Data-Driven Optimization of an Array of Steered Sub-Arrays for Enhanced Fairness in MU-MIMO

This paper investigates arrays of steered sub-arrays (SAs) for multi-user MIMO applications. We optimize SA steering angles based on rate-based and gain-based objective functions which naturally balance aggregate system performance and user fairness. Numerical simulations show that, compared to a full-digital array with the same number of RF-chains, rate-optimized arrays increase the array gain in small sector scenarios and compensate for low element gains in large sector scenarios. An array with 4-element SAs achieves the highest rate at all sectors at both low and high signal-to-noise ratios (SNRs). Improvements vary from 2% (120-degree sector, high SNR) to 113% (30-degree sector, low SNR). Hence, adjusting the steering angles to the long-term channel statistics yields significant rate improvements if users are confined to a small sector, without sacrificing performance if this is not the case. Similar results are obtained for gain-based optimized arrays, which is important and convenient for array design.