Deep Learning Accelerated Antenna Radiation Pattern Prediction for Undersampled Near-Field to Far-Field Transformation

We present a deep learning model based on U-Net architecture to estimate antenna radiation pattern and accelerate antenna characterization. This process is ordinarily time-consuming and expensive due to the need for an anechoic chamber and precision motion stages scanning through the antenna's 2D/3D near-field space. Significant developments in using deep neural network learning exhibit a non-linear relationship between high- and low-resolution datasets (e.g., images). In this work, we propose to build such a neural network to reconstruct the antenna radiation pattern from undersampled near-field data to accelerate pattern acquisition.