Scholarlink

<p>The rapid evolution of wireless technologies and the automotive industry have positioned Intelligent Vehicular Networks (IVNs) as a fundamental pillar of Intelligent Transportation Systems (ITS). This paper presents a comprehensive methodological framework for IVNs clustering using advanced Graph Neural Network (GNN) architectures. We address critical challenges such as cluster stability, connectivity duration, and accuracy in highly dynamic highway environments. Our improved GNN model surpasses traditional clustering methods, including k-means, spectral clustering, and graph auto encoders, achieving 99.4% accuracy while maintaining an average cluster connectivity duration of 890&#xa0;s (nearly 15&#xa0;min). This performance was observed in a realistic simulation environment modeling a 5&#xa0;km highway segment with 3 lanes, a vehicle density of 150, and a communication range of 200&#xa0;m, reflecting dynamic traffic conditions. The proposed approach integrates realistic mobility models based on PeMS database characteristics and introduces new evaluation metrics for temporal cluster stability analysis. Experimental results demonstrate significant improvements in cluster lifetime, stability, and connectivity compared to state-of-the-art techniques.</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

Improving clustering stability in intelligent vehicular networks using Graph Neural Networks

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月11日

查看全文
  • Hana Nouri,
  • Tarek Moulahi,
  • Salah Dhahri,
  • Salim El khediri

摘要

The rapid evolution of wireless technologies and the automotive industry have positioned Intelligent Vehicular Networks (IVNs) as a fundamental pillar of Intelligent Transportation Systems (ITS). This paper presents a comprehensive methodological framework for IVNs clustering using advanced Graph Neural Network (GNN) architectures. We address critical challenges such as cluster stability, connectivity duration, and accuracy in highly dynamic highway environments. Our improved GNN model surpasses traditional clustering methods, including k-means, spectral clustering, and graph auto encoders, achieving 99.4% accuracy while maintaining an average cluster connectivity duration of 890 s (nearly 15 min). This performance was observed in a realistic simulation environment modeling a 5 km highway segment with 3 lanes, a vehicle density of 150, and a communication range of 200 m, reflecting dynamic traffic conditions. The proposed approach integrates realistic mobility models based on PeMS database characteristics and introduces new evaluation metrics for temporal cluster stability analysis. Experimental results demonstrate significant improvements in cluster lifetime, stability, and connectivity compared to state-of-the-art techniques.