L-SLNR-Based Distributed Clustering Towards Cell-Free Massive MIMO Systems With Local ZF

Abstract

User-centric cell-free massive multiple-input-multiple-output (UC CF mMIMO) systems with local zero-forcing (ZF), where users are only served by a portion of access points (APs) and precoding is calculated by APs using local instantaneous channel state information (CSI), ensure the scalability of actual network deployment. However, limited by the spatial degrees of freedom of a single AP, allocating more degrees of freedom to suppress interference will reduce the precoding gain, making the performance of local ZF sensitive to the number of service users of APs. In this letter, in order to fully leverage the performance of local ZF, we firstly quantitatively analyzed the impact of the number of service users of APs on precoding gain and interference cancelation. Then, to provide an adaptive trade-off between precoding gain and interference cancellation, we design a local signal leakage and interference ratio (L-SLNR) based distributed clustering (DC) algorithm, which is executed locally in APs utilizing only local channel statistics. The numerical results indicate that our proposed DC outperforms all existing clustering schemes in terms of average spectral efficiency (SE).