Three-terminal Erasure Source-Broadcast with Feedback

TL;DR

This paper investigates the impact of feedback in a three-receiver erasure broadcast channel, proposing hybrid coding schemes to minimize transmission latency while meeting diverse user requirements.

Contribution

It introduces a queue-based hybrid digital-analog coding scheme and novel chaining algorithms for optimal or near-optimal performance in a three-user erasure broadcast setting with feedback.

Findings

Optimal performance achieved for some users with analog transmissions.

Linear programming characterizes the number of analog transmissions possible.

New chaining algorithm improves decoding strategies for multiple users.

Abstract

We study the effects of introducing a feedback channel in the erasure source-broadcast problem for the case of three receivers. In our problem formulation, we wish to transmit a binary source to three users over the erasure broadcast channel when causal feedback is universally available to the transmitter and all three receivers. Each receiver requires a certain fraction of the source sequence, and we are interested in the minimum latency, or transmission time, required to serve them all. We propose a queue-based hybrid digital-analog coding scheme that achieves optimal performance for the duration of analog transmissions. We characterize the number of analog transmissions that can be sent as the solution to a linear program and furthermore give sufficient conditions for which optimal performance can be achieved by all users. In some cases, we find that users can be point-to-point optimal regardless of their distortion constraints. When the analog transmissions are insufficient in meeting user demands, we propose two subsequent coding schemes. The first uses a queue preprocessing strategy for channel coding. The second method is a novel chaining algorithm, which involves the transmitter targeting point-to-point optimal performance for two users as if they were the only users in the network. Meanwhile, the third user simultaneously builds "chains" of symbols that he may at times be able to decode based on the channel conditions, and the subsequent reception of additional symbols.