Design and Optimization of Dielectric Resonator-Based Two-Port Filtering Radiator With Machine Learning for Sub-6 GHz 5G Wireless Applications

This paper proposes the structured and investigation of multi-port ceramic-based aerial with filtering, high gain, and circular polarization features. The alumina ceramic is excited using a cross-shaped microstrip line. Defected ground structure (DGS) is utilized to provide good impedance matching within the operating band. Metasurface suspension over the two-port aerial improves the gain level by an amount of 10.2 dBi. The application of polarization diversity idea helps to improve separation between aerial ports by 25 dB. The prediction of |S11| and axial ratio using random forest and XGBoost-built machine learning (ML) technique reduces the computational complexity. Fabricated prototype confirms that the designed antenna operates in between 2.55 and 3.45 GHz, and 3-dB axial ratio is found in between 2.75 and 3.15 GHz. Directive far-field attributes and suitable amount of diversity parameters confirm its applicability for sub-6 GHz-based 5G communication system.