On the maximum a posteriori decoding thresholds of multiuser systems with erasures

A fundamental connection between the belief propagation (BP) and maximum a posteriori (MAP) decoding thresholds was derived by Méasson, Montanari, and Urbanke using the area theorem for extrinsic information transfer (EXIT) curves. This connection allows the MAP threshold, for the binary erasure channel, to be evaluated efficiently via an upper bound that can be shown to be tight in some cases. In this paper, a similar analysis is used to extend these results to several multiuser systems, namely a noisy Slepian-Wolf problem and a multiple-access channel with erasures. The simplicity of these channel models allows for rigorous analysis and enables the derivation of upper bounds on the MAP thresholds using EXIT area theorems. In some cases, one can also show these bounds are tight. One interesting application is that the MAP thresholds can be compared with the BP thresholds of spatially-coupled codes to verify threshold saturation for the corresponding systems.