Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Improving Wi-Fi Network Performance Prediction with Deep Learning Models

Gabriele Formis, Amanda Ericson, Stefan Forsstrom, Kyi Thar, Gianluca Cena, Stefano Scanzio

TL;DR

The paper tackles the challenge of achieving robust industrial Wi-Fi performance by predicting the frame delivery ratio (FDR) using deep learning. It systematically compares CNN, LSTM, and Bi-LSTM on a large real-world 2.4 GHz Wi-Fi dataset across four channels, analyzing accuracy and computational efficiency for embedded deployment. Key findings show CNN offers strong real-time performance and suffices in many conditions, while LSTM and Bi-LSTM can provide accuracy gains in highly dynamic channels at the cost of higher latency and memory use; training on multi-channel data enhances generalization. The work demonstrates that proactive FDR prediction enables runtime parameter adaptation and supports practical edge deployments in industrial settings, with future directions including hybrid models and broader-band validations.

Abstract

The increasing need for robustness, reliability, and determinism in wireless networks for industrial and mission-critical applications is the driver for the growth of new innovative methods. The study presented in this work makes use of machine learning techniques to predict channel quality in a Wi-Fi network in terms of the frame delivery ratio. Predictions can be used proactively to adjust communication parameters at runtime and optimize network operations for industrial applications. Methods including convolutional neural networks and long short-term memory were analyzed on datasets acquired from a real Wi-Fi setup across multiple channels. The models were compared in terms of prediction accuracy and computational complexity. Results show that the frame delivery ratio can be reliably predicted, and convolutional neural networks, although slightly less effective than other models, are more efficient in terms of CPU usage and memory consumption. This enhances the model's usability on embedded and industrial systems.

Improving Wi-Fi Network Performance Prediction with Deep Learning Models

TL;DR

The paper tackles the challenge of achieving robust industrial Wi-Fi performance by predicting the frame delivery ratio (FDR) using deep learning. It systematically compares CNN, LSTM, and Bi-LSTM on a large real-world 2.4 GHz Wi-Fi dataset across four channels, analyzing accuracy and computational efficiency for embedded deployment. Key findings show CNN offers strong real-time performance and suffices in many conditions, while LSTM and Bi-LSTM can provide accuracy gains in highly dynamic channels at the cost of higher latency and memory use; training on multi-channel data enhances generalization. The work demonstrates that proactive FDR prediction enables runtime parameter adaptation and supports practical edge deployments in industrial settings, with future directions including hybrid models and broader-band validations.

Abstract

The increasing need for robustness, reliability, and determinism in wireless networks for industrial and mission-critical applications is the driver for the growth of new innovative methods. The study presented in this work makes use of machine learning techniques to predict channel quality in a Wi-Fi network in terms of the frame delivery ratio. Predictions can be used proactively to adjust communication parameters at runtime and optimize network operations for industrial applications. Methods including convolutional neural networks and long short-term memory were analyzed on datasets acquired from a real Wi-Fi setup across multiple channels. The models were compared in terms of prediction accuracy and computational complexity. Results show that the frame delivery ratio can be reliably predicted, and convolutional neural networks, although slightly less effective than other models, are more efficient in terms of CPU usage and memory consumption. This enhances the model's usability on embedded and industrial systems.

Paper Structure

This paper contains 16 sections, 5 equations, 2 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Method
  4. Data collection
  5. Data Preprocessing
  6. Model Selection
  7. Implementation
  8. Models
  9. Hyperparameter tuning
  10. Selected hyperparameters
  11. Results
  12. CNN accuracy
  13. LSTM / Bi-LSTM accuracy
  14. Computational Complexity
  15. Discussion
  16. ...and 1 more sections

Figures (2)

  • Figure 1: Main steps of the proposed methodology for predicting Wi-Fi network performance.
  • Figure 2: Box plot showing the loss on the validation dataset $\mathcal{V}$ for channel 1 concerning the seven best configurations related to the CNN model.