Iterative multiuser joint decoding : unified framework and asymptotic analysis

We present a framework for iterative multiuser joint decoding based on the application of the sum-product algorithm to the factor graph representation of the a posteriori joint probability mass function of the users information bits. Several low-complexity algorithms previously proposed based on parallel and serial hard and soft interference cancellation are derived in a simple and unified way. A wide class of these algorithms is analyzed by using the approach ofdensity evolution on graphs combined with results from the theory of large random matrices. In the case of parallel interference cancellation and equal power users we present a simple approximated analysis able to explain qualitatively (and often quantitatively) the behavior of iterative decoding via the study of the stable fixed points of a one- dimensional non-linear dynamical system.