Leveraging Artificial Intelligence for Enhanced Free Space Optical Communications: A Comprehensive Review

Abstract

The communication of Free Space Optical (FSO) has been approved to be a crucial high-bandwidth, technology possessing free license for 6G networks, UAV-to-ground, as well as satellite-to-ground backhaul. The well-known application related to turbulence of atmosphere, scintillation, as well as errors of reference has restricted its employment because of the signal stochastic fading. Viewed historically, the impairments reduction was mainly dependent upon frameworks that were complex as well as analytical. This required particular, actual-time Channel State Information (CSI). The present study offers a comprehensive literature survey concerning the change related to the transformation towards Artificial Intelligence data-driven systems as well as Machine Learning depending on the systems of FSO during the period of 2022 and 2026. The combination of Deep Learning (DL), Convolutional Neutral Network (CNNs.), as well as Deep Reinforcement Learning (DRL) are crucially evaluated. This evaluation is made within four functional areas: modeling of channel, the mitigation of turbulence, processing of signal, and adaptive resource allocation. In order to locate the need an organizing principle that is structured, this study offers a formal four-domain AI-FSO Taxonomy that divides all methods that are surveyed via their operational function within the FSO link, reinforcing logical comparison across studies and forming a framework that can be reproduced for future researchers. The findings of the present study have shown that neural auto encoders and cascaded CNNs can fulfill up to 12 dB Signal-to-Noise Ratio (SNR) gains at the 7% Hard-Decision Forward Error Correction (HD-FEC) threshold, while DRL-based relay-selection and power-allocation agents show 1-2 orders of magnitude improvements in Bit Error Rate (BER) over heuristic-based policies. Other research directions are also located with a focus on lightweight and hardware-aware The techniques of AI that can be explained for the coming generation resilient optical wireless networks.