Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Unsupervised Graph-based Learning Method for Sub-band Allocation in 6G Subnetworks

Daniel Abode, Ramoni Adeogun, Lou Salaün, Renato Abreu, Thomas Jacobsen, Gilberto Berardinelli

TL;DR

This work tackles sub-band allocation in dense 6G subnetworks by modeling interference as a graph and solving a graph-coloring-like problem with an unsupervised GGNN approach. A interference graph is constructed by linking each subnetwork to its $K-1$ strongest interferers, and a Potts-model-inspired loss $\\ ext{\\Psi}(\\hat{\\theta})=\\sum_{(n,m)\\in \\mathcal{E}} \\hat{\\theta}_n^T \\hat{\\theta}_m$ guides the learning without ground-truth labels. The GGNN learns to classify nodes into sub-band choices, providing results competitive with centralized greedy coloring while reducing signaling and runtime; it also shows robustness to density and channel-model variations and can operate in centralized or decentralized modes. This approach offers scalable, data-driven interference coordination suitable for large-scale subnetworks in dynamic 6G environments, with potential extensions to traffic and mobility scenarios.

Abstract

In this paper, we present an unsupervised approach for frequency sub-band allocation in wireless networks using graph-based learning. We consider a dense deployment of subnetworks in the factory environment with a limited number of sub-bands which must be optimally allocated to coordinate inter-subnetwork interference. We model the subnetwork deployment as a conflict graph and propose an unsupervised learning approach inspired by the graph colouring heuristic and the Potts model to optimize the sub-band allocation using graph neural networks. The numerical evaluation shows that the proposed method achieves close performance to the centralized greedy colouring sub-band allocation heuristic with lower computational time complexity. In addition, it incurs reduced signalling overhead compared to iterative optimization heuristics that require all the mutual interfering channel information. We further demonstrate that the method is robust to different network settings.

Unsupervised Graph-based Learning Method for Sub-band Allocation in 6G Subnetworks

TL;DR

This work tackles sub-band allocation in dense 6G subnetworks by modeling interference as a graph and solving a graph-coloring-like problem with an unsupervised GGNN approach. A interference graph is constructed by linking each subnetwork to its strongest interferers, and a Potts-model-inspired loss guides the learning without ground-truth labels. The GGNN learns to classify nodes into sub-band choices, providing results competitive with centralized greedy coloring while reducing signaling and runtime; it also shows robustness to density and channel-model variations and can operate in centralized or decentralized modes. This approach offers scalable, data-driven interference coordination suitable for large-scale subnetworks in dynamic 6G environments, with potential extensions to traffic and mobility scenarios.

Abstract

In this paper, we present an unsupervised approach for frequency sub-band allocation in wireless networks using graph-based learning. We consider a dense deployment of subnetworks in the factory environment with a limited number of sub-bands which must be optimally allocated to coordinate inter-subnetwork interference. We model the subnetwork deployment as a conflict graph and propose an unsupervised learning approach inspired by the graph colouring heuristic and the Potts model to optimize the sub-band allocation using graph neural networks. The numerical evaluation shows that the proposed method achieves close performance to the centralized greedy colouring sub-band allocation heuristic with lower computational time complexity. In addition, it incurs reduced signalling overhead compared to iterative optimization heuristics that require all the mutual interfering channel information. We further demonstrate that the method is robust to different network settings.
Paper Structure (17 sections, 12 equations, 5 figures, 2 tables)

This paper contains 17 sections, 12 equations, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. System Model and Problem Formulation
  3. Subnetworks System Model
  4. Interference Graph Model of Subnetworks
  5. Sub-band Allocation as a Node Classification Task
  6. Sub-band Allocation using Graph Neural Network
  7. Gated Graph Neural Network Architecture
  8. Training and Execution Procedure
  9. Results and Discussion
  10. Simulation Assumptions
  11. Subnetwork Deployment
  12. GGNN Model and Training Settings
  13. Benchmarks
  14. Network Performance Evaluation
  15. Complexity Analysis
  16. ...and 2 more sections

Figures (5)

  • Figure 1: Deployment of N subnetworks with J devices
  • Figure 2: Graph Neural Network Design
  • Figure 3: Cumulative distribution function (CDF) of the sum spectral efficiency for 10000 test snapshots
  • Figure 4: Cumulative distribution function (CDF) of the per-device spectral efficiency for 10000 test snapshots
  • Figure 5: Computational runtime for different number of subnetworks (ms) averaged over 10000 realizations