Uplink training with pilot optimisation for multicell massive multiple‐input multiple‐output systems

Abstract

The authors investigate pilot contamination in a multicell massive multiple-input multiple-output system, in which the estimated channel state information (CSI) is significantly different from the real CSI. Pilot contamination is an obstacle to achievable rates for users. The authors first propose an uplink training strategy to limit the effect of intercell pilot contamination on channel estimation with low complexity. As a result, the downlink data rate through the reciprocal channels is significantly improved. However, as the conventional uplink training, this strategy suffers from intra-cell pilot contamination when the length of pilots is smaller than the number of users in a cell, while increasing the length of pilots degrades the spectral efficiency. The authors, therefore, propose a pilot optimisation that can be incorporated into the proposed training strategy. Based on the minimum mean square error criterion, they derive a closed-form equation to determine the optimal pilots for all users in a multicell system. The joint pilot optimisation and proposed uplink training strategy is found to outperform other schemes, especially when the number of users exceeds the length of pilots in a cell. Numerical results verify that the downlink achievable rate with the proposed training strategy and pilot optimisation outperforms that with conventional approaches.