In recent years Internet of Things (IoT) and Artificial Intelligence (AI) technologies have been boosting the competitiveness of organizations in sectors like manufacturing, energy, healthcare, and smart cities. IoT technologies enable organizations to interact with the physical world through cyber-physical systems and internet-connected objects, improving the automation and efficiency of their business processes. At the same time, AI technologies like machine learning and industrial robots facilitate organizations in deriving insights from large volumes of structured and unstructured datasets, helping to optimize workflows and improve decision quality. Recent advances in IoT and AI systems have further streamlined the collection and processing of information from diverse distributed data sources, including sensors, automation devices, smart objects, and enterprise databases. These advances also ensure secure and scalable information management while enabling advanced analytics through high-performance AI techniques like deep learning.

Moreover, the emergence of Large Language Models (LLMs) has further transformed the landscape by enabling new capabilities for IoT applications. LLMs can process and analyze unstructured data at scale, generate human-like responses, and facilitate natural language interactions between humans and machines. When deployed at the edge within IoT ecosystems, LLMs bring opportunities for real-time decision-making while reducing latency and bandwidth requirements. However, these advancements also introduce challenges related to explainability, trustworthiness, and resource efficiency in highly decentralized environments. Addressing these challenges is critical for building human-centered IoT/AI systems that align with Industry 5.0 principles.

While these developments provide a solid foundation for Industry 4.0 applications, limitations remain in supporting the new wave of human-centered AI applications envisioned for Industry 5.0. The development of such applications requires an additional layer of trustworthiness to enhance security, safety, transparency, and human acceptability. This necessitates research into methods that promote trustworthiness across multiple levels—ranging from ensuring data integrity to explaining AI system operations to humans—and fostering trusted human-machine collaboration. These goals can benefit from recent advancements in cloud/edge/IoT continuum technologies (e.g., edge AI and edge LLM approaches that reduce the attack surface of industrial data), blockchain-based anti-tampering mechanisms, trusted human-centric AI paradigms like explainable AI (XAI), neurosymbolic learning, and LLM-based IoT applications such as IoT and digital twin applications for product design.

A key challenge lies in developing trusted applications in highly decentralized IoT/AI environments that orchestrate data sources and data-driven services from multiple providers, including AI/IoT agents. This requires federated architectures that enable secure information sharing across diverse platforms with varying security mechanisms. Moreover, novel methodologies are needed to design IoT/AI systems that adapt to human interaction loops rather than expecting humans to adapt to system operations.

Submission Guidelines

Prospective authors are invited to submit original technical papers reporting original research of theoretical or applied nature for presentation at the workshop and publication in the TI-2025 Proceedings. All papers will be reviewed and evaluated by independent experts and selected based on their originality, merit, and relevance to the workshop. Accepted papers will be published as part of the IEEE DCOSS-IoT 2025 conference proceedings and submitted to IEEE Xplore.

The authors of accepted papers will be required to prepare a presentation in PDF file format and provide it along with the camera-ready manuscript. All presentations will be made publicly available.

The manuscripts must be prepared in English, following the IEEE two-column Manuscript Templates for Conference Proceedings (available here) with a maximum length of eight (8) printed pages including text, figures, and references. Authors may add at most two (2) pages, but only for an appendix, i.e. these two pages contain supplementary material only. The additional two pages will incur overlength charges at $100/page.

List of Topics

In this context, the aim of the 2nd Workshop on Next Generation IoT and AI Systems for Trusted, Human-Centered Intelligence (TI-2025) is to present research results on technologies, tools, and methods that support the development, deployment, and operation of trustworthy and human-centered IoT/AI systems across industries such as manufacturing, smart cities, precision farming, energy, and healthcare. The workshop particularly welcomes contributions addressing IoT applications based on Generative AI and LLMs (including edge-deployed LLMs) and related explainability challenges.

The main topics of interest for this workshop includes:

• Architectures, tools, and techniques for trusted and reliable industrial data.
• Federated architectures and data spaces for trustworthy data sharing.
• Distributed data management for trusted AI/IoT applications in the cloud/edge continuum.
• Decentralized machine learning techniques for trusted and privacy-friendly IoT/AI applications (e.g., Federated Learning, Swarm Intelligence).
• Security, safety, and data protection in highly distributed IoT/AI systems.
• Distributed Ledger Technologies (DLTs) for trusted human-machine interactions.
• Decentralized AI paradigms fostering Human-AI collaboration (e.g., active learning, neurosymbolic learning).
• Novel techniques for Explainable Artificial Intelligence (XAI) and Interpretable Artificial Intelligence (IAI), including explainability challenges specific to GenAI-based and LLM-based IoT applications.
• Methods for deploying LLMs at the edge within IoT environments while addressing resource constraints.
• Human-centered enhancements to industrial IoT/AI technologies (e.g., digital twins integrated with LLMs).
• Trustworthy IoT/AI applications leveraging Generative AI and LLMs in sectors such as manufacturing, smart cities, precision farming, energy management, and healthcare.
• Techniques for aligning LLM-powered IoT systems with human interaction loops.
• Resource-efficient deployment strategies for LLMs and Generative AI systems within decentralized IoT ecosystems.

This workshop seeks to bring together researchers from academia and industry to discuss cutting-edge innovations that address these challenges while advancing trustworthy IoT/AI systems tailored to Industry 5.0 principles. With respect to its previous edition, the workshop puts emphasis on Generative AI and LLM-based IoT applications (including relevant edge deployments) alongside their explainability challenges and trustworthiness challenges.

Committees

Program Committee

• Luís Lino Ferreira:Instituto Superior de Engenharia do Porto, Portugal
• Martin Serrano:University of Galway, Ireland
• Salviano Pinto Soares:Universidade de Trás-os-Montes e Alto Douro, Portugal
• Filipe Moutinho:NOVA School of Science and Technology, Portugal
• João Rosas:NOVA School of Science and Technology, Portugal
• Pavlos Kranas:LeanXcale, Spain

Organizing committee

• Pedro Maló:NOVA School of Science and Technology, Portugal
• John Soldatos:Netcompany-Intrasoft, Luxemburg
• Tiago Teixeira:UNPARALLEL Research, Portugal

Contact

All questions about submissions should be emailed to Pedro Maló <pmm@fct.unl.pt>