Selective Coding Strategy for composite relay channels

Consider the composite relay channel (CRC) where the channel is randomly drawn from a set of conditional distributions with index θ ϵ Θ, which represents the vector of channel parameters. The specific draw θ is assumed to be unknown at the source, fully known at the destination and only partly known at the relay. In this setting, the coding would commonly be chosen regardless of the channel measurement at the relay end. This work introduces a novel coding which enables the relay to select-based on its channel measurement - the best coding strategy between Compress-and-Forward (CF) and Decode-and-Forward (DF). Indeed, provided that the channel source-to-relay is good enough for decoding the message, the relay decides on DF and otherwise it may switch to CF. The proposed Selective Coding Strategy (SCS) is based on superposition coding, DF and CF schemes, backward and joint decoding. We derive bounds on the asymptotic average error probability of the memoryless CRC. An application example to the fading Gaussian CRC is also investigated.