Multi-layer feed forward neural network application in adaptive beamforming of smart antenna system

In this paper an artificial Feed Forward Neural Network (FFNN) is applied for smart antenna adaptive beamforming. The neural network is used to calculate the optimum weights of the uniform linear antenna array to steer the radiation pattern of the toward the desired users and make nulling in the direction of interference sources. Levenberg Marquardt (LM) algorithm and Resilient Backpropagation (Rprop) algorithm are used to train the FFNN. Five element uniform linear array is used with spacing between element equal to the half wavelength. The simulation results of FFNN training using LM and Rprop algorithms showed that the Neural Network (NN) trained by LM training algorithm gives better performance than Rprop training algorithm, since it considers the fastest backpropagation training algorithm but it takes more memory than other algorithms.