Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Free-Space Optical Channel Turbulence Prediction: A Machine Learning Approach

Md Zobaer Islam, Ethan Abele, Fahim Ferdous Hossain, Arsalan Ahmad, Sabit Ekin, John F. O'Hara

TL;DR

This work addresses the challenge of predicting free-space optical channel turbulence without extra sensing hardware by applying machine learning to raw optical data. Using a laboratory turbulence chamber with six levels, the authors train an XGBoost classifier to map received signal characteristics to turbulence levels, achieving accuracies exceeding $98\%$ when turbulence has stabilized within about 1 minute. The study analyzes how data duration and the number of training instances influence performance and highlights time-scale dynamics of turbulence that affect prediction during transients. The findings suggest practical applications such as a "+communication weather map" for adaptive beam management and network-level decisions, while also noting limitations in not directly quantifying $C_n^2$ and proposing avenues to link ML outputs with physical turbulence metrics for real-world deployment.

Abstract

Channel turbulence is a formidable obstacle for free-space optical (FSO) communication. Anticipation of turbulence levels is highly important for mitigating disruptions but has not been demonstrated without dedicated, auxiliary hardware. We show that machine learning (ML) can be applied to raw FSO data streams to rapidly predict channel turbulence levels with no additional sensing hardware. FSO was conducted through a controlled channel in the lab under six distinct turbulence levels, and the efficacy of using ML to classify turbulence levels was examined. ML-based turbulence level classification was found to be >98% accurate with multiple ML training parameters. Classification effectiveness was found to depend on the timescale of changes between turbulence levels but converges when turbulence stabilizes over about a one minute timescale.

Free-Space Optical Channel Turbulence Prediction: A Machine Learning Approach

TL;DR

This work addresses the challenge of predicting free-space optical channel turbulence without extra sensing hardware by applying machine learning to raw optical data. Using a laboratory turbulence chamber with six levels, the authors train an XGBoost classifier to map received signal characteristics to turbulence levels, achieving accuracies exceeding when turbulence has stabilized within about 1 minute. The study analyzes how data duration and the number of training instances influence performance and highlights time-scale dynamics of turbulence that affect prediction during transients. The findings suggest practical applications such as a "+communication weather map" for adaptive beam management and network-level decisions, while also noting limitations in not directly quantifying and proposing avenues to link ML outputs with physical turbulence metrics for real-world deployment.

Abstract

Channel turbulence is a formidable obstacle for free-space optical (FSO) communication. Anticipation of turbulence levels is highly important for mitigating disruptions but has not been demonstrated without dedicated, auxiliary hardware. We show that machine learning (ML) can be applied to raw FSO data streams to rapidly predict channel turbulence levels with no additional sensing hardware. FSO was conducted through a controlled channel in the lab under six distinct turbulence levels, and the efficacy of using ML to classify turbulence levels was examined. ML-based turbulence level classification was found to be >98% accurate with multiple ML training parameters. Classification effectiveness was found to depend on the timescale of changes between turbulence levels but converges when turbulence stabilizes over about a one minute timescale.
Paper Structure (6 sections, 4 figures, 3 tables)

This paper contains 6 sections, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. System Design and Implementation
  3. Data Collection
  4. Turbulence Prediction
  5. Discussion
  6. Conclusion

Figures (4)

  • Figure 1: (a) System diagram showing the general connections between instruments and turbulence flow through the optical path, (b) View of the lab hallway with four turbulence generators marked
  • Figure 2: Received data plots from the beginning of the first files of turbulence levels 0 and 5.
  • Figure 3: Turbulence classification test accuracy scores for different files with zero stabilization time after activation of the turbulence generators (data were taken from the beginning of each file).
  • Figure 4: Turbulence classification test accuracy scores for different files with 10 minute stabilization time after activation of the turbulence generators (data were taken from the beginning of each file).