Efficient Federal Learning in USV-AUVs Collaborative Networks

Clusters of Unmanned Surface Vehicles (USVs) and Autonomous Underwater Vehicles (AUVs) can operate underwater while providing critical data support for machine learning. To address the challenges of data privacy protection, wireless resource constraints, and energy efficiency, we propose a Federated Learning (FL) framework for USVCAUV Collaborative Networks (UACNs) and introduce, for the first time, Reconfigurable Intelligent Surface (RIS)-assisted multi-hop Underwater Wireless Optical Communication (UWOC) technology. Building on this foundation, we further propose a Proximal Policy Optimization (PPO)-based resource scheduling and relay selection strategy designed to minimize communication time and energy consumption while enhancing training efficiency and learning accuracy in federated learning. Finally, simulation results validate the effectiveness and superiority of the proposed FL framework and strategy.