As IoT proliferates, the environmental and technical impact of billions of connected devices demands urgent attention. This workshop examines how AI-augmented agents should be designed, managed and coordinated while optimizing resource consumption and operating under frugal conditions and infrastructures. It also aims to review existing implementations of the Web of Things (WoT) standards with respect to long-term maintainability.
The workshop format will include a keynote talk, followed by long and short accepted paper presentations, based on peer-review.
Targeted audience
ATLAS targets researchers, practitioners, and policymakers interested in responsible WoT innovation.
Workshop participants will position their work within at least one of these key dimensions:
accurate decision-making in resource-aware agent architectures
sustainable design principles including life cycle management and modular components
efficient agent-mediated coordination protocols taking advantage of standardization and fog/edge architectures.
Participants will leave with practical strategies for implementing sustainable agentic systems and a collaborative network committed to advancing environmentally conscious IoT solutions.
Goals of the workshop
The primary goals of ATLAS are to:
Present cutting-edge research at the intersection of agent-based systems, IoT, and sustainability
Identify open research challenges and future directions for frugal, standards- compliant agentic systems
Build a community around sustainable WoT technologies
Workshop topics
The workshop topics include but are not limited to:
Federated learning for edge intelligence
Embedding intensive inference algorithms into constrained devices
Predictive maintenance to reduce e-waste
Frugal hardware-related concerns
Ethical frameworks for sustainable AI deployment
Resource-aware agentic LLM-based systems
Complexity of embedded semantic Web processes
Energy, network and/or memory management in decentralized Web-based architectures
Deployment of IoT / WoT platforms on Constrained devices
Lifecycle and maintenability of Cyber-Physical Systems
Robust and resilient distributed agentic systems
Environmental footprint of sensor and actuator-based infrastructures
Observability and governance of agents in decentralized architectures
Coordination architectures for agents with distributed knowledge
Application areas such as product passports and carbon footprint tracking
Pierre-Antoine Champin, W3C / LIRIS – Université Claude Bernard Lyon1, France
Victor Charpenay, Mines Saint-Étienne – LIMOS, France
Frédérique Laforest, LIRIS – INSA Lyon, France
Michael Mrissa, InnoRenew, Slovenia
Marta Sabou, Vienna University of Economics and Business, Austria
Ruben Taelman, Ghent University – imec, Belgium
Supported by
Workshop Program
-
Opening
-
Keynote
LETSGROW research: Soil monitoring for bio-fertilizer and treatment optimization Michael Mrissa, InnoRenew CoE, University of Primorska, Slovenia
-
Paper session I
[ATLAS] A Budget-Based Frugal Observability Governor for Decentralized Edge Agents: Design, Implementation, and Comparison Qiao Zhang - Orange, France
[BECS] An Edge-Cloud Framework for Scalable and Content-Aware Delivery of Whole Slide Images in Pathology Education Yu Sung Edward Kim - Korea Advanced Institute of Science and Technology (KAIST), South Korea Willmer Rafell Quinones Robles - Axinvent AI Inc., South Korea Katherine Po Sin Chung - KAIST, South Korea Nazar Novruzov - KAIST, South Korea Batyrkhan Mukhlis - KAIST, South Korea Byeong Uk Son - Axinvent AI Inc. Eunkyoung Jee - KAIST, South Korea Ji Young Park - Kyungpook National University, South Korea In-Young Ko - KAIST, South Korea
-
Lunch
-
Paper session II
[BECS] EDASC: Event-Driven Adaptive Stream Control for Edge-Cloud Environments Daniel Flores-Martin - COMPUTAEX. Extremadura Supercomputing Center, Spain Francisco Díaz-Barrancas - University of Extremadura, Spain Pedro J. Pardo - University of Extremadura, Spain Javier Berrocal - University of Extremadura, Spain Juan Manuel Murillo - University of Extremadura, Spain
[BECS] Post-Alignment Hard-Negative-Aware Contrastive Learning for Heterogeneous Defect Prediction in Big-Data Systems Sungu Lee - KAIST, South Korea Duksan Ryu - Jeonbuk National University, South Korea Jongmoon Baik - KAIST, South Korea
[ATLAS] Empowering Users toward Environmentally Sustainable Digital Practices: From Web-Based to Generative AI Interactions Mathyas Giudici - Politecnico di Milano, Italy Monica Vitali - Politecnico di Milano, Italy Marco Brambilla - Politecnico di Milano, Italy
[BECS] Towards a Client-Side HATEOAS-Based Recommender System Iuliia Kudiaeva - University of Primorska, Slovenia
-
Coffee Break & Poster Presentations
[BECS] Quality-Aware Structuring of Raw Robotic Stream Data Minjae Kim - Jeonbuk National University, South Korea Yunseo Choi - Jeonbuk National University, South Korea Duksan Ryu - Jeonbuk National University, South Korea Jongmoon Baik - KAIST, South Korea
[BECS] Pipelined Simulation for Autonomous Vehicle in Edge-Cloud Environment: A Case Study Byeongil Bae - KAIST, South Korea Eunchan Park - KAIST, South Korea In-Young Ko - KAIST, South Korea
-
Paper session III
[BECS] sLLM-Based QoS-Aware Server Recommendation for Mobile Edge Computing Yunseo Choi - Jeonbuk National University, South Korea Minjae Kim - Jeonbuk National University, South Korea Duksan Ryu - Jeonbuk National University, South Korea Jongmoon Baik - KAIST, South Korea
[BECS] A Spectrum of Switching Granularities for Energy-Aware Dynamic Neural Inference for Edge Devices Aviral Sharma - Indian Institute of Technology, Indore (IIT Indore), India Abhishek Srivastava - Indian Institute of Technology, Indore (IIT Indore), India
