Statistical massive MIMO design based on uplink eigenspace tracking

For multiuser massive MIMO systems, the acquisition and utilization of statistical channel information is very important. In this paper, we first adopt PASTd algorithm to track the uplink dominant eigenvectors (sub-eigenspace) of channel covariance matrix and then present a low-complexity algorithm to transform the uplink eigenvectors to the downlink eigenvectors. Thirdly, two scheduling algorithms based on downlink eigenvectors are proposed. Particularly, Iterative Comparison of Eigenvectors´Correlation (ICEC) algorithm selects users with quasi-orthogonal dominant eigenvectors; Grouping-based Same-Order scheduling (GSOS) algorithm clusters users with similar eigenvectors into one group, then orders users according to users´ location information and schedules users with the same order across groups. Numerical results show that ICEC algorithm outperforms GSOS algorithm in throughput while GSOS algorithm provides a tradeoff between fairness and throughput. Based on user scheduling, the most dominant (first) eigenvectors of the scheduled users are used as beamforming vectors to form a complete downlink transmission scheme. Finally, the effectiveness of the proposed design is validated via numerical results.