Remote Radio Head Multiclustering based Cell-Free Massive MIMO Systems

This paper considers a disaggregated Radio Access Network (RAN) architectural framework with the aim of eliminating cell boundaries and improving the bad performance of users near the cell-edge, which is an inherent limitation of cellular networks. The network consists of several Remote Radio Heads (RRHs), connected via a front-haul network to a number of Distributed Units (DUs), in turn connected to one Centralized Unit (CU). The achieved performance gains come from making each DU manage multiple overlapping clusters of RRHs, each tied to an orthogonal partition of the system bandwidth. This is done by assigning users to their preferred clusters, in such a way that each user has at least one cluster in which it does not suffer from low performance. For this, RRHs may be connected to multiple DUs, effectively extending the coverage area of each DU by adding new front-haul connections. After approximating the RRH clusters as having a circular shape, we formulate a geometric problem to guarantee that all users observe a predefined minimum ergodic rate. We minimize the number of partitions needed, so as to minimize the coordination traffic between the DUs and the CU. Additionally, we find the solution with the minimum number of extra front-haul connections needed. Simulation results show that the 5th percentile ergodic rate of users can be improved by 78%, compared to using a single partition.