Two-layer Precoding for Dimensionality Reduction in Massive MIMO

Massive MIMO (multiple-input multiple-output) is a promising technology for the upcoming 5G as it provides significant beamforming gains and interference reduction capabilities due to the large number of antennas. However, massive MIMO is computationally demanding, as the high antenna count results in high-dimensional matrix operations when conventional MIMO processing is applied. In this paper, we focus on two-stage digital beamforming, where the beamformer is split into a slow-varying statistics-based outer beamformer and an inner beamformer accounting for fast channel variations. We formulate two two-stage precoding optimization problems: weighted sumrate maximization and minimum user rate maximization for a single-cell downlink system. We also provide different heuristic methods of forming the outer precoder matrix via user channel covariance, while the inner precoder is obtained as a result for the optimization problem. Unlike most previous work, which consider the outer precoder design based on energy maximization and user group location, our aim is to design it to offer a tradeoff between energy maximization and interference reduction, and also take into account the fairness between users. We evaluate the performance of the different heuristic methods as a function of the number of statistical pre-beams and a fixed user angular spread to see the overall effect of complexity reduction on the system sum-rate and minimum user rate. We also evaluate the advantages of different methods in terms of user fairness.