Leveraging Conditional GANs for Performance Assessment of Free Space Optical Systems

Free-space optical communication (FSOC) is one of the promising technologies providing high-speed, secure, and interference-free communication. However, environmental impairments like turbulence, pointing errors, and amplified spontaneous emission noise restrict the extensive adoption of FSOC technology. While various machine learning techniques have been proposed to predict impairment and compensation, the use of Conditional Generative Adversarial Networks (CGANs) underexplored. This work addresses the potential usage of CGANs to predict the Probability of Outage (Pout), a measure of FSOC system reliability under varying atmospheric conditions, Unlike various other machine learning models which primarily evaluates turbulence intensity $\left( {{C^2}_n} \right)$, our study directly models Pout, capturing complex nonlinear relationships between meteorological parameters and system performance. The CGAN framework achieves a mean squared error (MSE) of 0.0020 and a root mean squared error (RMSE) of 0.0446 for Pout prediction, demonstrating its effectiveness in modeling turbulence-induced impairments. These results highlight the potential of CGANs in advancing FSOC system designs.