Relay Systems with Local Channel State Information
Jiening Zhan and Michael Gastpar
National Science Foundation
In some wireless systems, mobiles communicate with nearby base stations, which in turn process the received information before relaying it to the wired backhaul. We abstract this architecture into single source, parallel relays, and a single-destination communication model.
The source-to-relay part is modeled as a fading broadcast channel (BC) while the relay-to-destination part is modeled as a non-fading multiple access channel (MAC). We adopt the block fading model where the fades have a finite coherence time interval. We assume that the system has local channel state information. Hence, each relay knows its respective fading coefficient, but no channel state information is assumed at the destination. Such a situation may occur in scenarios where the source allocates a portion of each coherence time for training via pilot signals.
Our goal is to develop encoding strategies for the relays to send information about the source and channel to the destination. We compare the performance of our strategy to that of classical relaying schemes and also the cut-set upper bound. As a first step, we assume the MAC is the classical symmetric gaussian MAC. Next, we abstract the MAC into a general computation channel and identify it by its computation capacity. We use computation coding tools to develop improved relaying strategies.