| Summary: | A novel downlink scheduling strategy for time-division duplexing massive multiple-input multiple-output networks is proposed to relieve the pilot contamination. Specifically, users of slow-channel variation and simultaneous uplink pilot transmission are scheduled in different downlink subframes for adjacent cells, and hence the pilot contamination is significantly suppressed at the price of outdated channel state information. We investigate this trade-off by deriving the distribution of downlink signal-to-interference ratio in terms of Doppler effect, which can be approximated tightly and simply by Gaussian distribution. Based on it, a learning-based algorithm is developed for robust downlink rate allocation without priori knowledge on the users' distribution. It is shown that the proposed cross-frame strategy can effectively improve the overall downlink performance of cell-edge users.
|
|---|