Abstract:
Equipped with airborne base stations, low-altitude unmanned aerial vehicle (UAV) swarms feature rapid deployment, high mobility, and wide coverage, enabling them to quickly provide low-altitude wireless access services for affected ground personnel and rescuers in post-disaster rescue scenarios to seize the golden rescue window. Constrained by limited battery capacity, the energy efficiency (EE) of low-altitude UAV swarms directly determines the sustainable operation time of emergency communication missions. Improving the EE of UAV swarms is critical for extending communication duration and expanding service coverage in disaster emergency communication. This paper adopts graph neural networks (GNNs) and deep reinforcement learning (DRL) to optimize resource coordination in emergency communication systems and maximize the swarm energy efficiency. Specifically, GNNs aggregate observation and topological information from UAVs and ground users, and establish the mapping between such information and emergency communication resource coordination strategies; DRL is utilized to learn the optimal resource coordination policy for each UAV and user. Simulation results verify that the proposed method can significantly boost the energy efficiency of the UAV-enabled emergency communication system, realizing energy-efficient resource coordination for low-altitude UAV swarms.