Learning Antenna Radiation Patterns Through Gaussian Process Regression

Knowledge of an antenna system’s radiation pattern is important for the optimal deployment of radio systems in the field. However, while an antenna’s radiation pattern can be measured in an anechoic chamber, in real-world applications, the radiated signal can also be attenuated by the environment in a manner that is angle dependent. For directional antenna systems, in order to properly steer radiated energy in desired directions, the behavior of the radio system in the real world must be characterized. In this work, we propose a method for using Gaussian Processes regression to learn the deviations of an antenna’s array pattern due to its signal’s interactions with the environment. We evaluate this approach on publicly available benchmark datasets collected on antenna systems at 2.4GHz.