Research On AGV Mapping And Navigation Technology Based On Visual SLA

As an important branch of mobile robots,automated guided vehicles(AGVs)are widely used in manufacturing,warehousing,ports,dangerous and special operations and other industries.Simultaneous Localization and Mapping(SLAM),as a key technology of AGV,has always been a research hotspot.Visual SLAM obtains the surrounding environment information through the camera,completes the real-time positioning of the AGV itself and the construction of the environment map.It has the advantages of a large amount of information and a wide range of applications.In this paper,the research status of AGV and visual SLAM at home and abroad are reviewed and analyzed.On this basis,the autonomous localization,map construction and path planning algorithms of visual SLAM are theoretically analyzed and experimentally studied.The main research contents and results are as follows:(1)Establish camera imaging model according to geometric optical imaging principle,analyze coordinate transformation relationship of object image space and solve camera image distortion coefficient in detail.Zhang Zhengyou calibration method is used to complete the calibration of depth camera,and the calibration effect is good.(2)In view of the problems of redundancy and clustering of feature points in traditional ORB algorithms,Q＿tree uniformity algorithm is used to homogenize feature extraction.While reducing redundant feature points,feature points are distributed more evenly in the whole image,effectively reducing the calculation amount of back-end pose estimation.In back-end pose estimation,EPn P and ICP algorithm are combined to improve the accuracy of pose estimation.(3)Aiming at the problem that the sparse landmark map of sparse feature points cannot meet the requirements of path planning,the dense map construction thread is added to construct three-dimensional dense point cloud map,and the feasibility of the algorithm is verified on TUM dataset.(4)In view of the problem that A~* algorithm has repeated node search in the process of path planning,which affects the efficiency of optimal path planning,A~* algorithm is combined with jump point search to effectively screen out unnecessary nodes,which improves the path planning efficiency of the algorithm and has obvious effect on largescale maps.Finally,ROS is used as the software platform to conduct AGV diagram building,path planning and obstacle avoidance experiments in actual scenes on the self-built AGV experimental system.The experimental results show that the improved ORB-SLAM algorithm can construct the 3d dense point cloud map of the actual scene well,and the improved A~* algorithm can plan the complete path.At the same time,the improved A~*algorithm and the local path planning algorithm have good adaptability,and the real-time obstacle avoidance function of AGV can be realized in coordination with each other.