Modeling of Electromagnetic Wave Propagation through Turbulent Atmosphere using Machine Learning Techniques
Advances in Nonlinear Variational Inequalities · 27.12.2023 · 26 (3) · 74-82
DOI: 10.52783/anvi.v26.7288 ↗Kurz gefasst
Fachartikel in der Zeitschrift Advances in Nonlinear Variational Inequalities, online erschienen am 27.12.2023. Verfasst von Quaiser Hashmi.
Abstract (Originalsprache)
Electromagnetic (EM) wave propagation through the turbulent atmosphere is significantly affected by random fluctuations of the refractive index caused by variations in temperature, pressure, and wind velocity along the propagation path. These fluctuations give rise to beam wander, scintillation, beam spreading, and angle-of-arrival fluctuations, which severely degrade the performance of free-space optical (FSO) and radio-frequency (RF) communication links. Conventional analytical approaches, based on the Kolmogorov spectrum and the Rytov approximation, provide reasonable estimates under weak turbulence but become computationally intensive and less accurate under moderate-to-strong turbulence regimes. In this work, a hybrid physics-informed machine learning (ML) framework is proposed to model and predict the scintillation index of an EM wave propagating through a turbulent atmospheric channel. A synthetic dataset of 1500 samples was generated using the Hufnagel–Valley 5/7 refractive index structure parameter (Cn2) model coupled with Rytov-variance-based scintillation theory, incorporating stochastic turbulence-induced noise to emulate realistic channel conditions. Four supervised ML regressors — Random Forest (RF), Gradient Boosting (GB), Support Vector Regression (SVR), and an Artificial Neural Network (ANN) — were trained and benchmarked using wavelength, link distance, Cn2, wind speed, and elevation angle as input features. Simulation results indicate that the Support Vector Regression model achieved the lowest root-mean-square error (RMSE = 0.1125) and the highest coefficient of determination (R2 = 0.752) among the tested models, outperforming the Random Forest and Gradient Boosting regressors on this dataset. The proposed ML-assisted framework offers a computationally efficient and data-driven alternative to purely analytical turbulence models and can be extended for real-time adaptive optics and link-margin estimation in FSO and satellite communication systems.
Zitieren
APA
Hashmi, Q. (2023). Modeling of Electromagnetic Wave Propagation through Turbulent Atmosphere using Machine Learning Techniques. Advances in Nonlinear Variational Inequalities, 26(3), 74-82. https://doi.org/10.52783/anvi.v26.7288
Vancouver
Hashmi Q. Modeling of Electromagnetic Wave Propagation through Turbulent Atmosphere using Machine Learning Techniques. Advances in Nonlinear Variational Inequalities. 2023;26(3):74-82. doi:10.52783/anvi.v26.7288
BibTeX
@article{Hashmi2023,
author = {Hashmi, Quaiser},
title = {Modeling of Electromagnetic Wave Propagation through Turbulent Atmosphere using Machine Learning Techniques},
journal = {Advances in Nonlinear Variational Inequalities},
year = {2023},
volume = {26},
number = {3},
pages = {74--82},
doi = {10.52783/anvi.v26.7288}
}
EUVIMED