One-Hop Listening Based ARQs for Low-Latency Communication in Multi-Hop Networks

TL;DR

This paper introduces one-hop and multi-hop listening ARQ strategies for multi-hop wireless networks, achieving high reliability and low latency by utilizing unused ARQs across nodes, with optimized distribution to minimize packet drops.

Contribution

It proposes novel cooperative ARQ strategies with one-hop and multi-hop listening capabilities, improving reliability without increasing latency, and provides optimization solutions for ARQ distribution.

Findings

Strategies outperform existing methods in packet drop probability.

Theoretical analysis confirms improved reliability and delay profiles.

Optimization effectively minimizes packet drops under latency constraints.

Abstract

Inspired by emerging applications in vehicular networks, we address the problem of achieving high-reliability and low-latency communication in multi-hop wireless networks. We propose a new family of Automatic-Repeat-Requests (ARQs) based cooperative strategies wherein high end-to-end reliability is obtained using packet re-transmissions at each hop while the low-latency constraint is met by imposing an upper bound on the total number of packet retransmissions across the network. A hallmark of our strategies is the one-hop listening capability wherein nodes utilize the unused ARQs of their preceding node just by counting the number of failed attempts due to decoding errors. We further extend the idea of one-hop listening to multi-hop listening, wherein a set of consecutive nodes form clusters to utilize the unused ARQs of the preceding nodes, beyond its nearest neighbour, to further improve reliability. Thus, our strategies provide the high-reliability feature with no compromise in the original latency-constraint. For the proposed strategies, we solve non-linear optimization problems on distributing the ARQs across the nodes so as to minimize packet drop probability (PDP) subject to a total number of ARQs in the network. Through extensive theoretical results on PDP and delay profiles, we show that the proposed strategies outperform the best-known strategies in this space.