Scholarlink

<p>Wavelength division multiplexing (WDM) free space optical (FSO) systems enable high-speed wireless communication but face significant challenges from atmospheric attenuation, especially under adverse weather. To ensure reliable and energy-efficient WDM-FSO communication, this paper introduces a supervised machine learning (ML) approach for predicting optimal emission power tailored to specific weather conditions. Using decision trees (DT), support vector machines (SVM), and artificial neural networks (ANN), the method enables adaptive power control to mitigate signal degradation. A comprehensive dataset was generated by simulating an 8-channel WDM-FSO system at 1550&#xa0;nm, incorporating key parameters such as atmospheric attenuation, transmission distance, quality factor (Q-factor), and bit error rate (BER). Model performance was evaluated using root mean square error (RMSE) and the coefficient of determination (R²). The ANN model was the most accurate, achieving an R² of 0.99 and an RMSE of 0.71, outperforming SVM (R²=0.97, RMSE = 2.06) and DT (R²=0.96, RMSE = 2.48). These results show the effectiveness of ML-based methods in optimizing emission power to counteract atmospheric effects, ensuring reliable and energy-efficient communication across varying conditions. The findings highlight the potential of ML techniques to enable fast, adaptive prediction toward real-time power control, supporting the development of resilient and high-performance WDM-FSO 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

Machine learning-based, weather-aware optimal emission power prediction for reliable and energy-efficient WDM -FSO communication

  • 期刊论文

  • 发表日期：

    2026年7月31日

查看全文
  • Kebaili Rima,
  • Driz Samia,
  • Fassi Benattou

摘要

Wavelength division multiplexing (WDM) free space optical (FSO) systems enable high-speed wireless communication but face significant challenges from atmospheric attenuation, especially under adverse weather. To ensure reliable and energy-efficient WDM-FSO communication, this paper introduces a supervised machine learning (ML) approach for predicting optimal emission power tailored to specific weather conditions. Using decision trees (DT), support vector machines (SVM), and artificial neural networks (ANN), the method enables adaptive power control to mitigate signal degradation. A comprehensive dataset was generated by simulating an 8-channel WDM-FSO system at 1550 nm, incorporating key parameters such as atmospheric attenuation, transmission distance, quality factor (Q-factor), and bit error rate (BER). Model performance was evaluated using root mean square error (RMSE) and the coefficient of determination (R²). The ANN model was the most accurate, achieving an R² of 0.99 and an RMSE of 0.71, outperforming SVM (R²=0.97, RMSE = 2.06) and DT (R²=0.96, RMSE = 2.48). These results show the effectiveness of ML-based methods in optimizing emission power to counteract atmospheric effects, ensuring reliable and energy-efficient communication across varying conditions. The findings highlight the potential of ML techniques to enable fast, adaptive prediction toward real-time power control, supporting the development of resilient and high-performance WDM-FSO networks.