Robust Group Sparse Beamforming for Multicast Green Cloud-RAN with Imperfect CSI

TL;DR

This paper proposes a robust three-stage group sparse beamforming algorithm for multicast Cloud-RAN with imperfect CSI, aiming to minimize network power by adaptively selecting active RRHs and designing beamformers.

Contribution

It introduces a novel three-stage algorithm combining group sparsity induction, PhaseLift, and SDR techniques for power-efficient multicast beamforming under imperfect CSI.

Findings

Algorithm converges effectively in simulations.

Significant power savings demonstrated.

Robust beamforming achieved despite CSI imperfections.

Abstract

In this paper, we investigate the network power minimization problem for the multicast cloud radio access network (Cloud-RAN) with imperfect channel state information (CSI). The key observation is that network power minimization can be achieved by adaptively selecting active remote radio heads (RRHs) via controlling the group-sparsity structure of the beamforming vector. However, this yields a non-convex combinatorial optimization problem, for which we propose a three-stage robust group sparse beamforming algorithm. In the first stage, a quadratic variational formulation of the weighted mixed l1/l2-norm is proposed to induce the group-sparsity structure in the aggregated beamforming vector, which indicates those RRHs that can be switched off. A perturbed alternating optimization algorithm is then proposed to solve the resultant non-convex group-sparsity inducing optimization problem by exploiting its convex substructures. In the second stage, we propose a PhaseLift technique based algorithm to solve the feasibility problem with a given active RRH set, which helps determine the active RRHs. Finally, the semidefinite relaxation (SDR) technique is adopted to determine the robust multicast beamformers. Simulation results will demonstrate the convergence of the perturbed alternating optimization algorithm, as well as, the effectiveness of the proposed algorithm to minimize the network power consumption for multicast Cloud-RAN.