Scholarlink

<p>Terahertz (THz) satellite communications face challenges due to atmospheric attenuation and distortion caused by molecular interactions, including absorption and scattering. This research introduces a novel Physics-informed deep learning (PIDL) framework for adaptive atmospheric compensation in terahertz (0.1–10 THz) satellite communication networks. The framework integrates Maxwell’s equations for electromagnetic propagation and HITRAN-derived molecular absorption models directly into neural network architectures to overcome THz signal degradation caused by atmospheric attenuation and group velocity dispersion (GVD). This enables real-time adaptation to atmospheric variations, increasing the data transmission rate and compensating for atmospheric effects, making THz satellite systems feasible. Through extensive validation using 225,000 atmospheric samples and real satellite data in various climatic conditions, the framework achieves an unprecedented performance of &gt; 95% GVD compensation, sub-millisecond adaptation latency (5–15X faster compared to existing methods), 40% energy efficiency improvement, and 1.48 terabit per second (TB/s) average data rate with zero-shot generalisation to unseen atmospheric conditions. Utilising global atmospheric profiles and experimental satellite measurements, PIDL enables real-time physical consistency without external reference, thereby opening the door to THz satellite deployment for 6G and beyond wireless networks.</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

Physics-Informed Deep Learning for Adaptive Atmospheric Compensation in Terahertz Satellite Communication Networks

  • 期刊论文

  • 发表日期：

    2026年1月7日

查看全文
  • Bhagavan Konduri,
  • K. Chanthirasekaran,
  • Abhay Chaturvedi,
  • Anil Kumar Muthevi,
  • Suraya Mubeen,
  • Elangovan Muniyandy

摘要

Terahertz (THz) satellite communications face challenges due to atmospheric attenuation and distortion caused by molecular interactions, including absorption and scattering. This research introduces a novel Physics-informed deep learning (PIDL) framework for adaptive atmospheric compensation in terahertz (0.1–10 THz) satellite communication networks. The framework integrates Maxwell’s equations for electromagnetic propagation and HITRAN-derived molecular absorption models directly into neural network architectures to overcome THz signal degradation caused by atmospheric attenuation and group velocity dispersion (GVD). This enables real-time adaptation to atmospheric variations, increasing the data transmission rate and compensating for atmospheric effects, making THz satellite systems feasible. Through extensive validation using 225,000 atmospheric samples and real satellite data in various climatic conditions, the framework achieves an unprecedented performance of > 95% GVD compensation, sub-millisecond adaptation latency (5–15X faster compared to existing methods), 40% energy efficiency improvement, and 1.48 terabit per second (TB/s) average data rate with zero-shot generalisation to unseen atmospheric conditions. Utilising global atmospheric profiles and experimental satellite measurements, PIDL enables real-time physical consistency without external reference, thereby opening the door to THz satellite deployment for 6G and beyond wireless networks.