Abstract
In the multi-cell large-scale antenna TDD networks (Massive MIMO TDD networks), the channel state information (CSI) is explored by using uplink training sequences (pilots) which are transmitted by users during the training time. Due to the frequency reuse in the cellular networks and the finite length limit of pilot sequences (depending on the length of channel coherent time), the pilot sequences used in co-channel cells are unavoidable to be non-orthogonal. That is to say, the same set of pilot sequences is reused among the cells, which causes the pilot contamination problem. In this thesis, we propose a new pilot transmission scheduling scheme to completely eliminate the pilot contamination. Further, we apply least-square (LS) channel estimator to derive the estimates of both desired channels and inter-cell interference (ICI) channels, so that we do not need to know the second order statistics of channels. Besides the desired channel estimate, the ICI channel estimate also can be obtained in the this scheme. The achievable rate of downlink data transmission by Zero-Forcing (ZF) precoding is also analysed. Besides downlink data transmission, we use the estimates of interfering channels to help the base station when decoding the symbols users transmit during uplink data transmission.