Semantic Feature-Based Visual Automated Guided Vehicles Navigation and Localization System

Automated Guided Vehicles (AGV) serve as effective tools in the manufacturing industry, addressing different challenges in the procedures of intelligent loading, unloading, and autonomous goods distribution. To meet the demands of factory environments and practical applications, vehicles need to achieve high-precision localization within cost-effective constraints. Currently advanced AGV natural navigation mostly uses expensive LiDAR to navigate, sometimes fusing other sensors. On the other side, traditional visual-based navigation approaches, although cheaper, face many challenges such as texture deficiency, illumination changes, and pattern repetition. In this paper, we present a navigation and control method for AGV movement, using an optimized global map, upon which a high-precise visual odometry is constructed. The system presented in this paper encompass vehicle localization, navigation, and motion control. Experiments are carried out to verify the effectiveness of the proposed system. Our experimental results show that the system works well in a real scene (in size of 14mX20m), achieving centimeter-level localization accuracy with smaller than 3 degree average angular error.