Turbulence-aware free-space optical communication system hybrid CNN–LSTM channel modeling, adaptive prediction, and real-time mitigation

Atmospheric optical turbulence severely degrades the performance of free-space optical communication. Mitigating these effects without costly additional hardware remains an open challenge. This paper presents a fully machine learning–driven adaptive modulation and coding architecture that combines a first-principles neural surrogate model with a hybrid deep learning predictor to optimize the modulation and coding scheme. A feed-forward neural network, trained on exact Kolmogorov phase-screen statistics, learns the differentiable mapping from the refractive-index structure parameter C n 2 ${C}_{n}^{2}$ to the gamma–gamma channel parameters α and β , bypassing the limitations of standard Rytov approximations. Turbulence forecasting, a systematic ablation study demonstrates that a hybrid convolutional-recurrent architecture with temporal attention achieves a near-perfect coefficient of determination ( R 2 = 0.9956) for log 10 C n 2 ${\mathrm{log}}_{10} {C}_{n}^{2}$ , dramatically outperforming standalone CNN ( R 2 = −104.42) and LSTM ( R 2 = −0.06) models. The predicted channel state drives a threshold-based MCS selector evaluated through exact analytic integrals for outage probability, average bit error rate (BER), and spectral efficiency. When benchmarked against adaptive optics, the predictive scheme achieves a 9.3 % lower average BER and a 10.5 % lower outage probability, corresponding to absolute values of 2.04 × 10 −4 and 0.0068, respectively. In fact, a reduction in inference time by 15.7 % in comparison with the CNN-only model, amounting to an inference time of 86.45 ms. The proposed approach, therefore, provides a feasible, low-complexity solution for mitigating atmospheric turbulence in FSO systems.

Authors

Institutions

Publication Details

Journal
Journal of Optical Communications
Published
2026-11-18
DOI
https://doi.org/10.1515/joc-2026-0251
Primary Topic
Optical Wireless Communication Technologies
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Turbulence-aware free-space optical communication system hybrid CNN–LSTM channel modeling, adaptive prediction, and real-time mitigation

Selewondim Eshetu Ayana, Demissie Jobir Gelmecha, Ashenafi Paulos Forsido, Lami Nagara Mokonnen
4 links
Journal of Optical Communications
Optical Wireless Communication Technologies
article

Turbulence-aware free-space optical communication system hybrid CNN–LSTM channel modeling, adaptive prediction, and real-time mitigation

Selewondim Eshetu Ayana, Demissie Jobir Gelmecha, Ashenafi Paulos Forsido, Lami Nagara Mokonnen
article en

Abstract

Atmospheric optical turbulence severely degrades the performance of free-space optical communication. Mitigating these effects without costly additional hardware remains an open challenge. This paper presents a fully machine learning–driven adaptive modulation and coding architecture that combines a first-principles neural surrogate model with a hybrid deep learning predictor to optimize the modulation and coding scheme. A feed-forward neural network, trained on exact Kolmogorov phase-screen statistics, learns the differentiable mapping from the refractive-index structure parameter C n 2 ${C}_{n}^{2}$ to the gamma–gamma channel parameters α and β , bypassing the limitations of standard Rytov approximations. Turbulence forecasting, a systematic ablation study demonstrates that a hybrid convolutional-recurrent architecture with temporal attention achieves a near-perfect coefficient of determination ( R 2 = 0.9956) for log 10 C n 2 ${\mathrm{log}}_{10} {C}_{n}^{2}$ , dramatically outperforming standalone CNN ( R 2 = −104.42) and LSTM ( R 2 = −0.06) models. The predicted channel state drives a threshold-based MCS selector evaluated through exact analytic integrals for outage probability, average bit error rate (BER), and spectral efficiency. When benchmarked against adaptive optics, the predictive scheme achieves a 9.3 % lower average BER and a 10.5 % lower outage probability, corresponding to absolute values of 2.04 × 10 −4 and 0.0068, respectively. In fact, a reduction in inference time by 15.7 % in comparison with the CNN-only model, amounting to an inference time of 86.45 ms. The proposed approach, therefore, provides a feasible, low-complexity solution for mitigating atmospheric turbulence in FSO systems.

Journal of Optical Communications
Adama Science and Technology University (ET)
Optical Wireless Communication Technologies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon