Machine Learning–Assisted Hard Switching Scheme for Next Generation Optical Wireless Hybrid Communication Systems

ABSTRACT This study presents an end‐to‐end performance evaluation of a hybrid radio frequency/free space optic–visible light communication (RF/FSO‐VLC) communication link designed to transmit 10‐Gbps on–off keying (OOK)–modulated data. The data are transmitted over outdoor RF/FSO channels to an indoor VLC channel, ultimately reaching the end users. To evaluate the end‐to‐end performance of the RF/FSO‐VLC hybrid communication system, the outage probability is first computed using distinct channel models for each segment: the Nakagami fading model for the RF link, a combined gamma–gamma atmospheric turbulence and pointing error model for the FSO link, and the Lambertian propagation model for the VLC link. The probability density function (PDF) and cumulative distribution function (CDF) of the signal‐to‐noise ratio (SNR) are derived and used in a Monte Carlo simulation implemented in MATLAB. Subsequently, an optical system simulation is conducted by varying input parameters such as laser power, link distances, atmospheric attenuation, field‐of‐view (FOV) angle, and RF carrier frequency. System performance is assessed in terms of bit error rate (BER), quality factor ( Q factor), and SNR. Machine learning (ML) techniques—including artificial neural networks (ANNs), support vector machines (SVMs), decision trees (DTs), and random forests (RFs)—are employed for real‐time performance monitoring and dynamic link switching within the hybrid communication system. These models are trained using data generated from optical simulations. The simulation outputs are processed by an artificial intelligence module comprising predictive models based on the aforementioned ML algorithms. This module enables intelligent switching between communication links to maintain the required quality of service (QoS), based on predicted performance metrics. According to the obtained results, the outage probability analysis indicates that the RF‐VLC and FSO‐VLC hybrid links offer improved performance compared to standalone RF, FSO, and VLC links, with the FSO‐VLC configuration demonstrating superior performance over the RF‐VLC link. Additionally, the ML models, which enable hard switching based on optical system performance analysis, achieved prediction accuracies ranging from 95.64% to 98.75%, with an average switching accuracy of 97.5% ( R 2 = 0.975).