Combinação afim de filtros adaptativos RLS-LMS para conformação de feixe em antenas inteligentes com sintonia paramétrica baseada em redes neurais artificiais

Abstract

The Affine Combination of RLS-LMS Adaptive Filters for Beam Shaping in Smart Antennas with Parametric Tuning Based on Artificial Neural Networks, shows aspects of convergence of LMS (least mean-square) and RLS (recursive least-square) algorithms using an affine combination of filters for adaptive beamforming in smart antennas. The Feed Forward Artificial Neural Network is used to tune the weight parameters. The performance of the affine combination of two individual adaptive filters of different classes is investigated, considering the mixing parameter of the combination calculated adaptively through the stochastic gradient algorithm, called ρη-LMS. The purpose of the combination is to obtain, for a stationary environment, an adaptive RLS-LMS algorithm that outperforms the classical algorithms in terms of convergence speed and stability. The performance of the proposed RLS-LMS affine algorithm is evaluated through computational experiments. The artificial neural network calculates the ideal or optimal weights used in the signal input of the linear filters that adapt the antenna radiation pattern from the uniform linear array, directing several narrow beams to the desired users and minimizing interference or unwanted users. The application of this neural network provides the ability to increase efficiency and optimize the use of smart antennas. The obtained results are presented to be analyzed.