Pilot Contamination Reduction in Multi-cell TDD Systems with Very Large MIMO Arrays

Abstract

Very large multiple-input multiple-output (MIMO) technology has a potential of significantly improving the system performance of multi-cell time-division duplexing (TDD) networks, but in practice it is limited by pilot contamination. Different from the traditional channel estimation mean square error (MSE) expressions, the expressions that we derived have algebraic form, which no longer need hard matrix inversion as M, the number of the base station (BS) antennas, increasing. From them, we also found that the average transmitted power and length of training sequence almost does not help in enhancing the performance of MSE as $$M\rightarrow \infty $$ . Based on this, two pilot contamination reduction methods for the very large MIMO multi-cell TDD system were proposed. One is realized by grouping all the cells into two categories, using orthogonal pilots between these two types of cells or aligning the uplink pilot time slot of the cells in one category with the downlink data time slot of those cells in the neighboring categories. The other is to find the optimal division of pilot sequence length and the set of all users’s pilot transmission slots allocation through BSs’s coordination. The effectiveness of our proposed methods are verified via both theoretical analysis and numerical results.