Intelligent 3D Trajectory and Resource Control for Multi-UAV 6G Networks via GNN and Deep Unfolding

This paper investigates a multi-UAV downlink communication system, where multiple UAVs are deployed to serve ground users who move randomly within the coverage area. The primary objective is to maximize the users’ sum-rate over a specified time frame, by jointly optimizing UAV 3D trajectories, user association, channel estimation, and power allocation. To address this challenging problem, we introduce a novel deep learning-based solution. Specifically, a graph neural network (GNN) is employed to model and learn the interaction patterns between UAVs and users, yielding optimized 3D trajectories and user association strategies. Additionally, a deep unfolding network (DUN) is used to perform efficient channel estimation and power allocation with improved convergence efficiency. Simulation results demonstrate that the proposed scheme achieves superior communication performance while significantly increasing convergence speed.