Group-Sparse Beamforming for Sum-Spectral Efficiency Maximization in Cloud-RAN

A cloud radio access network (cloud-RAN) is a promising cellular architecture to increase both network spectral efficiency and energy efficiency. In the downlink transmission of cloud-RAN, a fundamental trade-off exists between the sum-spectral efficiency and the network power consumption induced by the fronthual links. To optimize this trade-off, it is essential to jointly identify a set of active remote radio heads (RRHs) and the beamforming vectors used at the active RRHs. To resolve this problem, this paper presents a novel group-sparse beamforming algorithm inspired by sparse principal component analysis (sparse-PCA). The key idea of the proposed method is to reformulate the sum-spectral efficiency maximization problem under a group-sparsity constraint into a generalized sparse-PCA problem, which is a tractable non-convex optimization problem. Using this reformulated optimization problem, a computationally efficient algorithm is proposed, which finds the solution that guarantees the first-order necessary optimality condition of the non-convex optimization problem. Simulation results demonstrate significant advantage of the proposed group-sparse beamfroming method.