Abstract 
The growing integration of artificial intelligence (AI) into measurement devices has accelerated the  development of intelligent sensors capable of performing complex tasks directly at the system edge. These  devices often operate under strict constraints, such as limited computational capacity, reduced memory  availability, and demanding energy requirements, which makes designing efficient solutions a significant  technical challenge. 

Beyond local processing capabilities, interoperability has become a critical factor for the adoption of  intelligent sensors in industrial, laboratory, and research environments. The use of standardized infrastructures, such as those provided by the IEEE 1451 family of standards, enables heterogeneous  sensors to communicate uniformly, supporting advanced semantic interoperability through ontologies and  data dictionaries. 

This Hands‑on Lab is structured into two complementary phases. In the first phase, participants will  develop an intelligent sensor capable of executing AI models locally, exploring optimization techniques  suitable for resource‑constrained devices. In the second phase, the sensor will be integrated into an IEEE  1451 infrastructure, demonstrating how standardization enablessemantic interoperability and integration  into larger systems. 

The ultimate goal is to provide a complete, practical experience: from building an intelligent sensor to  integrating it into an interoperable, semantically enriched ecosystem. 
 

Format of the Tutorial

Hands-on Lab

Length of the Tutorial:

3 Hours

  • António Espírito-Santo.jpg

    University of Beira Interior, Portugal

  • Headshot_Helbert.png

    University of Beira Interior, Portugal

This tutorial addresses emerging challenges at the intersection of embedded artificial intelligence,  intelligent sensing, and standardized interoperability. As artificial intelligence (AI) increasingly migrates from the cloud to the edge, sensors are expected not only to acquire data but also to process, interpret, and communicate it autonomously. Achieving this shift requires addressing several technical constraints that remain central to current research and industrial deployment. 

A first challenge concerns the execution of AI models on devices with severe resource limitations.  Intelligent sensors typically operate with reduced computational power, limited memory, and strict energy budgets. Designing and deploying machine‑learning models that remain functional under these constraints demands specialized optimization techniques, lightweight architectures, and careful hardware/software co‑design. These issues are particularly relevant today, as edge‑AI adoption accelerates across domains such as industrial monitoring, environmental sensing, and smart infrastructures. 

A second challenge relates to interoperability. As sensing systems grow in scale and heterogeneity, the ability of devices to integrate seamlessly into larger infrastructures becomes essential. The IEEE 1451  family of standards provides a structured framework for metadata representation, communication, and enabling device self‑description. However, effectively leveraging this framework, especially when integrating AI‑enabled sensors, requires a clear understanding of Transducer Electronics Data Sheet (TEDS) structures, communication models, and semantic extensions. The tutorial addresses how standardized metadata, ontologies, and dictionaries can support semantic interoperability, enabling systems to interpret sensor data consistently and automatically. 

The timeliness of this tutorial stems from the convergence of three trends: the rapid expansion of edge AI,  the growing demand for interoperable sensing ecosystems, and the increasing relevance of IEEE 1451 in both academic and industrial contexts. By guiding participants through the development of an AI‑enabled sensor and its integration into a standardized infrastructure, the tutorial provides practical insights into technologies that are becoming foundational for next‑generation cyber‑physical systems.