MIMO antenna optimization for millimeter-wave 5G applications using machine learning

This paper presents the design and machine learning-based optimization of a compact single-port MIMO antenna operating at 37 GHz for millimeter-wave 5G applications. The antenna is fabricated on a Rogers RT5880 substrate and optimized using a neural network trained on 670 CST simulation samples to enhance gain, bandwidth, and impedance matching. The data-driven model efficiently explores the design space, reducing the need for time-consuming full-wave simulations. The optimized single antenna achieves a compact size of 4.21 × 7 mm², bandwidth of 0.9 GHz (36.9–37.8 GHz), return loss better than − 30 dB, and radiation efficiency exceeding 83%. A 1 × 4 antenna array configuration based on the optimized element further improves the peak gain to 13.2 dB. Experimental validation confirms close agreement with simulated results, highlighting the reliability of the ML-driven design approach. The proposed methodology demonstrates a scalable and efficient path to antenna miniaturization and performance enhancement for 5G hardware. These results position the design as a strong candidate for integration into compact mobile and IoT devices operating in the mmWave band.