Performance prediction of dielectric resonator based mm-wave MIMO antenna using machine learning algorithm

Abstract The alumina ceramic-based dual port radiator is structured and analyzed in this communication. Cylindrical-shaped ceramic is fed by using modified circular aperture. The designed aperture creates dual orthogonal modes i.e. H E M 11 δ x and H E M 11 δ y mode inside the ceramic with 90° phase shifts. This property generates circularly polarized waves from 27.6 GHz–28.4 GHz. Anti-parallel placements of antenna ports introduce the concept of space diversity and improve the isolation level by more than 25 dB. To reduce the computational time, two machine learning algorithms i.e. Random Forest and XGBoost are utilized to predict the reflection coefficient variation. Testing of a fabricated prototype of the proposed radiating structure confirms that the designed radiator works well between 26.8 GHz–28.7 GHz. Asymmetrical placement of ports concerning substrate tilts the radiation pattern obtained from different ports in different directions i.e. ±30°. Stable radiation features and good diversity performance make the designed antenna suitable for mm-wave 5G communication system.