Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Scalable Interference Graph Learning for Low-Latency Wi-Fi Networks using Hashing-based Evolution Strategy

Zhouyou Gu, Jihong Park, Jinho Choi

TL;DR

A scalable interference graph learning (IGL) framework that learns optimal interference graph representations for graph coloring-based RTWT scheduling is proposed, which improves slot efficiency and reduces packet losses by up to 30% in dynamic environments.

Abstract

Wi-Fi 7 introduces the restricted target wake time (RTWT) mechanism, which is vital for Industrial IoT (IIoT) applications requiring periodic, reliable, and low-latency communication. RTWT enables deterministic channel access by assigning scheduled transmission slots to stations (STAs), minimizing contention and interference. However, determining efficient RTWT slot assignments remains challenging in dense networks, where conventional interference graph-based models lack flexibility and scalability. To overcome this, we propose a scalable interference graph learning (IGL) framework that learns optimal interference graph representations for graph coloring-based RTWT scheduling. The IGL leverages an evolution strategy (ES) to train a neural network (NN) using a single network-wide reward, avoiding costly edge-wise feedback. Furthermore, a deep hashing function (DHF) groups interfering STAs, limiting training and inference to relevant subsets and greatly reducing complexity. Simulation results demonstrate that the proposed IGL improves slot efficiency by up to 25\%, reduces packet losses by up to 30\% in dynamic environments. Thanks to DHF, it also reduces the training and inference time of IGL by 4 and 8 times, respectively, and the online slot assignment time by 3 times in large networks.

Scalable Interference Graph Learning for Low-Latency Wi-Fi Networks using Hashing-based Evolution Strategy

TL;DR

A scalable interference graph learning (IGL) framework that learns optimal interference graph representations for graph coloring-based RTWT scheduling is proposed, which improves slot efficiency and reduces packet losses by up to 30% in dynamic environments.

Abstract

Wi-Fi 7 introduces the restricted target wake time (RTWT) mechanism, which is vital for Industrial IoT (IIoT) applications requiring periodic, reliable, and low-latency communication. RTWT enables deterministic channel access by assigning scheduled transmission slots to stations (STAs), minimizing contention and interference. However, determining efficient RTWT slot assignments remains challenging in dense networks, where conventional interference graph-based models lack flexibility and scalability. To overcome this, we propose a scalable interference graph learning (IGL) framework that learns optimal interference graph representations for graph coloring-based RTWT scheduling. The IGL leverages an evolution strategy (ES) to train a neural network (NN) using a single network-wide reward, avoiding costly edge-wise feedback. Furthermore, a deep hashing function (DHF) groups interfering STAs, limiting training and inference to relevant subsets and greatly reducing complexity. Simulation results demonstrate that the proposed IGL improves slot efficiency by up to 25\%, reduces packet losses by up to 30\% in dynamic environments. Thanks to DHF, it also reduces the training and inference time of IGL by 4 and 8 times, respectively, and the online slot assignment time by 3 times in large networks.

Paper Structure

This paper contains 54 sections, 2 theorems, 41 equations, 18 figures, 1 table, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Works
  3. Markov Model Approach
  4. Neural Network Model Approach
  5. Human Intuition-Defined Interference Graph Approach
  6. Interference Graph Learning Approach
  7. Contributions
  8. Notation and Paper Organization
  9. System Model
  10. Wi-Fi Network with Multi-AP Coordination
  11. CSMA/CA with RTWT
  12. Wi-Fi Network States
  13. Measured Network States
  14. Unmeasured Network States
  15. The Proposed IGL Framework
  16. ...and 39 more sections

Key Result

Proposition 1

There exists an adjacency matrix $\mathbf{E}^*$ of the graph whose coloring scheme, $\{\mathbf{z}^*, Z^*\} = \chi(\mathbf{E}^*)$, solves the contention and interference management problem in eq:prob:cni_management_problem. In other words, $\mathbf{z}^*$ and $Z^*$ represent the optimal slot assignmen

Figures (18)

  • Figure 1: Illustration of a Wi-Fi network for IIoT applications with the proposed interference graph learning scheme using the hashing-based evolution strategy.
  • Figure 2: System model of the Wi-Fi IIoT network.
  • Figure 3: Illustration of the slot assignments for $K=3$ STAs with $Z=2$ slots, where STA $1$ and $3$ are assigned in the same slot, e.g., $z_1=z_3=1$, and STA $2$ is assigned with a different one, e.g., $z_2=2$.
  • Figure 4: Iterative training of the IGL NN $\mu(\cdot|\theta^\mu)$ for the contention and interference management problem in the Wi-Fi network.
  • Figure 5: The structure of the IGL NN $\mu(\cdot|\theta^\mu)$. Here, the state embedding with parameters $\theta^{\mu}_{\mathrm{SE}}$, the predictors with parameters $\theta^{\mu}_{\mathrm{PC}}$ and $\theta^{\mu}_{\mathrm{PH}}$ are trained using unsupervised/supervised learning in Appendix C, and the edge generator with parameters $\theta^{\mu}_{\mathrm{EG}}$ are trained using the ES in Algorithm \ref{['alg:es-ggm']}. Note that $s_{i,\hat{a}_i}$, $s_{i,\hat{a}_j}$, and $s_{j,\hat{a}_j}$ are extracted from the state sequences $\mathcal{S}_i$ and $\mathcal{S}_j$.
  • ...and 13 more figures

Theorems & Definitions (5)

  • Remark 1
  • Proposition 1
  • proof
  • Theorem 1
  • proof