Long-Range Fading Channel Prediction Using Recurrent Neural NetworkWei Jiang; Mathias Strufe; Hans Dieter Schotten
In: IEEE Consumer Communications and Networking Conference. IEEE Consumer Communications and Networking Conference (CCNC-2020), January 10-13, Las Vegas, Nevada, USA, IEEE, 2020.
Outdated channel state information (CSI) has a severely negative impact on the performance of a wide variety of adaptive transmission systems. Channel prediction is an effective method that can directly improve the quality of CSI. To realize the full potential of adaptive systems, the prediction horizon should be long enough to at least compensate for the time delay. In this paper, therefore, we focus on the problem of long-range prediction (LRP), i.e., how to forecast fading channels as far ahead as possible. Two different LRP approaches - Multi-Step Prediction and Fading Signal Processing - are proposed for the predictors based on classical Kalman filter and recently proposed recurrent neural networks. As an application example, we present an LRP-aided transmit antenna selection system, whose performance in noisy and correlated channels is evaluated. Numerical results reveal that the RNN predictor can achieve comparable performance with respect to the classical predictor while avoiding its drawbacks in parameter estimation and multi-step processing.
TACNET 4.0 - Hochzuverlässige und echtzeitfähige 5G Vernetzung für Industrie 4.0 - Das taktile Internet für Produktion, Robotik und Digitalisierung der Industrie