Efficient Beamforming Training and Limited Feedback Precoding for Massive MIMO Systems

In cellular communications, deploying a larger number of antennas at the base station, also called massive multiple-input multiple-output (MIMO), can offer a significant improvement in system throughput. In this paper, we exploit the spatial fading correlations in massive MIMO to reduce the downlink training and the corresponding feedback overhead in frequency division duplexing systems. We first study the user clustering, where the users with similar spatial channel correlations are clustered together. In the study, we provide the optimal metric and prove the convergence of the user clustering. Then, we propose an efficient eigenspace training and precoding (EETP) framework, where two different prebeamforming matrices are designed to minimize the channel estimation error and to manage the inter-user interference, respectively. In the results, we show that the channel estimation error for EETP decreases monotonically when either the number of prebeamforming vectors or the number of clusters increases. The spectral efficiency of the new algorithms is evaluated extensively with different user distributions, errors in channel correlations, different numbers of clusters, and different coherence block lengths, as well as with dynamic user scheduling for a large number of users. The new EETP not only achieves significant savings in the downlink training and the corresponding feedback, but also offers significantly higher system throughput compared with the existing schemes in the literature.