DMIMO and Machine Learning Techniques for Enhanced BER and Data Rates in 5G Free-Space Optical Communication

Abstract

This paper proposes a dual-MIMO (DMIMO) free-space optical (FSO) communication system integrated with hybrid machine learning models (CNN and LSTM) to enhance bit error rate (BER), Q-factor, and data rates for 5G backhaul networks under atmospheric disturbances. Using OptiSystem software, we simulated the system's performance in fog, rain, haze, and snow environments. Results demonstrate data rates up to 120 Gbps and up to 80% BER improvement compared to conventional MIMO systems, especially under dense fog and turbulence. Although the proposed system shows excellent performance in both simulated conditions and also in experimental validation, real-world challenges such as sudden weather changes, pointing errors, and computational complexity were tested in both simulation and experimental environment. Future work will focus on implementing adaptive beam steering to enhance its validation in real-world situations to optimise the system for dynamic environments.