A data-driven machine learning framework for the design of a compact dual-band 2 × 2 MIMO dielectric resonator antenna for 5G and IoT networks

The rapid growth of wireless technologies and smart devices has increased the demand for compact multiband antennas with efficient design methodologies. Conventional antenna design is computationally intensive and time-consuming. This work presents a Machine Learning (ML)-assisted design of a compact quad-port dual-band 2 × 2 MIMO dielectric resonator antenna with mechanical dimensions of 60 × 60 × 10 mm3 and electrical dimensions of 0.70λ0 × 0.70λ0 × 0.12λ0 at 3.52 GHz. Four ML models were evaluated to predict a critical antenna parameter for both operating bands. The optimized antenna covers 3.30–3.73 GHz and 5.10–6.05 GHz, achieving fractional bandwidths of 12.23% and 17.04%, with resonant frequencies at 3.52 GHz and 5.6 GHz. Isolation exceeds 20 dB in the lower band and 25 dB in the upper band. The proposed antenna supports 5G NR, Wi-Fi, IoT, radar, and satellite applications, demonstrating that ML simplifies antenna design while enhancing performance and design efficiency.