Design of Zero Clearance SIW Endfire Antenna Array Using Machine Learning-Assisted Optimization
Jin Zhang, Mobayode O. Akinsolu, Bo Liu, Shuai Zhang
Abstract
In this communication, a substrate integrated waveguide (SIW) end-fire antenna array with zero clearance is proposed for fifth-generation (5G) mobile applications using machine learning-assisted optimization. In particular, a novel impedance matching architecture that involves three arbitrary pad-loading metallic vias is investigated and adopted for the antenna element. Due to the stringent design requirements, the locations and sizes of the vias and pads are obtained via a state-of-the-art machine learning assisted antenna design exploration method, parallel surrogate model-assisted hybrid differential evolution for antenna synthesis (PSADEA). Keeping a very low profile, the array optimized by PSADEA covers an operating frequency bandwidth from 36 to 40 GHz. The in-band total efficiency is generally better than 60% and the peak gain is above 5 dBi. The beam scanning range at 39 GHz covers from −20° to 35°.