robots.Most of the current LiDAR SLAM approaches are based on the assumption of a static environment.However,there are many dynamic objects in the real environment.These dynamic objects can lead to reduced accuracy of LiDAR SLAM based on static environment assumptions.This thesis conducts the following research on LiDAR SLAM for the dynamic environment:Firstly,aiming at the problem of reduced accuracy of LiDAR point cloud map caused by dynamic objects,a dynamic object removal algorithm for LiDAR point cloud map is studied.Aiming at the problem of misidentification of dynamic objects,a descriptor combining LiDAR intensity and geometric structure is used to detect dynamic objects,thereby improving the accuracy of removing dynamic objects from a point cloud map.Aiming at the problem of low efficiency in ground plane fitting,the RANSAC(Random Sample Consumus)algorithm based on region growth is adopted to fit the ground plane in dynamic regions,further improving the operational efficiency of the algorithm.Verify the accuracy of the algorithm in removing dynamic objects and operational efficiency through comparative experiments with other mainstream algorithms on the Semantic KITTI dataset.Secondly,aiming at the problem of reduced localization accuracy of the LiDAR SLAM algorithms based on static environment assumption caused by dynamic objects,a LiDAR-inertial SLAM algorithm based on static feature matching is studied.Aiming at the problem of the unknown pose of the current frame,LiDAR odometry and IMU preintegration pose increment are used to initialize the pose of the current frame,thereby obtaining the initial pose value of the current frame.Aiming at the problem of reduced localization accuracy caused by dynamic feature points,the proposed dynamic object removal algorithm is combined to remove dynamic feature points from the current frame feature points.Obtain a more accurate current frame pose through static feature matching.Verify the localization accuracy of this algorithm through comparative experiments with other mainstream algorithms on the UrbanNav dataset.Finally,the effectiveness of the algorithms proposed in this thesis is verified by building an experimental platform.Build the hardware platform and software framework of the system,and calibrate the IMU internal parameters and LiDAR/IMU external parameters.the dynamic campus environment dataset is collected by the experimental platform for experimental validation and analysis.The experimental results demonstrate that this thesis provides a reliable experimental platform that can achieve highprecision localization and static point cloud maps construction in dynamic environments.