← Back to Research

Event-Driven and Edge AI for Structural Health Monitoring

Edge AISHMIoTEvent-driven sensingQualcomm
Summary figure for Event-Driven and Edge AI for Structural Health Monitoring

Overview

This research theme groups two related but distinct projects that both pursue power-efficient structural health monitoring—reducing continuous sensing, energy use, and network load—while attacking the problem from different layers of the stack.

Together they span vision-triggered edge intelligence and ultra-low-power wireless sensing hardware, and can be combined into an end-to-end pipeline from detection to local sensing to cloud analytics.

Projects under this theme

Complementary projects sit under this umbrella. Each has its own focus and materials below.

  1. Project 1

    Semantic Edge AI and Event-Driven Wake-Up Sensing

    In collaboration with Omar Khalifa and Ruslan Zhagypar, and under the guidance of Prof. Gilles Lubineau and Prof. Tareq Al-Naffouri, this track uses Qualcomm platforms for event-driven SHM: drone cameras detect cracks via semantic segmentation, and a wake-up transmitter activates local sensors only when anomalies are flagged. A split AI architecture keeps feature extraction at the edge and transmits compact feature tensors instead of raw frames, cutting end-to-end latency by about 56% and saving about 83% bandwidth on Qualcomm RB5 and RB3 Gen2 platforms. Presented and demonstrated at IEEE PIMRC, with related work under review for the IEEE ComSoc Student Competition.

    KAUSTQualcomm
    Pipeline for Project 1: semantic perception at the edge, split semantic computation with compressed feature tensors, and event-driven wake-up radio activation of ultra-low-power IoT sensors.
    Pipeline for Project 1: semantic perception at the edge, split semantic computation with compressed feature tensors, and event-driven wake-up radio activation of ultra-low-power IoT sensors.

  2. Project 2

    Hybrid Event-Driven SHM Node with Ultra-Low Power Adaptive Triggering

    A second project focused on the sensing and radio node itself: a hybrid event-driven structural health monitoring node with ultra-low-power adaptive triggering. The KAUST team developed and demonstrated this system in the EWSHM 2026 Young Professionals Challenge.

    KAUST
    Université Gustave Eiffel, Capturia, LAAS CNRS, AFENDA FrANDTB
    Hybrid event-driven SHM demo setup: gateway (dashboard, microcontroller, wake-up transmitter, wind-sensor mock-up) and IoT device (sensor, ESP32, wake-up radio, power profiler).
    Hybrid event-driven SHM demo setup: gateway (dashboard, microcontroller, wake-up transmitter, wind-sensor mock-up) and IoT device (sensor, ESP32, wake-up radio, power profiler).

Videos

  • Track 1 — PIMRC Demo: Event-Driven Edge-AI SHM

    Semantic edge-AI track: demonstration of the event-driven system presented at IEEE PIMRC.

  • Track 2 — EWSHM 2026 Young Professionals Challenge

    Ultra-low-power node track: competition video for the Hybrid Event-Driven SHM Node with Ultra-Low Power Adaptive Triggering.

Related publications

  • When Meaning Controls the Network: Event-Driven Semantic Communication at the Edge
    R. Zhagypar, O. Khalifa, H. A. Mahmoud, A. Rongoni, N. Kouzayha, G. Lubineau, A. Alloum, W. Doré, T. Y. Al-Naffouri · Under review — IEEE ComSoc Student Competition (2025)

Awards

  • EWSHM 2026 Young Professionals Challenge Award
    EWSHM 2026 Young Professionals Challenge Award