Achievable Rate Regions of the Butterfly Network With Noisy Links and End-to-End Error Correction
In this paper, the authors characterize achievable rate regions of different decoding strategies on the butterfly network with noisy links and end-to-end error correction. It has been shown, by simulations with low-density parity-check codes that joint decoding schemes can outperform serial decoding, which in turn outperforms independent decoding. In this paper, they derive the information-theoretic achievable rates of these three decoding strategies in the Shannon sense. They consider relays with minimal processing power. They are unable to decode received signals and can only perform simple operations on the received signals, e.g., adding and scaling, before relaying. This can be used to model low-complexity networks where nodes with limited capability are used to meet size, cost, and power constraints.