Recurrent Neural Networks with Long Short-Term Memory for Fading Channel Prediction

Wei Jiang, Hans Dieter Schotten

In: 2020 IEEE 91st Vehicular Technology Conference. IEEE Vehicular Technology Conference (VTC-2020) May 25-28 Antwerp Belgium IEEE 2020.


With the aid of accurate channel state information (CSI) at the transmitter, a wireless system can receive great performance by adaptively selecting its transmission parameters. However, the CSI becomes outdated quickly due to the rapid channel variation caused by multi-path fading, leading to severe performance degradation. Such an impact is applied to a wide variety of adaptive transmission systems, the fading channel prediction that can combat the outdated CSI is therefore of great significance. The aim of this paper is to propose two novel recurrent neural networks (RNN)-based predictors, leveraging the strong time-series prediction capability of long short-term memory or gated recurrent unit. Performance evaluation is conducted and the results in terms of prediction accuracy verify that the proposed predictors notably outperform the conventional RNN predictor.


PID1247988.pdf (pdf, 208 KB)

German Research Center for Artificial Intelligence
Deutsches Forschungszentrum für Künstliche Intelligenz