Overview

INFINITY (Intelligent and Frugal Infrastructure for Innovation and Technology) is a research project funded through the AIRE 2026 program and jointly developed by PROMES-CNRS and the Génie Industriel et Maintenance (GIM) department of the University of Perpignan Via Domitia.

The project investigates intelligent and frugal diagnostic methodologies for Internet of Things (IoT) sensor networks within the emerging paradigm of Maintenance 5.0. Unlike conventional maintenance strategies that focus on equipment failures, INFINITY shifts the attention toward the reliability and health of the sensor networks themselves.

The project aims to develop self-diagnostic capabilities capable of detecting sensor drifts, signal inconsistencies, communication losses and measurement anomalies through the combination of physical coherence models and lightweight artificial intelligence algorithms.

Funding

AIRE 2026 Research Initiative

PROMES-CNRS

Participating Institutions

PROMES-CNRS (UPR 8521)

University of Perpignan Via Domitia (UPVD)

Génie Industriel et Maintenance (GIM)

Project Team

• Edgar Hernando Sepúlveda-Oviedo
• François Vernay

Scientific Context

The rapid deployment of connected industrial systems and IoT infrastructures has transformed modern maintenance strategies. However, the reliability of these systems increasingly depends on the quality and consistency of the measurements provided by distributed sensor networks.

Sensor drifts, communication losses, calibration errors, noisy measurements and synchronization issues may significantly affect monitoring systems and compromise decision-making processes.

Current Maintenance 5.0 approaches primarily focus on diagnosing failures occurring in industrial assets. Comparatively little attention has been devoted to diagnosing the health of the sensing infrastructure itself.

INFINITY addresses this gap by developing methodologies capable of monitoring, validating and diagnosing the reliability of distributed IoT sensor networks in real operating conditions.

Main Objectives

The project pursues the following scientific objectives:

• Detect sensor drifts and measurement inconsistencies.
• Develop coherence indicators for distributed sensor networks.
• Investigate lightweight diagnostic algorithms suitable for Edge AI applications.
• Combine physical coherence models with data-driven approaches.
• Quantify uncertainty associated with sensor measurements.
• Develop self-diagnostic capabilities for IoT infrastructures.
• Improve reliability and resilience of Maintenance 5.0 systems.

Methodology

The project follows a progressive experimental and analytical methodology composed of four main stages.

Stage 1 – Instrumentation and Data Acquisition

Deployment and operation of an IoT sensor network based on LoRa communication technologies.

The experimental infrastructure includes:

• Environmental sensors
• Electrical sensors
• Energy monitoring devices
• Vibration sensors
• Distributed acquisition nodes

Special attention is given to connectivity losses, synchronization issues and data traceability.

Stage 2 – Drift Characterization and Statistical Analysis

Identification and characterization of sensor drifts using statistical signal analysis techniques.

Investigated approaches include:

• Change-point detection
• CUSUM analysis
• EWMA monitoring
• Distribution analysis
• Uncertainty quantification
• Sensor-to-sensor correlation analysis

The objective is to identify weak drifts before they become critical.

Stage 3 – Intelligent Diagnostic Module

Development of a hybrid diagnostic framework combining physical coherence models with lightweight machine learning algorithms.

Methods under investigation include:

• Incremental PCA
• Sparse learning
• Regularized regression
• Physical consistency indicators
• Edge AI architectures

The resulting framework aims to provide interpretable and computationally efficient diagnostics.

Stage 4 – Experimental Validation

Validation of the developed methodologies on real sensor networks deployed at:

• IUT de Perpignan
• PROMES-CNRS Tecnosud facilities
• PROMES-CNRS Odeillo facilities

Cross-validation campaigns will be conducted to assess robustness and transferability.

Scientific Contributions

INFINITY introduces several innovative aspects:

• Diagnosis of sensor networks rather than monitored assets.
• Combination of physical and statistical coherence indicators.
• Integration of uncertainty-aware diagnostics.
• Edge AI implementation under resource constraints.
• Experimental validation on real industrial IoT infrastructures.

Expected Outcomes

The project is expected to deliver:

• A methodology for intelligent diagnosis of IoT sensor networks.
• Novel coherence indicators for distributed sensing systems.
• A frugal self-diagnostic module.
• Edge AI implementations.
• Experimental validation campaigns.
• Scientific communications.
• Foundations for future ANR, BQR and FREE projects.

Scientific Impact

INFINITY contributes to the emerging fields of:

• Maintenance 5.0
• Industrial Artificial Intelligence
• Edge Computing
• Smart Monitoring Systems
• Distributed Diagnostics
• Industrial Internet of Things (IIoT)

The methodologies developed within the project may be transferred to photovoltaic systems, thermal plants, smart grids, industrial facilities and cyber-physical systems requiring reliable distributed sensing infrastructures.

Educational Impact

The project directly supports practical training activities within the GIM department by providing students with access to a real experimental platform dedicated to IoT technologies, industrial monitoring and intelligent diagnostics.

The resulting infrastructure will contribute to strengthening research-teaching interactions and promoting project-based learning activities.

Keywords

Maintenance 5.0 · Internet of Things · Industrial IoT · Edge AI · Frugal Artificial Intelligence · Sensor Networks · Reliability · Predictive Maintenance · Distributed Diagnostics · Smart Monitoring · LoRa · Signal Processing

Project Status

Ongoing