Beamforming design with combined channel estimate and covariance CSIT via random matrix theory

The Interfering Broadcast Channel (IBC) applies to the downlink of (cellular and/or heterogeneous) multi-cell networks, which are limited by multi-user (MU) interference. The interference alignment (IA) concept has shown that interference does not need to be inevitable. In particular spatial IA in the MIMO IBC allows for low latency transmission. However, IA requires perfect and typically global Channel State Information at the Transmitter(s) (CSIT), whose acquisition does not scale well with network size. Also, the design of transmitters (Txs) and receivers (Rxs) is coupled and hence needs to be centralized (cloud) or duplicated (distributed approach). CSIT, which is crucial in MU systems, is always imperfect in practice. We consider the joint optimal exploitation of mean (channel estimates) and covariance Gaussian partial CSIT. Indeed, in a Massive MIMO (MaMIMO) setting (esp. when combined with mmWave) the channel covariances may exhibit low rank and zero-forcing might be possible by just exploiting the covariance subspaces. But the question is the optimization of beamformers for the expected weighted sum rate (EWSR) at finite SNR. We propose explicit beamforming solutions and indicate that existing large system analysis can be extended to handle optimized beamformers with

the more general partial CSIT considered here.