Design And Realization Of Positioning System Based On Vision For AGV In Automated Warehouse

Fiducial markers have become the preferred method for pose estimation in the applications of AGV(Automatic Guided Vehicle)in automated warehouse.Sticking to the ground of the warehouse,fiducial tags are not affected by surroundings and the top,suitable for 2D matrix layout,and a single tag can provide global positioning for AGV.The AGV positioning system consists of three positioning elements: fiducial tags,camera,grid map,and three pose transformations between fiducial tags and grid map,fiducial tags and camera,camera and vehicle center.The positioning system of the warehouse AGV is consist of the three positioning elements and the three pose transformations.In applications,warehouse AGV positioning system need to be evaluated and improved,since imperfect positioning elements and inaccurate pose transformations.In this paper,the warehouse AGV positioning system is studied systematically.Firstly,a kind of fiducial tag called Grid Tag is designed for the warehouse scenario.The total number of Grid Tag is more than 260,000 and it has two-dimensional data information,which overcomes the flexibility problem and the shortage in number.Although fiducial tag's detection are easily influenced by dust,debris shielding and uneven light,Grid Tag using large black and white grids,and using the original data and inverse data verification mechanism,have robust detection and recognition.In addition,a fast detection and recognition algorithm of Grid Tag is developed.The image downsampling process and image restoration strategies are adopted to ensure the detection rate and the accuracy of pose estimation.Secondly,the original grid map of tags need every tag to be attached accurately,it's a big work,and the accuracy of grid map is difficult to guarantee.This paper proposed a grid map creation method using the optimization of nonlinear least squares,and also proposed a map data acquisition method using warehouse AGV.So that the accuracy of tag attaching can be relaxed,also greatly reduced the attaching work.Thirdly,to make the camera position calibration process even simpler,according the kinematics model of the warehouse AGV,a simple and easy method for the calibration of Grid Tag sensor is proposed.Finally,the vehicle control rate equation was established,and the reliability and stability was verified by the vehicle running test under multiple paths.