Coding Schemes for Discrete Memoryless Multicast Networks with Rate-limited Feedback

TL;DR

This paper introduces new coding schemes for discrete memoryless multicast networks with rate-limited feedback, improving upon existing methods like noisy network coding and distributed decode-forward coding, and extends ideas to feedback-free scenarios.

Contribution

The paper proposes novel coding schemes utilizing block-Markov, superposition, and hybrid strategies for multicast networks with feedback, achieving higher rates than prior methods.

Findings

Schemes recover known results for relay channels with feedback.

Proposed schemes outperform noisy network coding in certain settings.

New feedback-free scheme still improves existing bounds for some channels.

Abstract

Coding schemes for discrete memoryless multicast networks (DM-MN) with rate-limited feedback from the receivers and relays to the transmitter are proposed. The schemes improve over the noisy network coding proposed by Lim et al.. For the single relay channel with relay-transmitter feedback, our coding schemes recover Gabbai and Bross's results, and strictly improve on noisy network coding, distributed decode-forward coding proposed by Lim et al., and all known lower bounds on the achievable rates proposed for the setup without feedback. The coding schemes are based on block-Markov coding, superposition coding, sliding-window/backward decoding and hybrid relaying strategies. In our Scheme 1A, the relays and receivers use compress-forward strategy and send the compression indices to the transmitter through the feedback pipes. After obtaining the compression indices through feedback, the transmitter sends them together with the source message. Each receiver uses backward decoding to jointly decode the source message and all compression indices. Our Scheme 1B is similar, except that here each relay not only uses compress forward to compress its channel outputs, but also performs partial decode forward to decode a common part of source message. Our Scheme 1C allows different relays to decode different parts of the source message (in Scheme 1A, no relay decodes any part of the source message; in Scheme 1B, all relays decode the same part of source message), which can achieve higher rates than Scheme 1A and Scheme 1B. Motivated by the feedback coding scheme, we propose a new coding scheme for DM-MN without feedback, where the transmitter decodes the compression indices instead of obtaining them directly through feedback. It is shown that the scheme still improves noisy network coding and distributed decode-forward coding for some channels.