Evaluation of an Antenna Selection Strategy for Reduced Massive MIMO Complexity

Massive Multiple-Input Multiple-Output (MIMO) is emerging as one promising technology for the fifth generation (5G), but the hardware and software complexity arising from the sheer number of transmitting elements is a bottleneck. Antenna selection strategies have been reported as an appealing solution for hybrid beamforming architectures to select a number of radio-frequency (RF) chains less than the total number of antennas but are yet to be fully defined and evaluated. In this work, a generic strategy relying on the receiver (Rx) spatial correlation is investigated to select the best antenna subset from a full array. It is evaluated from ray-traced massive MIMO radio channels using propagation metrics, and also sum-rate capacity computation. The results demonstrate that (1) for a fixed number of users, a subset with wisely selected distributed elements outperforms a collocated one for all studied metrics with performance close to the full array and (2) using a proposed optimization algorithm for a dynamic number of users, the number of active Tx antennas can be further optimized for the different studied subsets. Moreover, for a fixed number of users, the best antenna subset presenting the lowest Rx correlation values is found to reach the optimal sum-rate capacity using simple linear precoders compared with dirty paper coding. This is achieved with only a third of the initial RF chain hardware complexity, thus validating the proposed approach. Also, experimental measurements, presented for an industrial scenario, validate the proposed approach.