Obstacle modeling in array synthesis using neural networks

A new neural network-based method able to perform synthesis over real radiating structures in presence of obstacles is presented. The method is based on the ability of neural networks to establish a relation between inputs and outputs from a previous training procedure with known sets of pairs input/output. If such sets are obtained using a full-waveform analysis tool, a neural network can estimate the voltages that must be applied to the radiating elements of an array in order to get a desired radiated field distribution, taking into account interactions between elements and the presence of obstacles in a near environment