A joint channel estimation and beamforming separation principle for massive MIMO systems

TL;DR

This paper proves that in massive MIMO systems, separating channel estimation from beamforming does not reduce optimality, under broad conditions applicable to various models, simplifying system design.

Contribution

It establishes conditions where optimal joint processing can be separated into MMSE channel estimation and beamforming, applicable to both centralized and distributed architectures.

Findings

Separation incurs no loss of optimality under general conditions.

Conditions are provided for optimal processing in ergodic achievable rates.

Numerical simulations validate the theoretical results.

Abstract

We demonstrate that separating beamforming (i.e., downlink precoding and uplink combining) and channel estimation in multi-user MIMO wireless systems incurs no loss of optimality under general conditions that apply to a wide variety of models in the literature, including canonical reciprocity-based cellular and cell-free massive MIMO system models. Specifically, we provide conditions under which optimal processing in terms of ergodic achievable rates can be decomposed into minimum mean-square error (MMSE) channel estimation followed by MMSE beamforming, for both centralized and distributed architectures. Applications of our results are illustrated in terms of concrete examples and numerical simulations.