Generative Adversarial Networks for Total Electron Content Prediction

The correction of ionospheric delay is an important part of insuring the precision of Global Navigation Satellite Systems. Real time measurements are taken by the various services at ground stations and broadcast to terminals for corrections. The correction works well for terrestrial areas near the stations but errors increase for marine measurements and remote terrestrial locations especially near the Equator. We attempt to create a neural network that can accurately infer the local Vertical Total Electron Content (VTEC) of a receiver from as few parameters as possible. To this end we use a neural network called IONONET that we attempt to improve by augmenting the training dataset with synthetic data from a Generative Adversarial Network (GAN).