Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Real-Time Edge Computing for Road Surface Classification Using Multi-IMU Data and a Hybrid CNN-LSTM Classification

Luis A Arce-Saenz1, Luis A Salazar-Calderón1, Renato Galluzzi1,2

  • 1School of Engineering and Sciences, Tecnologico de Monterrey, Mexico City 14380, Mexico.

Sensors (Basel, Switzerland)
|July 15, 2026
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Adequate Segmentation in Marker-Based Motion Capture Studies for Hyperflexion and Hyperextension Lumbar Exercises.

Bioengineering (Basel, Switzerland)·2025
Same author

Demystifying Infusion Pumps: Design of a Cost-Effective Platform for Education and Innovation.

Journal of diabetes science and technology·2025
Same author

Optical Marker-Based Motion Capture of the Human Spine: A Scoping Review of Study Design and Outcomes.

Annals of biomedical engineering·2024
Same author

Development of a Framework for the Communication System Based on KNX for an Interactive Space for UX Evaluation.

Sensors (Basel, Switzerland)·2023
Same author

Development of an Intelligent System for the Monitoring and Diagnosis of the Well-Being.

Sensors (Basel, Switzerland)·2022
Same author

Visualizing Street Pavement Anomalies through Fog Computing V2I Networks and Machine Learning.

Sensors (Basel, Switzerland)·2022

This study optimized deep learning models for real-time road quality monitoring using vibration data. The edge-optimized system significantly reduced processing time, enabling effective road condition classification and anomaly detection for enhanced vehicle safety.

Area of Science:

  • Engineering
  • Computer Science
  • Transportation Science

Background:

  • Road quality monitoring is crucial for vehicle safety, ride comfort, and advanced driver-assistance systems.
  • Deep learning models excel at road surface classification but require optimization for real-time embedded applications.
  • Effective road feature knowledge supports proactive measures for both vehicles and infrastructure.

Purpose of the Study:

  • To deploy an edge-optimized deep learning model for real-time road condition classification using vibration data.
  • To evaluate the system's performance in terms of inference latency and classification accuracy.
  • To demonstrate the feasibility of real-time temporal deep learning models in a cyber-physical system.

Main Methods:

  • A hybrid model combining convolutional neural networks (CNNs) and long short-term memory (LSTM) networks was utilized.
Keywords:
deep learningedge computinggeospatial mappingintelligent transportation systemsreal-time processingroad surface monitoring

Related Experiment Videos

  • Continuous vibration data from multiple inertial measurement units (IMUs) were processed.
  • The system was deployed on a MicroAutoBox III Embedded PC for edge processing at speeds of 5.0–25.0 km/h.

    • Main Results:

    • Inference latency was reduced by 88% (from 33.8 ms to 4.05 ms) compared to offline deployment.
    • A weighted-average F1-score of 0.8751 was achieved in real-world conditions, meeting the 86 Hz sensor polling rate requirement.
    • Geospatial mapping successfully localized structural anomalies, demonstrating robustness to varying lighting conditions.

    Conclusions:

    • The edge-optimized deep learning architecture enables real-time road condition classification within strict processing limits.
    • The cyber-physical system proves the feasibility of deploying temporal deep learning models for road monitoring.
    • Future research will focus on highway-speed validation and domain adaptation for broader applicability.