Achieving Adaptive sum-capacity of fading MACs with distributed CSI and non-identical links by rate-splitting

We consider a two user slow fading Gaussian MAC where each transmitter knows only its own channel fading coefficient prior to transmission and where the receiver knows the fading coefficients of both users. The slow fading channel is modeled as a block fading channel. Each user performs block by block encoding for communication with the common receiver. The receiver decodes the messages of both users from the received signal, also on a block by block basis. Each user has its own fixed average power constraint in each block. A useful metric to measure the throughput performance in such a scenario is the adaptive sum capacity which is the maximum expected sum throughput that can be achieved. The adaptive sum capacity in such a setting was derived by the authors in [4]. We show that the adaptive sum capacity can be achieved by an asymptotically optimal rate-splitting scheme with low complexity single-user decoding. This eliminates the need for joint decoding of the two users which, although optimal, has a high complexity.