Machine Learning-Based Design and Optimization of a Circularly Polarized Dielectric Resonator-Based MIMO Antenna for 5G Sub-6 Ghz

Abstract

In this work, we develop and analyse a 5G sub-6 GHz band multiple-input multiple-output (MIMO) antenna based on a two-port dielectric resonator (DR). A ceramic cylinder is excited by an L-shaped opening that has been slanted at an angle. The circularly polarised (CP) waves and HEM11 mode are produced inside the ceramic by this feeding system. The suggested antenna's left-handed circular polarisation features allow it to cover the possible 5G range between 2.8 and 3.05 GHz. Datasets are generated using the suggested antenna using parametric analysis in HFSS. Several machine learning methods are used to the datasets in order to optimise the suggested antenna design's reflection coefficient (S11), isolation (S21), and axial ratio (AR). The approach suggested here employs a number of different machine learning techniques, including Decision Tree (DT), Deep Neural Network (DNN), k-Nearest Neighbours (KNN), Random Forest (RF), and EXtreme Gradient Boosting (XGBoost). The algorithms produce accurate findings that agree with HFSS simulations. DNN, KNN, and RF, however, produce results that vastly outperform those of DT and XGBoost.