Neural network processing for adaptive array antennas

Abstract

Currently, several algorithms can be used to perform the direction finding or angle of arrival of signals from mobile users. One drawback of these algorithms is the difficulty of their implementation in real-time because of their intensive computational complexity. Neural networks, on the other hand, due to their high-speed computational capability, can yield results in real-time. Moreover, conventional beamformers require highly calibrated antennas with identical element properties. Performance degradation often occurs due to the fact that these algorithms poorly adapt to element failure or other sources of errors. Neural network-based array antennas do not suffer from this shortcoming.