Compress-and-forward in the multiple-access relay channel: With or without network coding?

We consider compress-and-forward-based transmission strategies for the multiple-access relay channel. In particular, we study the impact of the capacity on the relay-destination link on the usefulness of network coding in this context. To this end, we compare a transmission scheme with network coding to a simple forwarding scheme without network coding. For both schemes we design optimal log-likelihood ratio (LLR) vector quantizers using the information bottleneck method. Moreover, we provide closed-form expressions for the LLR statistics at the relay which are required for the vector quantizer design. Numerical simulation results show that the usefulness of network coding depends strongly on the capacity of the relay-destination channel as well as on the number of sources.