Multi-AGV Tracking System Based on Global Vision and AprilTag in Smart Warehouse

With the development of smart warehouses in Industry 4.0, scheduling a fleet of automated guided vehicles (AGVs) for transporting and sorting parcels has become a new development trend. In smart warehouses, AGVs receive paths from the multi-AGV scheduling system and independently sense the surrounding environment while sending poses as interactive information. This navigation method relies heavily on on-board sensors and significantly increases the information interactions within the system. Under this situation, a solution that locates multiple AGVs in global images of the warehouse by top cameras is expected to have a great effect. However, traditional tracking algorithms cannot output the heading angles required by the AGV navigation and their real-time performance and calculation accuracy cannot satisfy the tracking of large-scale AGVs. Therefore, this paper proposes a multi-AGV tracking system that integrates a multi-AGV scheduling system, AprilTag system, improved YOLOv5 with the oriented bounding box (OBB), extended Kalman filtering (EKF), and global vision to calculate the coordinates and heading angles of AGVs. Extensive experiments prove that in addition to less time complexity, the multi-AGV tracking system can efficiently track a fleet of AGVs with higher positioning accuracy than traditional navigation methods and other tracking algorithms based on various location patterns.