Joint Centralized and Distributed Precoding in Scalable Cell-Free Massive MIMO Systems

Abstract

We consider the precoding design for scalable cell-free massive multiple-input multiple-output (MIMO) systems. In-spired by the adaptable trade-off between the spectral efficiency and centralization overhead, we propose a joint centralized and distributed (JCD) precoding scheme. In the proposed precoding, only parts of the access points (APs) are collaborated to apply centralized precoding through the central processing unit (CPU) while the rest of the APs perform distributed precoding using local channel state information (CSI) to reduce the fronthaul signaling overhead and computational complexity. We adopt the partial minimum mean square error (P-MMSE) and local P-MMSE (LP-MMSE) for centralized and distributed precodings, respectively. In addition, simple AP selection and heuristic power allocation based on the large-scale parameter are also presented. It is demonstrated through the system-level simulation that the proposed J CD precoding significantly outperforms the existing local precodings, such as LP-MMSE and maximum-ratio (MR), even if a small number of APs are centralized.