Abstract
Cooperative communication suffers from multiplexing loss and low spectral efficiency due to the half duplex constraint of relays. To improve the multiplexing gain, successive relaying, which allows concurrent transmission of the source and relays, has been proposed. However, the severe inter-relay interference becomes a key challenge. In this paper, we propose a channel aware successive relaying protocol, also referred to as CAO-SIR, which is capable of thoroughly mitigating inter-relay interference by carefully adapting relays' transmission order and rate. In particular, a relay having a poorer link to the source is scheduled first to forward a message, the data rate of which is adapted to the link quality of the source-relay and relay-destination channels. By this means, each relay may decode the messages intended for the preceding relays, and then cancel these relays' interference in a low complexity which is equal to that of Decision Feedback Equalizer (DFE). To further optimize and analyze CAO-SIR, we present its equivalent parallel relay channel model, based upon which the adaptive relay selection and power allocation schemes are proposed. By employing M half duplex relays, CAO-SIR is capable of achieving an diversity-multiplexing tradeoff (DMT) given by d(r) = max (M + 1) 1- {(M+2/M+1r)} , (1 - r) , where d(r) and r denote the diversity and multiplexing gains, respectively. Its DMT asymptotically approaches the DMT upper bound achieved by (M + 1) × 1 MISO systems or M full duplex relays, when M is large.
Funding Information
  • National Basic Research Program of China (973 Program) (2013CB336600)
  • NSFC Excellent Young Investigator (61322111)
  • 973 Program (2012CB316000)
  • National innovative talents promotion program (2013RA2151)
  • Beijing Nova Program (Z121101002512051)
  • MoE new century talent program (NCET-12-0302)
  • MoE self-determined research project (20131089396)
  • National science & technology key projects (2013ZX03003006-005, 2013ZX03003004-002)

This publication has 23 references indexed in Scilit: