Machine Learning Design Approach of 3D Printed Pixelated Lens MIMO Antenna for CubeSat

A Machine Learning (ML) synthesis approach is proposed and presented to design a high gain 3D printed pixel lens MIMO antenna for CubeSat at V-band. A realized gain of 12 dBi and an fractional bandwidth of 11.59% (59.3-66.6 GHz) with S11<-20 dB are obtained. Simulated and ML performance results indicate that the ML-based synthesis approach required only a few seconds to predict results with a low error rate, whereas the classic optimization approach, such as GA in HFSS, required 48 hours of continuous iteration to solve the optimization problem.