Cross-Domain UAV Swarm Cooperation based on Decentralized Federated Learning with Hierarchical Aggregation
Low Altitude Economy (LAE) is emerging as a transformative field, with applications in aerial logistics, intelligent surveillance, urban air mobility, and public safety. However, the deployment of large-scale UAV swarms faces challenges such as limited resources, unstable communication links, and data security and privacy concerns. Federated learning, with its distributed advantages, can be considered a key technology to address these issues. We have discussed the current applications of federated learning in UAV swarms and the challenges it faces. We propose a cross-domain hierarchical aggregation federated learning framework. This framework dynamically adjusts the transmission order based on the resource monitoring of drones and adopts a cross-domain heterogeneous distribution setting that is closer to real-life scenarios. It enhances generalization ability through hierarchical model aggregation, effectively improving the resource utilization efficiency of drone swarms. This paper elaborates on the design requirements of the framework, verifies its effectiveness, and outlines future research directions.