We consider the two-way relay network where two nodes communicate via a relay. We assume that the data at the nodes are correlated (e.g., measurements in a sensor network) and that there is no direct communication between the nodes. The nodes communicate via the relay using a two-phase protocol consisting of an uplink part over an orthogonal multiple access channel and a downlink part over a broadcast channel. The individual codes as well as the overall system can be represented by a joint factor graph consisting of a source code at each node, a channel code for the each uplink and a channel code for the downlink. The optimality of separation of source and channel coding implies that it is optimal to individually design these codes. We focus on low-density parity-check codes where code design corresponds to the optimisation of their degree distributions.
2011 IEEE Data Compression Conference (DCC 2011). Proceedings of the IEEE Data Compression Conference, DCC 2011 (Snowbird, UT 29-31 March, 2011) p. 466