Reliability of Relay Networks under Random Linear Network Coding

TL;DR

This paper analyzes the reliability of relay networks employing Random Linear Network Coding, providing a new upper bound for decoding success probability that accounts for multiple relays, correlation, and arbitrary field sizes, validated by simulations.

Contribution

It introduces a novel upper bound for decoding success probability in relay networks with RLNC, accurate for any number of relays and considering relay correlation.

Findings

Proposed upper bound is highly accurate with mean square error as low as 10^{-6}.

Demonstrated throughput gains over alternative coding strategies.

Bound becomes exact for a single relay.

Abstract

We consider a single-source, multiple-relay, single-destination lossy network employing Random Linear Network coding at all transmitting nodes. We address the problem of calculating the probability of successful decoding at the destination node. In contrast with some previous studies, we assume the classical RLNC scheme, in which the relaying nodes simply re-encode packets, without resorting to decoding. In addition, we consider an arbitrary field size and take into account correlation between the relay nodes. We propose a novel upper bound for an arbitrary number of relays, which becomes exact for a single relay. Using Monte Carlo simulations, we show that the proposed bound is very accurate, exhibiting the mean square error as low as $10^{-6}$. We also demonstrate the throughput gain of the proposed scheme over alternative coding and relaying strategies.