Performance Enhancement Of FSO Communication Systems Using Machine Learning With F-Distribution Modeling For 5G/6G And Iot Applications Under Varying Weather Conditions

Abstract

In this paper, we propose a hybrid performance improvement solution to Free Space Optical (FSO) communication, a solution, namely, facing reliability constraint due to atmospheric turbulence and weather variability in 5G/6G an IoT networks. Due to turbulence, precipitation and pointing errors, traditional FSO channels also have suffered signal degradation, commonly modeled as e.g. log-normal or gamma-gamma distributions. These methods do not however translate well to real time circumstances. To mitigate the latter shortcoming, the suggested model combines the F-distribution-based channel modeling of the signal and a Signal prediction mechanism based on machine learning in the form of Support Vector Machines (SVM). Intensity Modulation with Direct Detection (IM/DD) is replicated by the hybrid model, applying the use of On-Off Keying (OOK) and the channel impairments are modeled using the F-distribution, simulating different levels of turbulence. Real-time attenuation is predicted using SVM regression model trained on synthetic weather data (visibility, humidity, precipitation, wind speed, and cloud cover), allowing dynamic signal thresholding. The simulation findings reveal that the hybrid system may considerably reduce Bit Error Rate (BER) under varied weather situations as compared to traditional Channel State Information (CSI)-based detection. The system is injury resistant to fog, rain, and pointing inaccuracies and has an R² of 0.96 regarding predicting attenuation. This is an easy-to-install and open-source solution, fully operable in Python, and can be customized to be deployed in the next-generation wire systems. It offers a scalable, data-based substitute for hardware-intensive techniques for enhancing FSO connection dependability.