Energy-Efficient User Clustering and Downlink Beamforming for MIMO-SCMA in C-RAN

Abstract

Multiple-inputmultiple-output (MIMO) sparse code multiple access (SCMA) is of great interest for future wireless networks to achieve higher spectral efficiency and support massive connectivity. In this paper, we investigate the key problems of user clustering and downlink beamforming for MIMO-SCMA in a cloud radio access network (C-RAN). Using channel state information available at the central processor, an efficient user clustering algorithm based on the constrained $K$ -means method is proposed. Subsequently, two iterative algorithms for beamforming design are developed by minimizing the total transmission power under quality-of-service (QoS) and fronthaul capacity constraints. In the first approach, we approximate the continuous non-convex constraints by convex conic ones using first-order Taylor expansion and iteratively solve a sequence of mixed-integer second order cone programs (MI-SOCPs) to achieve high quality solution, but with higher complexity. In the second approach, a two-stage low-complexity solution is developed in which beamforming matrices obtained from each stage are combined to form a single beamformer for each user. In the first stage, cluster beamformers are designed by taking advantage of block diagonalization, while in the second stage, user-specific beamformers are determined by minimizing transmission power. The performance of the proposed user clustering and downlink beamforming approaches for MIMO-SCMA in C-RAN is validated through simulations over mmWave channels. Compared to benchmark approaches, the results show significant improvements in terms of transmit power and spectral efficiency.