Outage Efficient Strategies for Network MIMO with Partial CSIT

We consider a multi-cell MIMO downlink (network MIMO) where $B$ base-stations (BS) with $M$ antennas connected to a central station (CS) serve $K$ single-antenna user terminals (UT). Although many works have shown the potential benefits of network MIMO, the conclusion critically depends on the underlying assumptions such as channel state information at transmitters (CSIT) and backhaul links. In this paper, by focusing on the impact of partial CSIT, we propose an outage-efficient strategy. Namely, with side information of all UT's messages and local CSIT, each BS applies zero-forcing (ZF) beamforming in a distributed manner. For a small number of UTs ($K\leq M$), the ZF beamforming creates $K$ parallel MISO channels. Based on the statistical knowledge of these parallel channels, the CS performs a robust power allocation that simultaneously minimizes the outage probability of all UTs and achieves a diversity gain of $B(M-K+1)$ per UT. With a large number of UTs ($K \geq M$), we propose a so-called distributed diversity scheduling (DDS) scheme to select a subset of $\Ks$ UTs with limited backhaul communication. It is proved that DDS achieves a diversity gain of $B\frac{K}{\Ks}(M-\Ks+1)$, which scales optimally with the number of cooperative BSs $B$ as well as UTs. Numerical results confirm that even under realistic assumptions such as partial CSIT and limited backhaul communications, network MIMO can offer high data rates with a sufficient reliability to individual UTs.