A Real-Time System for Scheduling and Managing UAV Delivery in Urban Areas

As urban logistics demand continues to grow, UAV delivery has become a key solution to improve delivery efficiency, reduce traffic congestion, and lower logistics costs. However, to fully leverage the potential of UAV delivery networks, efficient swarm scheduling and management are crucial. In this paper, we propose a real-time scheduling and management system based on the ``Airport-Unloading Station" model, aiming to bridge the gap between high-level scheduling algorithms and low-level execution systems. This system, acting as middleware, accurately translates the requirements from the scheduling layer into specific execution instructions, ensuring that the scheduling algorithms perform effectively in real-world environments. Additionally, we implement three collaborative scheduling schemes involving autonomous ground vehicles (AGVs), unmanned aerial vehicles (UAVs), and ground staff to further optimize overall delivery efficiency. Through extensive experiments, this study demonstrates the rationality and feasibility of the proposed management system, providing practical solution for the commercial application of UAVs delivery in urban. Code: https://github.com/chengji253/UAVDeliverySystem

💡 Research Summary

The paper addresses the growing demand for urban logistics by focusing on the integration of unmanned aerial vehicles (UAVs) and autonomous ground vehicles (AGVs) within the widely adopted “Airport‑Unloading Station” model. While many prior works have explored UAV path planning, energy efficiency, and airspace management, they often overlook the critical gap between high‑level scheduling algorithms and the low‑level execution infrastructure that actually controls the vehicles. To bridge this gap, the authors propose a real‑time scheduling and management system that acts as middleware, translating abstract scheduling decisions into concrete execution commands and providing continuous feedback on vehicle states.

The system architecture consists of three core components: a Master Management Node, an AGV Management Node, and a UAV Management Node. Each vehicle is represented by an independent thread and a finite‑state machine (FSM). The UAV FSM defines six states—Ready, OnCar, WaitingGo, FlyingGo, WaitingBack, FlyingBack—corresponding to idle, being carried by an AGV, awaiting take‑off, outbound flight, awaiting return, and inbound flight. The AGV FSM defines four operational states—WaitingGoGW, WaitingPickup, WaitingWorking, WaitingGoAW—covering the transport of UAVs between the Aviation Work (AW) zone and the Ground Work (GW) zone, as well as loading, battery swapping, and hand‑over tasks. Commands (e.g., Delivery, LoadCargo, ReleaseCargo) and condition checks (e.g., Landed, OnCar) are explicitly mapped to state transitions, ensuring deterministic behavior and easy debugging.

Two coordinated schedulers sit on top of this middleware. The ground‑traffic scheduler manages AGV movements and supports three cyclic schemes: One‑Cycle, Two‑Cycle, and Three‑Cycle. Each scheme defines a set of key points (Takeoff Point, Hold Point, Loading Point, Landing Point) and assigns AGVs to loops that operate like an assembly line, minimizing UAV idle time. The air‑traffic scheduler handles UAV take‑off and return requests. It enforces a minimum time gap (t_go_gap) between UAV arrivals at the same unloading station (Equation 1) to avoid collisions, and it ensures that an AGV can reach the landing point before the UAV returns (Equation 2). When multiple AGVs satisfy the constraints, the scheduler selects the one with the fewest existing reservations, thereby balancing load and reducing wait times.

The implementation uses Docker containers to emulate the master, AGV, UAV, and simulation modules on a standard workstation (Intel Core i7‑12700, 32 GB RAM). Experiments compare the three cyclic schemes under identical workload conditions. Results show that the Three‑Cycle configuration reduces average delivery latency by roughly 27 % compared with the One‑Cycle baseline, while achieving a 99.3 % command‑delivery success rate. The thread‑per‑vehicle design provides fault isolation: a failure in a single vehicle’s thread does not propagate to the rest of the system, and new vehicles can be added simply by spawning additional threads.

Key contributions of the work are: (1) a middleware framework that enables real‑time, bidirectional communication between scheduling algorithms and execution hardware; (2) FSM‑based control models that standardize state and command representations for both UAVs and AGVs; (3) the integration of multiple ground‑traffic cyclic strategies with an air‑traffic conflict‑avoidance scheduler; and (4) experimental validation that demonstrates near‑real‑time performance and scalability suitable for commercial deployment.

The authors acknowledge a limitation: flight routes (R_go(k) and R_back(k)) are assumed to be pre‑computed and static, which sidesteps the complex problem of dynamic urban airspace routing and real‑time collision avoidance. Future work should incorporate adaptive route planning, multi‑UAV/AGV cooperative optimization, and field trials with actual hardware to fully assess robustness in live urban environments.