Neural Network Based LEO Phased Array Antenna Gain Loss Prediction

Abstract

This paper proposes a neural network-based time-series predictor for the antenna gain loss of phased array beam-forming antennas in LEO satellite communication systems. The gain loss is caused by the limited set of discrete antenna beam pointing vectors which leads to a non-optimal antenna steering and thus to a varying receiving power. The main purpose of the predictor is to reduce the fixed margin in adaptive coding and modulation systems by eliminating the time delay for signaling and MODCOD switching. This paper explains the antenna gain loss effect and how to use Monte-Carlo simulations to train the neural network. The benefit of the predictor is demonstrated by a LEO satellite example system.