Self-Learning Detector for the Cell-Free Massive MIMO Uplink: The Line-of-Sight Case

TL;DR

This paper introduces a self-learning method for estimating path losses in cell-free massive MIMO uplink systems, especially in line-of-sight environments, improving accuracy despite pilot contamination.

Contribution

It proposes a novel pilot transmission scheme with phase rotation and an estimation algorithm effective for both Rayleigh and line-of-sight channels.

Findings

Significantly improves path loss estimation accuracy.

Effective in environments with rapid channel changes.

Outperforms baseline state-of-the-art methods.

Abstract

The precoding in cell-free massive multiple-input multiple-output (MIMO) technology relies on accurate knowledge of channel responses between users (UEs) and access points (APs). Obtaining high-quality channel estimates in turn requires the path losses between pairs of UEs and APs to be known. These path losses may change rapidly especially in line-of-sight environments with moving blocking objects. A difficulty in the estimation of path losses is pilot contamination, that is, simultaneously transmitted pilots from different UEs that may add up destructively or constructively by chance, seriously affecting the estimation quality (and hence the eventual performance). A method for estimation of path losses, along with an accompanying pilot transmission scheme, is proposed that works for both Rayleigh fading and line-of-sight channels and that significantly improves performance over baseline state-of-the-art. The salient feature of the pilot transmission scheme is that pilots are structurally phase-rotated over different coherence blocks (according to a pre-determined function known to all parties), in order to create an effective statistical distribution of the received pilot signal that can be efficiently exploited by the proposed estimation algorithm.