On the Capacity Scaling of Large Multipair Relay Networks With Successive Relaying Protocol

In this paper, we consider large multi-pair relay networks with K fixed source-destination pairs and M relays randomly distributed in a given area, where each node is equipped with a single antenna and works on half duplex. Each source communicates with its corresponding destination under the aid of the relays. With the conventional two-slot relaying protocol, the sum capacity was found to scale as (K/2) log(M) + O(1), where K is fixed and M -> infinity. This paper proves that the capacity scaling law can be further improved to K log(M) + O(1) with successive relaying protocol, as if the relays became "full duplex." To prove the scaling law, a distributed amplify-and-forward scheme is proposed, which only requires local channel state information (CSI) at each relay and statistical global CSI at the sources and destinations. Furthermore, we prove that imperfect CSI at the relays would not affect the scaling law.