High Gain and Circularly Polarized Two Port Array Based Dielectric Resonator Antenna using Machine Learning for mm-wave IOT Applications

Abstract

The rapid growth of millimeter-wave (mm-wave) Internet-of-Things (IoT) networks necessitates compact, efficient, and high-performance antennas capable of meeting the challenges of high data rates, limited power budgets, and dense device integration. This paper presents a high-gain, circularly polarized two-port dielectric resonator antenna (DRA) array, optimized using machine learning techniques for mm-wave IoT applications. The proposed array employs orthogonal mode excitation and controlled inter-element coupling to achieve stable circular polarization, with an axial ratio below 3 dB across the target band. The two-port configuration provides spatial diversity, enhanced gain, and simplified feeding structure while maintaining compactness, making it suitable for integration into small-form-factor IoT devices. To address the computational complexity of mm-wave antenna design, a machine learning-based surrogate model is developed to predict key antenna performance metrics, including S-parameters, axial ratio, and realized gain, thereby reducing the reliance on time-consuming full-wave electromagnetic simulations. Furthermore, machine learning-driven optimization identifies optimal design parameters and ensures robustness against fabrication tolerances and material variations. Simulation results demonstrate that the proposed design achieves high realized gain and reliable circular polarization within the desired mm-wave spectrum. This integration of DRA technology with machine learning-assisted optimization provides a promising pathway for developing efficient, low-cost, and scalable antennas tailored to next-generation IoT applications.