Skip to main content Skip to main navigation

Publication

Neural Network Based Wireless Channel Prediction

Wei Jiang; Hans Dieter Schotten; Ji-Ying Xiang
In: Fa-Long Luo (Hrsg.). Machine Learning for Future Wireless Communications. Chapter 16, Pages 1-30, John Wiley and IEEE Press, 2019.

Abstract

The advantages of wireless communication over wired are its flexibility, scalability, mobility, convenience, and economic efficiency, thanks to the free propagation of electromagnetic waves through a wireless channel from the transmitter to the receiver. Due to the reflection, diffraction, and scattering of electromagnetic waves traveling along different paths, as well as the mobility of surrounding objects or mobile stations, a wireless channel exhibits an extremely challenging condition for the design and implementation of wireless systems. By adapting transmission parameters such as the constellation size, coding rate, transmit power, time or frequency resources, transmit or receive antennas, and relaying nodes to instantaneous channel conditions, adaptive transmission systems can potentially aid the achievement of great performance. To fully realize this potential, the transmitter needs to know accurate channel state information (CSI). In a frequency-division duplex (FDD) system, the CSI is estimated at the receiver and fed back to the transmitter through a limited feedback channel. Owing to time delays in the process of channel estimation, signal processing, and feedback, the available CSI at the transmitter may become outdated before its actual usage. In a time-division duplex (TDD) system, although the feedback delay can be avoided by taking advantage of channel reciprocity, it is still possible that the CSI is outdated, especially in high-mobility environments.