Distributed User-Centric Cell-Free Massive MIMO with Architectural Constraints

We consider a network-level coordination architecture operating on a disaggregated Radio Access Network (RAN). We introduce the concept of Remote Radio Head (RRH) multi-partitioning, which allows users to be served by clusters of RRHs in a user-centric manner. The network consists of a number of Centralized Units (CUs) connected to several Distributed Units (DUs), each controlling a priori a cluster of multiple-input multiple-output (MIMO) RRHs. Aiming at providing uniform coverage and user performance over the network area, RRHs at cluster boundaries are connected to two or more DUs, leading to a multi-partitioning architecture in which RRHs may belong to more than one cluster. Each DU, in turn, manages multiple overlapping clusters of RRHs. The total system bandwidth is divided into orthogonal resource parts, each of which is assigned to a different partition. We consider a coordination framework to allocate resources, where each DU, in a distributed way, assigns radio resources to the different RRH clusters that it manages. Each DU manages the scheduling of its users to their corresponding serving clusters. Simulation results show that the 5th percentile downlink user spectral efficiency improves by 45% when five RRH partitionings are used, compared to only one.