King's College London

Research portal

Recurrent Neural Network Channel Estimation Using Measured Massive MIMO Data

Research output: Chapter in Book/Report/Conference proceedingConference paper

Original languageEnglish
Title of host publication2020 IEEE 31st Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC)
Publication statusPublished - 2020

King's Authors


In this work, we develop a novel channel estimation
method using recurrent neural networks (RNNs) for massive
multiple-input multiple-output (MIMO) systems. The proposed
framework alleviates the need for channel-state-information
(CSI) feedback and pilot assignment through exploiting the
inherent time and frequency correlations in practical propagation
environments. We carry out the analysis using empirical MIMO
channel measurements between a 64T64R active antenna system
and a state-of-the-art multi-antenna scanner for both mobile
and stationary use-cases. We also capture and analyze similar
MIMO channel data from a legacy 2T2R base station (BS)
for comparison purposes. Our findings confirm the applicability
of utilising the proposed RNN-based massive MIMO channel
acquisition scheme particularly for channels with long time
coherence and hardening effects. In our practical setup, the
proposed method reduced the number of pilots used by 25%.

View graph of relations

© 2018 King's College London | Strand | London WC2R 2LS | England | United Kingdom | Tel +44 (0)20 7836 5454