Capacity of the Gaussian two-hop full-duplex relay channel with residual self-interference

Nikola Zlatanov, Erik Sippel, Vahid Jamali, Robert Schober

Research output: Contribution to journalArticleResearchpeer-review

27 Citations (Scopus)

Abstract

In this paper, we investigate the capacity of the Gaussian two-hop full-duplex (FD) relay channel with residual self-interference. This channel is comprised of a source, an FD relay, and a destination, where a direct source-destination link does not exist and the FD relay is impaired by residual self-interference. We adopt the worst case linear self-interference model with respect to the channel capacity, and model the residual self-interference as a Gaussian random variable whose variance depends on the amplitude of the transmit symbol of the relay. For this channel, we derive the capacity and propose an explicit capacity-achieving coding scheme. Thereby, we show that the optimal input distribution at the source is Gaussian and its variance depends on the amplitude of the transmit symbol of the relay. On the other hand, the optimal input distribution at the relay is discrete or Gaussian, where the latter case occurs only when the relay-destination link is the bottleneck link. The derived capacity converges to the capacity of the two-hop ideal FD relay channel without self-interference and to the capacity of the two-hop half-duplex (HD) relay channel in the limiting cases when the residual self-interference is zero and infinite, respectively. Our numerical results show that significant performance gains are achieved with the proposed capacity-achieving coding scheme compared with the achievable rates of conventional HD relaying and/or conventional FD relaying.

Original languageEnglish
Article number7817779
Pages (from-to)1005-1021
Number of pages17
JournalIEEE Transactions on Communications
Volume65
Issue number3
DOIs
Publication statusPublished - 1 Mar 2017

Keywords

  • Channel capacity
  • full-duplex
  • relay
  • self-interference

Cite this