Massive MIMO Channel Estimation for Vehicular Communications: A Deep Learning Based Approach

Abstract

The estimation of the massive multiple input multiple output (MIMO) channel for vehicular communications is very challenging due to the variation of the channel and the requirement of low latency. To improve the accuracy and reduce the delay of the massive MIMO channel estimation, the recently emerging and popular deep neural network is exploited in this paper to learn the sparse structural information of the MIMO channel and estimate the channel more accurately and more rapidly. Firstly, a novel deep learning based massive MIMO channel estimation (DLCE) scheme is proposed, which achieves an efficient trade-off between the accuracy and the delay in channel estimation. Furthermore, exploiting the spatial correlation of the multiple antennas channel, an enhanced scheme called spatial-correlated DLCE (SC-DLCE) is proposed to further improve the channel estimation accuracy, especially in low signal-to-noise ratio environment. Simulation results demonstrate that the two proposed schemes can significantly improve the accuracy of massive MIMO channel estimation with a much shorter processing delay in practical vehicular communications terminals compared with the state-of-the-art benchmark schemes.