Dynamic Sectorization in Multi-Panel Massive MIMO Systems

In this paper, a multi-panel multiuser massive multiple-input multiple-output (MIMO) system is considered. Recently, it has been shown that the spectral efficiency in the massive MIMO downlink can be significantly improved by distributing the base station antennas over several panels located at some distance from each other. In order to ensure a close-to-optimum spectral efficiency, a dynamic and fast procedure is required for determining the best possible grouping of panels and assignment of users to panel groups, respectively. We address such a dynamic sectorization problem assuming that in all panel groups low-complexity maximum ratio transmission (MRT) precoding is employed and propose an efficient algorithm based on a sequential method. Here, first a simplified version of the original mixed-integer nonlinear optimization problem is considered. The corresponding linearized integer quadratic problem (IQP) is solved globally with respect to the energy usable in precoding using a state-of-the-art mixed-integer programming (MIP) solver from mathematical optimization. The obtained assignment is further improved with respect to sum rate by a greedy algorithm. It turns out that this approach is fast and leads to high-quality groupings of panels and assignments of users to panel groups for realistic channel data.