Inverse design of antenna beam pattern for millimeter-wave base station applications

Abstract

Beam pattern design for beamforming is an essential requirement for base station operation in the 5G NR millimeter wave band. A beam pattern design requires tuning the phase shifter value of the antenna, and it takes expensive computational cost to find the desired beam pattern. In this paper, we propose a method to obtain the phase shifter value corresponding to the desired beam pattern through a DNN (Deep neural network) based model to reduce the iterative computational cost. The DNN model is trained based on the extracted data through the simulation tool, and the validation is performed by comparing a simulated beam pattern using estimated phase shifter values through trained DNN model with desired beam pattern.