Parallel Computations in Two-Dimensional Adaptive Antenna Arrays

Abstract

This paper considers a parallel version of the real-valued blind Linearly-Constrained (LC) Recursive Least Squares (RLS) adaptive filtering algorithm for weights computation in symmetric adaptive antenna arrays. The algorithm can be implemented using two processors for any number of weights in an array. Parallel computations make it possible to decrease the computation load per processor. This is an important property for use in arrays with a large number of antennas/weights. A particular case of a large array is a two-dimensional configuration, where a relatively small number of antennas along of each dimension results in a large total number of the antennas. As the adaptive algorithm under consideration is rea 1- valued, the paper describes the condition of using real-valued arithmetic in two-dimensional arrays and a computational procedure of the algorithm. Simulation results of both real-valued and complex-valued arithmetic blind LC RLS adaptive filtering algorithms in the two-dimensional array with 8×4=32 antennas/weights are presented. In steady-state the real-valued algorithm provides on average 3 dB greater notch depth in the radiation pattern of the array towards the interference source, compared to the complex-valued arithmetic algorithm.