Doubly hermitian precoding for parallel MIMO relay networks

We consider a parallel multiple-input multiple-output (MIMO) relay network, in which a source node communicates with a destination node assisted by multiple parallel relays. In such a network, it is costly to acquire global channel state information (CSI) at every relay node. In this regard, we assume local CSI, i.e., each node in the network knows perfect CSI of the links from and to this node, but only has some statistical information of the other links. We propose an amplify-and-forward relaying strategy, termed doubly Hermitian precoding to efficiently exploit the potential benefit of local CSI. We show that the proposed relaying strategy is asymptotically capacity-approaching as the number of relays tends to infinity. Numerical results demonstrate that the proposed scheme performs close to the performance upper bound obtained by assuming global CSI.