Scholarlink

<p>This paper addresses a critical gap in Unmanned Aerial Vehicle (UAV)-assisted Internet of Things (IoT) networks, where existing works inadequately integrate UAV deployment optimization with privacy-preserving Federated Learning (FL) and adaptive resource allocation under dynamic network conditions. The research explores the deployment of multi-UAV networks in IoT environments, emphasizing their dual roles in expanding cellular network coverage and facilitating efficient data collection. Unlike prior studies that treat UAV placement and FL-driven resource optimization separately, we present a unified hybrid framework leveraging Deep Reinforcement Learning (DRL) and FL. The proposed framework incorporates the Multi-UAV Network Formation (MUNF) algorithm, which employs Particle Swarm Optimization (PSO) to improve the Signal-to-Noise Ratio (SNR) for effective data collection. Additionally, the Dynamic Adaptive Strategy (DAS) utilizes a Deep Deterministic Policy Gradient (DDPG) approach to optimize resource allocation, reduce latency, and enhance throughput. Extensive simulations demonstrate a 26% increase in data throughput, an 18% reduction in latency, and more stable SNR distribution compared to state-of-the-art baselines. These results indicate a consistent improvement in network efficiency and scalability, validating the proposed framework’s capability to address real-world UAV-assisted IoT challenges more effectively than prior work.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

UAV-asisted IoT network framework with hybrid deep reinforcement and federated learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月23日

查看全文
  • Andreas Andreou,
  • Constandinos X. Mavromoustakis,
  • Evangelos Markakis,
  • Athina Bourdena,
  • George Mastorakis

摘要

This paper addresses a critical gap in Unmanned Aerial Vehicle (UAV)-assisted Internet of Things (IoT) networks, where existing works inadequately integrate UAV deployment optimization with privacy-preserving Federated Learning (FL) and adaptive resource allocation under dynamic network conditions. The research explores the deployment of multi-UAV networks in IoT environments, emphasizing their dual roles in expanding cellular network coverage and facilitating efficient data collection. Unlike prior studies that treat UAV placement and FL-driven resource optimization separately, we present a unified hybrid framework leveraging Deep Reinforcement Learning (DRL) and FL. The proposed framework incorporates the Multi-UAV Network Formation (MUNF) algorithm, which employs Particle Swarm Optimization (PSO) to improve the Signal-to-Noise Ratio (SNR) for effective data collection. Additionally, the Dynamic Adaptive Strategy (DAS) utilizes a Deep Deterministic Policy Gradient (DDPG) approach to optimize resource allocation, reduce latency, and enhance throughput. Extensive simulations demonstrate a 26% increase in data throughput, an 18% reduction in latency, and more stable SNR distribution compared to state-of-the-art baselines. These results indicate a consistent improvement in network efficiency and scalability, validating the proposed framework’s capability to address real-world UAV-assisted IoT challenges more effectively than prior work.