Synthetic Generator for Ionospheric Amplitude Scintillation Fading Channels Using Generative Adversarial Networks

The demand for global navigation satellite system positioning processing techniques resilient to low-latitude ionospheric effects, notably signal fading and the related scintillation, has driven the exploration of innovative solutions. Various signal processing and machine learning methods have been employed to enhance receiver performance under scintillation conditions, but some possible machine learning approaches and simulations have still not been tested. Also, to design and test sophisticated receivers capable of functioning amid scintillation effects, a dataset of cases of strong scintillation is essential for comprehensive coverage of severe environments. This study introduces an ionospheric amplitude scintillation simulator based on neural networks, utilizing autoencoders, generative adversarial networks (GANs), and particle swarm optimization. The synthetic scintillation simulator generates time series that closely adhere to the statistical characteristics of the α–μ fading model. The proposed simulator comprises a GAN explicitly trained through supervised learning, preserving the temporal dynamics of the dataset. In addition to the GAN architecture, the simulator includes an autoencoder that learns a low-dimensional latent space, facilitating the reproduction of the temporal relationships observed in historical data by the generator. The implemented simulator was trained and validated using fading and scintillation data collected in São José dos Campos, Brazil, during the period of 16–30 November 2014, when scintillation was severe. Results demonstrate that the proposed simulator accurately generates time series with precise values of amplitude scintillation, quantified by the index <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">S</i><sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">4</sub> and first-order statistics following the α–μ fading model. This represents the first instance of a simulator achieving such a high degree of statistical fidelity and successful validation, indicating the promising potential of this approach in simulating ionospheric fading channels.