Research On Cluster Scheduling AGV Based On Stereo Vision Technology

With the rapid development of e-commerce and new retail industries,the warehousing and logistics industry is developing towards intelligent and unmanned.At the same time,the sorting efficiency and spatial density are improved.People spend more and more efforts to the research and development of AGV,which is the most important intelligent link of warehousing and logistics.Basing on an actual project,this paper proposes an solution of intelligent warehousing AGV that can perform target detection and binocular ranging to solve the problems of AGV obstacle avoidance,maintaining warehouse unmanned and the risk of human-computer interaction in the project.Firstly,this paper introduces the application environment and technology status of AGV,and designs the vehicle body structure,motion control mode,positioning mode and obstacle avoidance system to meet the functional requirements of long-term unmanned and harmonious human-machine interaction proposed by AGV in large-scale integrated scheduling warehouse logistics.Secondly,the target detection technology in computer vision is introduced,and the appropriate detection network is selected according to the field operating environment and the function requirements of the vehicle.The training set was obtained by manual labeling according to the actual operation images collected on site.Training to get the model suitable for the AGV operation site in the project,and deployment to the AGV obstacle avoidance system,information preparation for the subsequent obstacle avoidance strategy.The binocular ranging algorithm is implemented according to the binocular ranging theory,and a complete obstacle avoidance strategy is designed based on the content information provided by the target detection.An intelligent obstacle avoidance system which can avoid obstacles independently and distinguish the types of obstacles is proposed.And this system is verified in this project.Finally,the binocular obstacle avoidance AGV was tested in the actual project,and the failure of stereoscopic detection at close range was found.Aiming at the problem,the solution is proposed from both the algorithm and the AGV.Through a new cost function and adding body feature points,the obstacle avoidance scheme was improved..