Multi-Uav Data Collection Trajectory Optimization in Three-Dimensional Urban Environments: A Federated Reinforcement Learning Framework

With the acceleration of smart city construction, unmanned aerial vehicles (UAVs) have demonstrated tremendous potential in the field of Internet of Things (IoT) data collection. Although trajectory planning has been extensively researched, learning-based cooperative trajectory planning for multiple UAVs is often limited to planar designs, making it difficult to adapt to complex urban terrains with high-rise buildings. This paper extends learning-based multi-UAV trajectory planning from two-dimensional planes to three-dimensional space, improving data collection efficiency through vertical maneuverability capabilities and precise three-dimensional urban modeling. Comparative experiments show that compared to traditional two-dimensional methods, our approach achieves approximately 10.2% improvement in data collection rate while reducing energy consumption by around 15.0%. These results provide a more practical technical approach for multi-UAV collaborative data collection in urban scenarios.