Efficient modeling of graphene-dielectric resonator based hybrid MIMO antenna for THz application using machine learning algorithms.

In this communication, a hybrid MIMO antenna (combination of graphene and ceramic) is designed and investigated in THz frequency range. Circular Polarization is achieved by feeding the cylindrical ceramic using moon shaped aperture with L-shaped Microstrip line. Mutual coupling between the antenna ports has been reduced to − 25 dB by using the concept of polarization diversity. Graphene material is used to introduce the re-configurability in the proposed radiator. For avoiding the difficulty of simulation of radiator at THz regime (very high simulation time), machine learning algorithms i.e. Artificial Neural Network (ANN), Random Forest and XGBoost are utilized to predict the reflection coefficient and axial ratio of designed antenna. The designed radiator works from 2.4 to 3.1 THz with 3-dB axial ratio from 2.6 to 3.0 THz. Good value of diversity parameters and stable radiation features make the designed radiator appropriate for various wireless applications in THz frequency band. [ABSTRACT FROM AUTHOR]