Federated Learning for Complex AI Models in multi-UAV Systems Assisted by Mobile Edge Servers

This paper proposes a federated learning (FL)-assisted multi-UAV system incorporated with an intermediate mobile edge server (MES) to enhance the computational capabilities of UAVs. This system architecture allows UAVs to utilize complex AI models, such as YOLO, for object detection, improving decision-making and adaptability in dynamic environments. The proposed system includes an asynchronous cooperative network for continuous model updates, facilitated by MESs positioned near UAVs to reduce energy consumption and improve efficiency. Experimental results demonstrate the feasibility and effectiveness of this approach, as well as the improvements in training and inference speeds when leveraging MESs, suggesting that the proposed method can provide UAVs with improved accuracy, enhanced decision-making, and better adaptability to dynamic environments.