Global Navigation Satellite Systems (GNSS) are undoubtedly the most preferred navigation method in these days. However, GNSS signals can frequently become a target for undesired jammer signals. The most known solution for this problem is beamforming methods. In this study, recurrent neural network (RNN) structures are used to determine the beamforming coefficients of signals with known angle of arrivals. A new encoder-decoder (ED) beamforming approach is presented, which is built with long short-term memory (LSTM) cells. ED beamforming approach is compared with a previously proposed multilayer LSTM network and known null steering beamforming (NSB), which ED beamforming is resulted more successfully in terms of beam and null divergences in directions of signal arrivals and with a shorter response time.
Paper
Full text
Adaptive Beamforming Based on Recurrent Deep Learning for GNSS Bands
Semantic Scholar · Engineering · 2024
Abstract
Global Navigation Satellite Systems (GNSS) are undoubtedly the most preferred navigation method in these days. However, GNSS signals can frequently become a target for undesired jammer signals. The most known solution for this problem is beamforming methods. In this study, recurrent neural network (RNN) structures are used to determine the beamforming coefficients of signals with known angle of arrivals. A new encoder-decoder (ED) beamforming approach is presented, which is built with long short-term memory (LSTM) cells. ED beamforming approach is compared with a previously proposed multilayer LSTM network and known null steering beamforming (NSB), which ED beamforming is resulted more successfully in terms of beam and null divergences in directions of signal arrivals and with a shorter response time.