Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Congestion-aware Distributed Task Offloading in Wireless Multi-hop Networks Using Graph Neural Networks

Zhongyuan Zhao, Jake Perazzone, Gunjan Verma, Santiago Segarra

TL;DR

The paper tackles congestion-aware computational offloading in wireless multi-hop networks by integrating a graph neural network into a fully distributed offloading framework. It models the network as $\mathcal{G}^{n}$ and $\mathcal{G}^{c}$, extends to $\mathcal{G}^{e}$ and $\mathcal{G}^{\ell}$, and formulates a latency objective $\boldsymbol{1}^{\top}\mathbf{u}^{j}$ that is NP-hard as a generalized assignment problem. A GCNN predicts congestion-aware extended-link weights $\hat{\boldsymbol{\delta}}^{\ell}$ and a distributed procedure $\varphi$ estimates per-packet latency, enabling near-distributed offloading decisions. Numerical experiments on synthetic Barabasi-Albert networks show substantial latency reductions and congestion mitigation compared with a context-agnostic baseline, supporting the practicality of edge intelligence for congestion-aware offloading.

Abstract

Computational offloading has become an enabling component for edge intelligence in mobile and smart devices. Existing offloading schemes mainly focus on mobile devices and servers, while ignoring the potential network congestion caused by tasks from multiple mobile devices, especially in wireless multi-hop networks. To fill this gap, we propose a low-overhead, congestion-aware distributed task offloading scheme by augmenting a distributed greedy framework with graph-based machine learning. In simulated wireless multi-hop networks with 20-110 nodes and a resource allocation scheme based on shortest path routing and contention-based link scheduling, our approach is demonstrated to be effective in reducing congestion or unstable queues under the context-agnostic baseline, while improving the execution latency over local computing.

Congestion-aware Distributed Task Offloading in Wireless Multi-hop Networks Using Graph Neural Networks

TL;DR

The paper tackles congestion-aware computational offloading in wireless multi-hop networks by integrating a graph neural network into a fully distributed offloading framework. It models the network as and , extends to and , and formulates a latency objective that is NP-hard as a generalized assignment problem. A GCNN predicts congestion-aware extended-link weights and a distributed procedure estimates per-packet latency, enabling near-distributed offloading decisions. Numerical experiments on synthetic Barabasi-Albert networks show substantial latency reductions and congestion mitigation compared with a context-agnostic baseline, supporting the practicality of edge intelligence for congestion-aware offloading.

Abstract

Computational offloading has become an enabling component for edge intelligence in mobile and smart devices. Existing offloading schemes mainly focus on mobile devices and servers, while ignoring the potential network congestion caused by tasks from multiple mobile devices, especially in wireless multi-hop networks. To fill this gap, we propose a low-overhead, congestion-aware distributed task offloading scheme by augmenting a distributed greedy framework with graph-based machine learning. In simulated wireless multi-hop networks with 20-110 nodes and a resource allocation scheme based on shortest path routing and contention-based link scheduling, our approach is demonstrated to be effective in reducing congestion or unstable queues under the context-agnostic baseline, while improving the execution latency over local computing.
Paper Structure (5 sections, 6 equations, 2 figures, 1 algorithm)

This paper contains 5 sections, 6 equations, 2 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. System Model and Problem Formulation
  3. Distributed Task Offloading with GNNs
  4. Numerical experiments
  5. Conclusions

Figures (2)

  • Figure 1: Challenges in distributed multi-hop offloading: (a) probing: nodes $1$ and $2$ query the communication and computing bandwidth of three servers. (b) offloading: nodes $1$ and $2$ both select server A based on collected information, however, such decisions lead to congestion at both the server A and link (3,A) in execution.
  • Figure 2: (a) Scale of instances: the numbers of edge nodes, server nodes, relay nodes, and tasks by network size. (b) Task congestion probability by network size with $95\%$ confidence interval. (c) Average per-task execution latency ratio w.r.t. the baseline policy across network sizes.