DeFeL-FUN: Decentralized Federated Learning Framework for UAV-Enabled Networks

Unmanned Aerial Vehicles (UAVs) are envisioned as vital components of next-generation networks, driving diverse applications across various domains. UAV-enabled networks typically encompass diverse edge devices, including UAVs, ground stations, sensors, and other IoT devices, all generating vast amounts of data. Hence, in recent years, there has been a significant surge of interest in federated learning (FL) within UAV-enabled networks. However, most of the approaches in the state-of-the-art suffer from a single point of failure as well as massive communication overhead and excessive delay due to the existence of a central aggregator. To tackle this shortcoming, this paper proposes a framework for decentralized FL in UAV-enabled networks. Our approach consists of two parts; i) First, we propose a UAV placement scheme while ensuring the connectivity of the network of deployed UAVs. ii) Next, we use a consensus-based approach for decentralized FL among the UAV agents. We verify the effectiveness of our approach via extensive numerical simulation.