Recurrent Neural Network Application Features to Suppress Non-stationary Interference in an Adaptive Antenna Array

Modern radio engineering systems often operate in a complex and dynamically changing signal-interference environment. Adaptive antenna arrays (AAAs) in these systems perform interference suppression, as a rule, using classical quasioptimal algorithms based on linear algebra methods. Such algorithms do not always effectively suppress non-stationary interference, especially in a practical implementation. This research focuses on the development and application features of the adaptation algorithm based on a recurrent neural network for a digital AAA. The results of numerical simulation show that the LSTM neural network provides comparable efficiency of suppression of non-stationary noise and pulse interference with the classical adaptation algorithm.