Research On AGV Autonomous Localization Technology Based On Lidar And Vision

As a kind of intelligent handling equipment,Automated Guided Vehicle(AGV)has been widely used in the manufacturing industry.At present,the relatively mature AGV localization guidance methods all depend on artificial beacon,such as magnetic nail,magnetic tape,reflector and QR code.However,there are problems such as low degree of flexibility,a high cost and poor autonomy,and working requirements in the complex environment of factory cannot be met.Therefore,the AGV self-localization guidance method with a low cost and high degree of flexibility in the complex environment of factory has become the development trend of the industry.Simultaneous Localization and Mapping(SLAM)technology is the key technology for robots to realize self-localization.It only uses its own sensor and does not need to arrange any artificial beacon.With high degree of flexibility of the system,it has become the research hotspot in the field of robot.This thesis studies the realization of self-localization of AGV by SLAM technology;analyses problems such as incomplete and inaccurate mapping,low localization accuracy and re-localization of AGV when AGV uses SLAM technology in the complex environment;and gives the corresponding improvement method.The main research contents of this thesis are as follows:(1)There is less data information of 2D laser radar,resulting in incomplete and inaccurate map.To solve this problem,this thesis gives the SLAM algorithm integrating laser and vision,which improves the integrity and accuracy of map.By integrating data of RGB-D camera and laser radar as data input of the algorithm,closed-loop points are rapidly judged with visual information,and the map is optimized simultaneously.The experiment results show that in the 3m×5m experiment scene the map constructed by the algorithm in this thesis clearly shows barriers in the environment,which improves the integrity of map.In the 7m×13m experiment scene in the process of mapping by the algorithm in this thesis closed-loop points are judged immediately,the map is optimized and the accuracy of the map is improved.(2)In order to further improve the localization accuracy of SLAM algorithm in the complex environment,the pose graph optimization method of adding constraints based on rich information is given.The error function is constructed by laser radar matching error and visual feature point matching error,the error function is minimized and the optimal pose estimation is obtained.The experiment results show that the localization accuracy of SLAM algorithm in this thesis is within 5cm.(3)In the process of visual re-localization,the path is relied on too much,so it is very difficult to realize global re-localization.To solve this problem,this thesis gives an improvement method.By combining the visual re-localization method based on bag-of-words(BoW)model and Adaptive Monte Carlo Localization(AMCL)method,the global re-localization is realized.The results of re-localization are used for iterative calculation as the initial values of the Iterative Closet Point(ICP)algorithm to further improve the localization accuracy.The experiment results show that compared with the re-localization method based on ORB-SLAM2 the recall rate of the method in this thesis increases by 40.9%,the translation error is 3cm-5cm,the rotation error is 1°-3°,the accuracy obviously increases,and the reusability of environment map is effectively improved.To sum up,this thesis studies SLAM technology in terms of mapping and localization and AGV in terms of re-localization.The experiment results show the feasibility of the method in this thesis,lays a good foundation for solution selection and rapid development of the use of SLAM technology by AGV in the complex working environment,and provides a basis for the reuse of environmental map.