Research On Simultanious Localization And Mapping For Unmanned Vehicle In Dynamic Environment

Self-driving technology can not only reduce traffic pressure and traffic accident rate in cities,but also reduce logistics costs and improve the efficiency of materials distribution in factories.Autonomous driving vehicles need to have a complete and reliable perception system in the complex and variable environment,especially in the situation where GPS signal cannot be received,reconstruct the surrounding environment and self-positioning in real-time.Therefore,Simultaneous Localization And Mapping(SLAM)in dynamic environments is a significant subject research.In this paper,the SLAM method in dynamic outdoor environment is studied by using lidar as the sensor and four-wheel differential steering AGV trolley as the experimental platform.Firstly,the coordinate system model and sensor model of the unmanned vehicle system are established.Based on the traditional point cloud filtering method,a constant velocity motion model was established for the correction of distorted point cloud caused by the car’s motion.After separating the ground points and non-ground points in the environment,a point cloud segmentation method based on geometry is used to segment the different targets in the environment,and the filtering process is carried out by combining the direct filtering and voxel filtering methods.In view of the characteristics of multi-line lidar,the valuable edge points and surface points are extracted from the environment point cloud by calculating the curvature of the points.And then the line and surface features can be constructed.The two adjacent frame point clouds are correlated by point-to-line matching and point-to-surface matching.Then the constraint equation can be constructed,and the constraint equation is solved by LM optimization algorithm.In addition,for the dynamic outdoor scene,a dynamic target recognition strategy combining geometric features and motion information is designed to reduce the interference of dynamic target on SLAM.Based on the iterative closest point(ICP)algorithm,a loop detection method is designed,and the graph optimization framework is integrated to further optimize the point cloud map and the posture.Based on the self-built unmanned vehicle platform,experiments were carried out on the point cloud pretreatment algorithm,laser range calculation method and dynamic scene building algorithm in this paper,and theoretical analysis was made on the experimental results of the above algorithms in different scenarios.The results verify that the SLAM algorithm can achieve good localization and mapping effect in the actual environment and has practical application value.