On the benefit of delayed CSIT in fading MIMO broadcast channel with CSIR locality

Abstract

The benefit of delayed channel state information (CSI) at the transmitter is investigated in fast-fading MIMO broadcast channels (BC) with local CSI at receivers (CSIR). The seminal work by Maddah-Ali and Tse shows that delayed CSI at the transmitter (DCSIT) can increase the degrees of freedom (DoF) in fast-fading MIMO BC. A caveat, however, is that each receiver was assumed to know the channel coefficients of other receivers, and the cost of distributing such global CSIR was not considered. In this paper, we propose a model to capture the cost of CSIR dissemination. In this model, each receiver only knows its own channel (local CSIR) while rate-limited capacitated links among the receivers are available. These rate-limited links can be used for exchanging CSIR or other purposes such as facilitating cooperative MIMO. Our main contribution is the evaluation of achievable DoF of two classes of schemes in two-user MISO BC. One class leverages delayed CSIT following the Maddah-Ali-and-Tse (MAT) scheme together with the rate-limited links for exchanging CSIR. The other uses cooperative MIMO and does not leverage delayed CSIT at all. We first show that when the transmitter has delayed CSI of both receivers, leveraging delayed CSIT with MAT scheme cannot outperform cooperative MIMO in terms of DoF. Next, we turn our attention to the hybrid CSIT scenario where the CSI of one receiver is instantaneously known at the transmitter and the other receiver. We show that delayed CSIT can strictly improve DoF in this case.