Joint User and Antenna Selection in Massive-MIMO Systems With QoS-Constraints

This article considers a downlink transmission in a multiuser massive multi-input-multi-output (massive-MIMO) system in both the large-scale and small-scale fading environments. While the base station has many antennas, the associated circuit costs require some antennas to be turned off, thus necessitating antenna/RF-chain selection. Furthermore, the users have quality-of-service (QoS) constraints and cannot be served within the specified power budget and also with arbitrarily low signal-to-interference-plus-noise ratios. Instead, the base stations seek to schedule only a subset of users for which the QoS-constraints can be met, while the remaining users must be scheduled in other time-frequency slots. The problem is considered under both long and short time-scales. The longer time-scale problem entails solving for the power, user allocation, and number of active antennas to be used over multiple coherence intervals. While the need to select the optimum set of users makes the problem combinatorial, a low-complexity algorithm is proposed that allows us to solve it in polynomial time. For the more computationally intensive per-coherence interval problem, we employ a convex relaxation-based approach in order to obtain an approximate solution. Detailed simulations are carried out to establish that the proposed algorithms outperform standard greedy approaches and are close to optimal for several settings.