A Novel Antenna Peaking Technique Using Machine Learning in Noisy Satellite Communication Channels

This paper introduces a novel approach to antenna tracking and pattern reconstruction using machine learning as demonstrated here using a single class support vector machine. It assumes a global estimation technique peaks a noisy main beam with finite pointing error but then the proposed method finetunes the peaking to produce the true beam maximum. The finetuning processes uses a u-v coordinate raster input to a single class support vector machine to classify an n-dB main beam contour and provide a true peak likelihood score therein (essentially filtering out the outlier noise). The results show a 97% improvement over picking the global maximum. The intention of the work is to provide a theoretical foundation from which an actual tracking algorithm could be based.