Distributed Two-Stage Beamforming with Power Allocation in Multicell Massive MIMO

We consider a multicellular system in the downlink where at each base station (BS), massive multiple-input multiple-output (MIMO) is utilized via large antenna array. However, big numbers of antennas and users result in a huge channel state information (CSI) overhead and big computational complexity. To cope with these challenges, we propose a two-stage beamforming with power allocation using the signal to leakage and noise ratio (SLNR) as a performance metric since it is known to decouple the optimization problems compared to conventional design methods, allowing independent, distributed processing at each BS. As a first contribution, we derive a deterministic equivalent of the SLNR in the distributed two-stage beamforming context using random matrix theory which provides very accurate approximations in closed-form. The second contribution consists of beamforming design and power allocation. More precisely, in the first stage of the proposed beamforming, an outer beamformer is designed using the deterministic equivalent of the SLNR which requires only statistical CSI and produces an effective system of lower dimension. In the second stage, an inner beamformer applies regularized zero forcing on the low-dimensional effective channel to combat the interference. Additionally, to improve the user fairness, power allocation which maximizes the minimum effective SLNR is proposed and shown to be a convex problem. Simulation results confirm that both the proposed beamforming as well as the combination of the proposed beamforming with power allocation reduce the system dimensionality without degradation of system performance in terms of throughput and user fairness.