Research On Agvautonomous Navigation Technology Based On Visual SLAM

With the rapid development of modern scientific and technological information and intelligence,the level of automation in production and transportation has made unprecedented progress.Automatically guided vehicles integrate the perception of the external environment,intelligent control,and path planning decisions,which have become an indispensable part of modern production logistics.Synchronous positioning and map construction technologies are the key to achieving AGV autonomous positioning and navigation.This article is based on the ROS operating system and uses Xiaoqiang XQ5 as the main carrier.Monocular vision SLAM is applied to the autonomous positioning and navigation of Xiaoqiang XQ5.The algorithm proposed in this article is verified by building the Xiaoqiang XQ5 autonomous navigation experiment platform.The main research work is as follows:1)Monocular vision mileage calculation method design and pose optimization.A monocular vision mileage calculation method with good real-time performance and accuracy is proposed.Firstly,the pre-processing operation is performed on the acquired consecutive inter-frame images,and then the SIFT algorithm and the ORB algorithm are combined to quickly perform feature points processing on the images.,Efficient extraction,and on the basis of the Brute-Force Matcher algorithm and the Fast Library for Ap-proximate Nearest Neighbors algorithm,continue Introduce a Progressive Sampling Consensus algorithm to match and remove mismatched extracted feature points,and then based on the improved nonlinear optimization method,Iterative optimization to solve the least squares equation and accurately estimate the motion pose of the AGV between consecutive frames.Through experimental verification,the monocular visual odometer designed in this paper has higher accuracy and real-time performance,which can greatly reduce mismatching and effectively improve AGV positioning accuracy.2)AGV global path planning.Based on the precise positioning of the AGV by the monocular visual odometer,based on the consideration of the shortest path and the shortest time in the global path planning of the AGV,this paper proposes a global path planning method combining A* algorithm and particle swarm algorithm,The global grid map constructed is used for simulation verification.The results show that the hybrid algorithm proposed in this paper has further improved the running time,number of turns,the number of search nodes,and the total path cost in the global path search process,thereby verifying that the hybrid algorithm proposed in this paper can obtain relatively On the basis of short paths,it also has better real-time performance.3)Set up an outdoor simulation environment and complete the AGV autonomous navigation experiment.Based on the improved monocular visual odometry to accurately locate the AGV,obtain a sparse point cloud map,and continuously iteratively create a monocular dense map,and then based on the integrated A* algorithm and particle swarm algorithm based on this monocular dense map,Carry out AGV global path planning,finally based on the ROS operating system,using Xiaoqiang XQ5 as a carrier,and verify it by building a simulation experiment platform in a real scenario,The monocular vision SLAM studied in this paper has better robustness in AGV autonomous navigation.