Simulation of Machine Vision AGV Autonomous Navigation Based on Monocular SLAM

With the promotion and application of unmanned delivery, the navigation method of Automated Guided Vehicle (AGV) has become a hot research topic, especially in the research of AGV navigation technology combined with machine vision system. This article introduces an AGV navigation method based on monocular visual SLAM technology. This method uses a monocular camera to collect environmental images and constructs a 2D grid map in real time through the SLAM algorithm to achieve vehicle positioning and navigation functions. The SLAM technology used in this method is ORB-SLAM, which uses the ORB algorithm and Bag of Words (BOW) model for visual feature description and matching, describing and matching feature points through visual word bags to further improve matching efficiency and accuracy. ORB-SLAM performs pose estimation and map construction based on the Bundle Adjustment (BA) graph optimization algorithm, optimizing the pose of all key frames and the position of map points by minimizing the reprojection error, further improving the accuracy and robustness of the map. Finally, the scene map construction, positioning, and autonomous navigation functions of the AGV were simulated and tested based on ROS, and the experimental results show its effectiveness.