Multi-Layer Downlink Precoding for Cloud-RAN Systems Using Full-Dimensional Massive MIMO

Abstract

Full-dimensional (FD) massive multiple-input multiple-output (MIMO) has recently emerged as one of the physical layer enablers of 5G systems. Grounded on prior work on layered precoding, a novel framework for low-complexity multi-layer downlink precoding in multi-cluster cloud radio access network (C-RAN) systems using FD massive MIMO is introduced in this paper. The precoding matrix used at the cluster of remote radio heads (RRHs) associated to a given central control unit (CCU) of the C-RAN is decoupled as a multiplication of three precoding sub-matrices (or layers), a precoding architecture that leverages the special characteristics of the elevation component of the channel correlation matrix to manage both the C-RAN inter-cluster interference and the massive MIMO pilot contamination based on the availability of statistical channel state information. The proposed multi-layer approach is then adapted to the compress-after-precoding-based CCU-RRH functional split where the first and second precoding layers are locally applied at each of the RRHs in a cluster, whereas only the third precoding layer is implemented at the CCU. Optimal and suboptimal solutions are provided, which take into account both the per-RRH transmit power constraints and the capacity constraints of the fronthaul links between the CCU and the associated RRHs. The suboptimal approach, which is rooted on the vector normalization techniques used in massive MIMO precoding schemes with uniform power allocation, is shown to provide minor spectral efficiency losses when compared with the optimal solution. Numerical simulations are used to illustrate the potential of the proposed C-RAN-based framework when benchmarked against the classical FD massive MIMO scheme.