Introduction
The integration of Unmanned Aerial Vehicles (UAVs) into urban logistics promises significant efficiency gains. However, a critical "execution gap" exists between high-level scheduling algorithms and the low-level physical execution of drone missions. Real-world operations require complex coordination between UAVs, Automated Guided Vehicles (AGVs), and ground personnel. This paper presents a real-time middleware system designed to bridge this gap by synchronizing multi-agent operations through a hierarchical architecture and state-aware management.
System Architecture
The proposed system acts as a middleware layer between high-level schedulers and physical actuators. As shown in the hierarchical diagram, the architecture is organized into three distinct layers:
- Scheduling System: Contains the Ground Traffic Scheduler and Air Traffic Scheduler, which handle high-level path planning and window allocation.
- Management System: The core of the middleware, featuring a Master Management Node that coordinates specialized AGV and UAV Management Nodes. These nodes utilize multithreading and Finite State Machines (FSMs) to ensure real-time responsiveness and fault isolation.
- Actuators: The low-level control layer for individual AGVs and UAVs.
Ground Traffic and Cyclic Schemes
A key innovation of this system is the use of AGVs as safety buffers, separating the "Ground Work" (GW) area where humans load cargo from the "Aviation Work" (AW) area where flight operations occur. To optimize the movement of drones within the "airport" facility, the authors evaluated three cyclic ground traffic schemes: One-Cycle, Two-Cycle, and Three-Cycle loops.
As illustrated in the cyclic scheme comparison, these configurations vary in complexity and resource allocation. The One-Cycle scheme uses a single loop, while the Two-Cycle scheme employs two loops sharing a central takeoff point. The Three-Cycle scheme further divides the area into three parallel lines.
State-Aware Coordination
The system ensures operational safety and efficiency through synchronized Finite State Machines (FSMs). Both UAVs and AGVs follow strict state transitions (e.g., Waiting_Pickup to On_Car to Flying_Go). This bi-directional communication allows the Air Traffic Scheduler to query the real-time state of ground resources. For instance, a returning UAV is only granted landing permission if the system confirms an AGV will be available at the landing point before the drone's arrival, effectively preventing aerial bottlenecks.
Experimental Results
The system was evaluated using a custom simulation platform built on Docker and ROS. Key findings include:
- Optimal Throughput: The Two-Cycle scheme emerged as the most efficient configuration. It provided the optimal balance between ground staff utilization and drone throughput. While the Three-Cycle scheme reduced individual waiting times, the increased travel distance for AGVs led to lower overall system efficiency.
- Scalability: Testing with 4 to 16 UAVs showed that while throughput increases with the number of drones, the system eventually reaches a plateau dictated by the physical capacity of the "airport" infrastructure.
- Performance Metrics: The system maintained an average delivery score of approximately 80 points, demonstrating high effectiveness despite the inherent constraints of urban flight environments.
Conclusion
This research provides a robust framework for managing the complex logistics of urban drone delivery. By treating ground transport as a critical constraint of the flight path and utilizing a hierarchical FSM-based management system, the "execution gap" in drone logistics can be effectively narrowed. Future work will focus on integrating dynamic 4D trajectory management and scaling the system to coordinate multiple interconnected delivery hubs across a city.
Summary updated