Massive multi-user (MU) multiple-input multiple-output (MIMO) promisessignificant gains in spectral efficiency compared to traditional, small-scaleMIMO technology. Linear equalization algorithms, such as zero forcing (ZF) orminimum mean-square error (MMSE)-based methods, typically rely on centralizedprocessing at the base station (BS), which results in (i) excessively highinterconnect and chip input/output data rates, and (ii) high computationalcomplexity. In this paper, we investigate the achievable rates of decentralizedequalization that mitigates both of these issues. We consider two distinct BSarchitectures that partition the antenna array into clusters, each associatedwith independent radio-frequency chains and signal processing hardware, and theresults of each cluster are fused in a feedforward network. For botharchitectures, we consider ZF, MMSE, and a novel, non-linear equalizationalgorithm that builds upon approximate message passing (AMP), and wetheoretically analyze the achievable rates of these methods. Our resultsdemonstrate that decentralized equalization with our AMP-based methods incursno or only a negligible loss in terms of achievable rates compared to that ofcentralized solutions.
展开▼
