Deep Learning-Aided Analysis for UAV-Assisted NOMA Networks Using FSO-RF Communications

This paper aims to evaluate the efficacy of a hybrid free space optical-radio frequency (FSO-RF) non-orthogonal multiple access (NOMA) network using an unmanned aerial vehicle (UAV) relay. We examine the spectral efficiency and enhanced reliability provided by FSO under the inherent distortion caused by atmospheric turbulence by analyzing the outage performance of the system. Further, we derive the asymptotic outage behavior for both the near and far NOMA users. Additionally, we build a deep learning model utilizing deep neural networks (DNN) to predict the outage probability of the proposed system with great accuracy and less time. We then compare its performance with pre-existing regression models such as random forest and xgboost. In contrast to the orthogonal multiple access and the benchmark RF-Rayleigh channel, we highlight the performance advantage of the proposed UAV assisted NOMA network. Our study highlights the potential of DNN-based analysis to enhance efficiency and accuracy in evaluating hybrid FSO-RF communication networks, addressing challenges relevant to fifth-generation and beyond.