IEEE Publication: Energy-Efficient ML for IIoT Systems
I'm pleased to announce my contribution as co-author to a paper published in IEEE Internet of Things Magazine
Publication Details
Title: "Energy Efficiency Improvement Methods for Edge AI IIoT Networks: An Overview"
Published in: IEEE Internet of Things Magazine
IEEE Xplore: ieeexplore.ieee.org/document/11244844
Authors
Multi-institutional collaboration across France and Japan:
| Author | Affiliation |
|---|---|
| Yousef N. Shnaiwer | NII, Tokyo, Japan — now Al-Zaytoonah University, Amman, Jordan |
| Julien Weber | R&D Team, Wavely, Villeneuve d'Ascq, France |
| Julien Roland | R&D Team, Wavely, Villeneuve d'Ascq, France |
| Megumi Kaneko | NII, Tokyo, Japan — The University of Tokyo |
| Robin Gerzaguet | GRANIT team, IRISA, University of Rennes, France |
| Kenichi Kawamura | NTT Inc., Kanagawa, Japan |
| Olivier Berder | GRANIT team, IRISA, University of Rennes, France |
| Salah Berra | NII, Tokyo, Japan |
| Pascal Scalart | GRANIT team, IRISA, University of Rennes, France |
| Keisuke Wakao | NTT Inc., Kanagawa, Japan |
| Yasushi Takatori | NTT Inc., Kanagawa, Japan |
Abstract
The Industrial Internet of Things (IIoT) is transforming modern industries by enabling intelligent, data-driven operations at the edge of networks. As these systems grow in scale and complexity, optimizing the energy consumption of Machine Learning (ML) techniques becomes essential for sustainable and reliable performance.
This article examines the architectural foundations of IIoT systems and offers a structured classification of methods for reducing the energy use of ML. A real-world case study — the Energy Efficient Internet of Emergency Services — illustrates the practical deployment of energy-optimized IIoT infrastructures in demanding operational contexts.
The article concludes with a discussion of current challenges and future directions focused on:
- Ultra-low-power architectures for embedded inference
- Scalable deployment strategies across heterogeneous edge fleets
- Robust security for next-generation industrial intelligence
Key Contributions
- Architectural Survey: Structured overview of IIoT system designs and their energy profiles
- ML Energy Taxonomy: Classification of techniques to reduce ML energy consumption at the edge (pruning, quantization, knowledge distillation, hardware-aware NAS)
- Case Study — IoES: Real-world deployment of energy-optimized ML in emergency services infrastructure
- Research Roadmap: Identified open challenges in ultra-low-power ML, scalability, and edge security
Context
This work was carried out in collaboration with Wavely (France), IRISA / University of Rennes (France), National Institute of Informatics (Japan), and NTT — Nippon Telegraph and Telephone (Japan), within the ANR LIGHT-SWIFT research program — the same program driving our embedded acoustic sensor development.