Simultaneous Localization And 3d Environment Mapping For Unmanned Vehicle Based On Lidar Perception Simultaneous Localization And 3d Environment Mapping For Unmanned Vehicle Based On Lidar Perception

In recent years,more and more researchers are involved in the research on the technology of autonomous driving and mobile robot.The ability of simultaneous localization and mapping(SLAM)is an important prerequisite for the realization of unmanned system.Lidar is widely used as the main sensor in SLAM technology due to its high precision and all-weather operation.In this thesis,multi-line lidar is used as the main data acquisition sensor to carry out the research on the location and environment modeling of unmanned vehicle.First of all,this article gives a brief introduction to the construction of the software and hardware research platform for unmanned vehicles,mainly including the sensors used in the experiment,software development and the operating environment.Secondly,considering the appearance of unreliable feature points caused by small objects in complex environments,this thesis uses the method of point cloud preprocessing to avoid the extraction of unreliable feature points.The point cloud pre-processing stage separates the points from the ground and small objects to avoid the occurrence of erroneous constraints caused by unreliable feature points.The subsequent processing based on the processed point cloud can improve the accuracy of the odometry and increase the robustness of the algorithm in complex environments.Then,a high-precision lidar localization and mapping algorithm is introduced in this thesis.It mainly includes three parts: feature points extraction,feature points matching and inter-frame pose calculation.At the same time,the method of tight coupling with IMU is proposed to improve the accuracy of the odometer.The loop closure detection module using the odometer information as the position prior is also added to the system.In the backend,a graph optimization method based on factor graph is used to correct the final pose trajectory and obtain a globally consistent map.In addition,for some application scenarios where the constraints of feature points are insufficient,such as scenes with relatively empty environments and structured scenarios,additional constraints obtained from other odometer information are added to the optimization of pose graph,which makes the algorithm ensure the accuracy of localization and mapping even in the case of misalignment of the lidar odometer.Finally,based on the unmanned vehicle research platform,a program is developed to realize the algorithm and verify the effect of the algorithm.Indoor experiments are carried out in the environment of the teaching building.At the same time,outdoor data sets are used to verify the effectiveness of the algorithm for large-scale mapping and loop closure detection.The experimental results are displayed with the help of visualization tools provided by Robot Operating System(ROS).The results show that the effect is improved compared to the similar Lidar SLAM algorithm.