A Novel Realistic Approach of Adaptive Beamforming Based on Deep Neural Networks

A new deep neural network (NN) approach applied to antenna array adaptive beamforming is presented in this article. recurrent NN (RNN) based on the gated unit (GRU) architecture used as a beamformer order produce proper complex weights for feeding of array. The proposed RNN utilizes four hidden GRU layers and one extra layer linear transformation. produced are subsequently compared with respective derived by null steering (NSB) technique measure accuracy RNN. training performed using large dataset from an NSB realistic microstrip array, consider real-world effects, such nonisotropic radiation pattern element mutual coupling between elements. performance examined root-mean-square error metric, whereas its evaluated estimating mean value standard deviation divergences main lobe nulls directions their desired directions. comparison various structures overall study RNN-based also presented.