Channel Estimation for mmWave Massive MIMO Based Access and Backhaul in Ultra-Dense Network

TL;DR

This paper introduces a structured compressive sensing-based channel estimation method for mmWave massive MIMO in ultra-dense networks, significantly reducing pilot overhead by leveraging angular sparsity, and demonstrating near-optimal performance in simulations.

Contribution

It proposes a novel SCS-based channel estimation scheme that exploits angular sparsity in mmWave massive MIMO to reduce pilot overhead in ultra-dense networks.

Findings

The proposed scheme outperforms traditional methods in simulations.

It approaches the theoretical performance bound.

Pilot overhead depends only on the number of dominant multipaths.

Abstract

Millimeter-wave (mmWave) massive MIMO used for access and backhaul in ultra-dense network (UDN) has been considered as the promising 5G technique. We consider such an heterogeneous network (HetNet) that ultra-dense small base stations (BSs) exploit mmWave massive MIMO for access and backhaul, while macrocell BS provides the control service with low frequency band. However, the channel estimation for mmWave massive MIMO can be challenging, since the pilot overhead to acquire the channels associated with a large number of antennas in mmWave massive MIMO can be prohibitively high. This paper proposes a structured compressive sensing (SCS)-based channel estimation scheme, where the angular sparsity of mmWave channels is exploited to reduce the required pilot overhead. Specifically, since the path loss for non-line-of-sight paths is much larger than that for line-of-sight paths, the mmWave massive channels in the angular domain appear the obvious sparsity. By exploiting such sparsity, the required pilot overhead only depends on the small number of dominated multipath. Moreover, the sparsity within the system bandwidth is almost unchanged, which can be exploited for the further improved performance. Simulation results demonstrate that the proposed scheme outperforms its counterpart, and it can approach the performance bound.