Neural network-based pattern synthesis of array antennas with element specification

In this paper, a neural network (NN) based solution for obtaining the electric voltages applied at every port of the array feeding network, from radiation pattern information, is proposed. This is to avoid complexity from the designer´s point of view, and establish a relation between the desired radiation pattern and the voltages in the real antenna. The NN training phase must be performed through antenna modeling (including the feeding network) and full-waveform analysis methods, so that the data used for training implicitly consider a realistic model with interaction between elements.