Global Internet of Things and Edge Computing Summit

As distributed energy resources (DERs) such as solar PV, batteries and electric vehicles become increasingly prevalent at the edge, maintaining grid stability requires advanced monitoring and control mechanisms. This paper presents a scalable smart grid gateway architecture that enables interoperability between Modbus-based inverters and IEEE 2030.5 cloud-based control systems. The proposed solution leverages Azure cloud services and gateway devices located at the edge, to support dynamic configuration, telemetry ingestion, remote control and Volt-VAR Curve deployment. A microservice architecture ensures flexibility and scalability across diverse deployment scenarios, including both gateway-mediated and direct-to-cloud device communication. Results demonstrate the successful mapping of a Fronius Primo inverter’s Modbus registers to IEEE 2030.5-compliant telemetry and control functions. Additionally, we evaluate real-time VVC updates and their impact on local voltage regulation, showcasing dynamic cloud-to-edge control with minimal latency. This work highlights the potential of virtualised, standards-based control infrastructures to support DER integration and active grid participation, while remaining adaptable to evolving smart grid architectures.

This paper presents a comprehensive comparison between two leading Low Power Wide Area Network (LPWAN) technologies: LoRa and mioty. While both are designed for energy-efficient, long-range communication in IoT applications, they differ significantly in their modulation schemes, spectral efficiency, robustness, and network capacity. The study combines theoretical analysis with extensive laboratory measurements to evaluate critical performance metrics such as on-air time, packet error rate (PER), sensitivity, and achievable network throughput. LoRa’s performance is analyzed across various spreading factors, highlighting its limitations under high-load conditions in dense deployments. Energy consumption per packet and estimated device lifetime under typical usage scenarios are also assessed. The results show that mioty operates closer to theoretical performance limits and offers substantial advantages for large-scale IoT deployments. These findings provide practical insights for network designers selecting between LoRa and mioty, particularly in dense or capacity-critical environments.

Parkinson’s Disease (PD), a progressive neurodegenerative condition, sometimes eludes early diagnosis owing to its subtle and diverse motor symptoms. Conventional diagnostic methods are expensive, intrusive, and often unattainable. This paper presents the NeuroKey Prediction Tool, a lightweight, browser-based AI system that utilises passive keystroke dynamics gathered during regular typing to detect and track Parkinson's disease. Utilising a modular Streamlit framework, NeuroKey extracts statistical and temporal biomarkers, including key hold time and inter-key intervals, employing machine learning and deep learning models for classification and regression applications. The Random Forest classifier attained an accuracy of 88% in Parkinson's Disease detection, surpassing XGBoost (65%) and Logistic Regression (47%). In terms of predictive performance for motor severity estimation (UPDRS/nQi scores), Ridge Regression (R² = 0.75), LSTM (0.74), and Random Forest Regressor (0.69) exhibited robust results. These findings validate the feasibility of digital phenotyping using typing behavior as an effective, non-invasive biomarker for neuromotor evaluation. NeuroKey’s lightweight, scalable, and privacy-conscious architecture facilitates home-based ambient assisted living and telemedicine processes, thereby improving proactive, patient-centered treatment. Future endeavors will enhance functionality by utilizing varied, real-world typing data, incorporating personalized baselines, and investigating federated learning to protect privacy while augmenting performance. NeuroKey showcases the integration of AI and IoT for accessible, continuous, and precise neurological healthcare.

Federated Learning (FL) enables collaborative model training without centralizing raw data, but distributed deployments remain exposed to typical poisoning and inference attacks and must operate across resource-constrained edge environments. The REMINDER project addresses these challenges by designing an edge-centric framework that provides privacy and security mechanisms with byzantine-robust learning approaches. This paper reports some of the project’s mechanisms and their implications for the development of robust and secure FL deployments, including: (i) a threat model addressing poisoning and inference risks; (ii) a modular architecture with differential privacy, secure authenticated updates, and robust aggregation against malicious clients; and (iii) two representative validation scenarios, such as eHealth and smart buildings, which ground design choices and highlight domain-specific constraints. Building on these contributions, the present work formalizes the end-to-end workflow, specifies component interfaces, and links attack classes to concrete mitigations within REMINDER, while outlining open challenges such as verifiable aggregation and the privacy–utility trade-off introduced by differential privacy in common FL settings.

The rapid adoption of AI technologies is outpacing our ability to assess their long-term societal and economic impacts. Initially, AI was expected to automate only repetitive, low-skill tasks. However, presently, highly skilled roles such as in software development, manufacturing, and finance are being automated. Tasks that once required many professionals can now be managed by a few, disproportionately benefiting those with greater resources and economic power, such as large corporations. This trend may lead to a growing socio-technological imbalance, for instance, a growing mismatch between the rapid advancements in artificial intelligence and society’s ability to adapt to and govern these changes in a fair, ethical, and inclusive manner. As AI-driven automation is increasingly being adopted across almost all business domains, covering all process-level functions. Additionally, there is a lack of practical methods for the ethical and responsible adoption of AI, which are either not being implemented or not well understood by stakeholders in the business ecosystem. This emphasizes the role of Responsible AI (RAI), which is becoming critical and essential in addressing the socio-technological imbalance in the organizational context regarding adopting/adapting AI. RAI ensures that systems are developed and deployed with core principles such as explainability, fairness, ethics, transparency, accountability, human oversight, and privacy. Furthermore, AI itself presents as a technological capability to promote responsible behavior by assessing, predicting, supporting, and regulating its own societal (i.e., organization-specific) and technological impacts. This study explores this dual perspective by reviewing the literature to propose a RAI Adoption/Adaptation (RAA) framework, focusing on the business process as the primary unit of AI intervention. A conceptual system implementation context is proposed to illustrate RAA enablement at the process level and highlight the need for an RAI agentic architecture capable of measuring, analyzing, and forecasting AI’s impact on business processes to support ethical, balanced innovation.

The current data landscape presents several challenges in data governance and integration. This paper presents the integration of the Data Fabric implemented in the Horizon Europe project aerOS, which integrates and unifies data available in the IoT-Edge-Cloud continuum, with the Data Space implemented in the RE4DY project in order to extend the capabilities of the Data Fabric by enabling the definition fine-grained data access policies and data monetization. This integration would promote collaboration with other participants in the Data Space based on the reuse of data assets from the Data Fabric by sharing them in a trusted environment and the integration of new data or services with the Data Fabric, thus contributing to the creation and integration of new data ecosystems.

This paper describes a framework for web-based monitoring, orchestration and simulation for industrial settings such as highly automated factories and warehouses. HTML5 is used for a 2.5D visualization with local prediction for smooth animation and 3D models for robot arms. The web server aggregates data from the devices for streaming to the web page, for generating situation reports, and enabling users to intervene as needed. Simulation is possible using simple looping sequences of actions. A cognitive architecture is described for richer behaviour that dynamically adapts to the context, decoupling reasoning from real-time control using asynchronous intents. Stochastic rules allow agents to escape from futile patterns of behaviour. Agents can control multiple devices. For a larger numbers of devices, control can be distributed across multiple agents in a way that decouples applications from the underlying protocols and addressing schemes. Iterative refinement is applied as new requirements come to light, e.g. to handle faults, or where humans have intervened in unexpected ways. The paper closes with a short summary of previous work and suggestions for future work on integrating generative AI to further simplify development and provide supervisory control.

Open Science is increasingly central to European research, enabling cross-sectoral access to and reuse of scientific data. To support secure, FAIR-aligned data sharing, the European Commission has launched flagship initiatives such as the European Open Science Cloud (EOSC) and the Data Spaces framework. This paper aims to discuss EOSC and Data Spaces, two of the main European initiatives designed to achieve this goal. This study will be carried out by analyzing the different strategies to achieve such a secure and cross-domain environment, as well as how these strategies can be used to contribute to the development of open science in Europe. Within this scope, the EU-funded TITAN-EOSC project will be discussed, describing its technical architecture, main objectives, and relating the key points to EOSC and Data Spaces. The main synergies between the project presented and the initiatives discussed will also be considered, highlighting how they concur with each other.

This paper introduces a stateless orchestration architecture and mechanism for managing distributed data pipelines in multi-agent systems. Workflows are decomposed into independent subworkflows by a splitting module. A caching mechanism preserves execution context without relying on local state, supporting true statelessness. Communication is event-driven via Apache Kafka, and a unified data model ensures consistent interaction among components. The architecture is validated on a Kubernetes environment, showing scaling and low-latency features. This approach combines the control of orchestration with the scalability of event-driven systems, offering a robust solution for modular workflow execution.

Respiratory disorders continue to pose a significant global health concern, necessitating prompt, precise, and scalable solutions for diagnosis and monitoring. Traditional respiratory evaluations frequently rely on face-to-face examinations, restricting accessibility and the continuity of care. This work tackles the research issue of attaining continuous, distant, and precise prediction of pulmonary conditions through the use of multimodal AI combined with IoT-based e-Health systems. The main goal was to create and evaluate the Respiratory Analysis and Prediction Tool (RAPT), a system that integrates audio data and patient metadata to categorise respiratory disorders. Employing a dual-branch convolutional neural network trained on mel spectrograms and standardised clinical data (age, sex, BMI), RAPT attained a validation accuracy of 77.08% on a subset of the Respiratory Sound Database. The Gradio-powered interface facilitates real-time inference, and the tool is engineered for compatibility with IoT-enabled smart stethoscopes and wearable devices. Principal findings underscore robust efficacy in classifying COPD (precision 0.81, recall 1.00), however class imbalance constrained accuracy for less prevalent conditions such as Bronchiectasis. The paper finds that RAPT shows potential viability for personalised and remote respiratory monitoring but necessitates enhancements via bigger balanced datasets, calibration of wearable sensors, secure edge-based processing, and interoperability with 5G and Zigbee networks. This research establishes a basis for enhancing AI-driven, IoT-integrated e-Health platforms to transform respiratory health monitoring, facilitating continuous, accessible, and scalable patient care.

The study incorporates the concept of a digital product passport as a data carrier for tracking the end-of-life of wind turbine blades across the processes value chain, integrating circular economy principles into the business model. Evaluation of a novel chemical recycling process for wind turbine blades using Life Cycle Assessment (LCA) and Life Cycle Cost Analysis (LCCA) suggests that the chemical recycling process has significantly reduces environmental impacts when compared to landfill disposal and offers potential economic benefits. However, scalability and material recovery efficiency remain key factors for widespread adoption.

Organisations preparing for the European Sustainability Reporting Standards (ESRS) mandate often lack a clear, structured approach to assess the maturity of their Environmental, Social, and Governance (ESG) capabilities.

This study develops an ESG-specific Capability Maturity Model (ESG-CMM) through a six-step Design Science Research process. A systematic review distilled 104 descriptors from 15 capability maturity model articles. Open coding, cross-checked against ESRS topics, condensed these into 13 parameters arranged within five transversal clusters and five classic CMM levels.

The artefact was demonstrated in a mid-sized manufacturing firm via seven semi-structured interviews. Evidence-backed ratings placed the organisation at Level 2.0 (“Repeatable”), highlighting strengths in governance policy and gaps in analytics capability and integrated risk management.

The ESG-CMM artifact thus bridges a documented academic gap by translating a software-derived maturity logic to ESG practice, offering managers a self-assessment tool that is both regulator aligned and implementation oriented. While the single-case design limits generalisability, the research sets a foundation for multi-site validation and for sensor-driven, real-time scoring extensions.

This paper presents a novel approach to biological resource traceability and certification through an IoT-ready platform based on NGSI-LD data models. The BioReCer ICT Tool (BIT) addresses significant challenges in sustainable supply chain management by providing an interoperable, standardized framework for data collection, visualization, and verification. Our implementation leverages the Stellio Context Broker together with Apache NiFi for Internet of Things (IoT) data integration, creating a powerful platform for sustainability assessment and certification. Both the data model and the user interface have been co-developed and validated with stakeholders across four distinct case studies, demonstrating its flexibility and effectiveness across diverse bio-resource value chains. Results indicate significant improvements in data interoperability, traceability, and verification capabilities compared to traditional approaches.

With the rapid growth of hyperconnected devices and decentralized data architectures, safeguarding Internet of Things (IoT) transactions is becoming increasingly challenging. Blockchain presents a promising solution, yet its effectiveness depends on the underlying consensus algorithm. Conventional mechanisms, such as Proof of Work and Proof of Stake, are often impractical for resource-constrained IoT environments. To address these limitations, this work introduces a fair and lightweight hybrid consensus algorithm tailored for IoT. The proposed approach minimizes resource demands on the nodes while providing a fair and secure agreement process. Specifically, it utilizes a distributed lottery mechanism to ensure fair block proposals without requiring dedicated hardware. In addition, to enhance trust and establish finality, a reputation-based voting mechanism is incorporated. Finally, we experimentally validated the key features of the proposed consensus algorithm.