Research On Intelligent AGV Map Building And Autonomous Navigation And Obstacle Avoidance System

In recent years,the mapping and navigation technology of AGV(Automated Guided Vehicles)has developed rapidly.The key technologies mainly include Simultaneous Localization and Mapping and path planning.Although visual feature-based methods can effectively detect 3D point clouds,they cannot work in non-luminous environments and have problems such as limited observation range and poor stability.Navigation and obstacle avoidance methods based on 2D laser have begun to be applied,but they still have some shortcomings,such as their observation range being limited to a certain plane and their inability to effectively observe 3D obstacles.To address these issues,this paper proposes a 2D grid map construction method that integrates depth cameras,2D Li DAR,and IMU,and realizes dynamic path planning for AGV through the observation information from depth cameras and 2D Li DAR.The main research content of this paper is as follows:Firstly,a 2D grid map construction method with multi-sensor information fusion is designed to address the problem of incomplete information in 2D laser SLAM mapping.By fusing the data from 2D Li DAR and depth images using point cloud filtering,a laser scan frame is constructed,which preserves richer obstacle information in the local map.After the local map is constructed,the motion distortion of the scan frame data is corrected using IMU data to reduce the cumulative error of local positioning.At the same time,scan matching is used to correct the pose of the local map.When a similar scene is detected,loop closure detection is used to eliminate the cumulative error in the positioning process,and global scan matching is accelerated using branch and bound method.Secondly,a relocation method for AGVs based on an adaptive Monte Carlo algorithm is proposed to address the problem of manually specifying the pose during AGV relocation.When the navigation function package is launched,the sensor observation information is matched with the map through self-rotation,enabling the AGV to perform global relocation at any position in the map and improving the intelligence of AGV navigation.Then,a path planning method for AGV is designed based on a fusion of the improved A*algorithm and the dynamic window approach to address the issue of excessive search nodes and ineffective avoidance of 3D obstacles.This method enables the AGV to adjust the heuristic function weight adaptively according to the obstacle weight from the parent node to the target point,reducing the number of search nodes and improving planning efficiency.At the same time,the global path turning points are optimized twice to ensure the optimality of the global path.Subsequently,the global path turning points are set as sub-goals of the DWA algorithm in sequence to achieve dynamic path planning.Additionally,the distance information of obstacles observed by depth camera is introduced to enable the AGV to effectively avoid newly-added 3D obstacles in the scene.Finally,the performance of the AGV map building and navigation and obstacle avoidance system is tested.The experimental results show that the algorithm is able to build 2D raster maps in real scenarios in real time,and the root mean square error of the system can reach0.033 m in positioning accuracy and 2.52 cm in map building accuracy.Compared with the traditional fusion algorithm,the planning time of this improved fusion algorithm is reduced by 53.39% and the path length is shortened by 16.76%,and the system can avoid 3D obstacles that cannot be detected by 2D LiDAR.