Tutorial Presentations

Like any science and engineering field, Instrumentation and Measurement (I&M) is currently experiencing the impact of the recent rise of Artificial Intelligence (AI) as an applied tool. In fact, there is an intertwined relationship between the two: measurement is used to collect data, which are then used to train AI models, which in turn are used for indirect measurement. The applications are vast: medical diagnosis, surveillance, fault detection, condition monitoring, etc. At the same time, uncertainty, which is a fundamental component of measurement, must be quantified for risk management and better decision making in these AI-assisted measurement systems.

In this talk, we will learn about uncertainty from the perspective of both AI literature and measurement standards, such as VIM and GUM. We then show how AI is used for indirect measurement and how to quantify the uncertainty of AI-assisted measurements to design more reliable and practical measurement systems. We will cover AI regression, classification, and Large Language Models (LLM). Additionally, we will provide a unified taxonomy of uncertainty quantification methods, and point to some measurement-standard noncompliance in AI literature, which readers need to be aware of. Finally, we go over a few specific examples from existing literature.

Photoacoustic (PA) sensing, a non-invasive hybrid technique, provides high-contrast, label-free information for biological tissues and industrial materials by combining optical contrast with acoustic depth. This talk focuses on its transformation from complex lab instruments into compact, field-deployable handheld systems. We will discuss PA sensing's potential for predictive machine maintenance, paint manufacturing, and coal mining safety. Since PA generates vast, complex data, the session emphasizes integrating AI tools to enhance sensitivity, speed, specificity, and to solve challenging inverse problems in real-time. Finally, we outline future directions for robust, clinically and industrially translatable AI-enabled PA systems.

Electromagnetic (EM)/ radio-frequency (RF) sensing enables unobtrusive, non-contact, and privacy-preserving monitoring. While AI enhances data extraction, its success hinges on high-fidelity instrumentation and data quality. This tutorial emphasizes the overlooked necessity of reliable system-level implementation to prevent the failure of blind AI application. We provide a comprehensive overview, showcasing research in biomedical monitoring (radar-based vital signs, implant assessment, oedema detection) and industrial safety (subsurface imaging, fall-of-ground prediction in mines). The session provides practical insights into the intersection of EM/RF systems and AI for addressing real-world challenges.

The use of Artificial Intelligence techniques to process medical images is receiving increasing attention from academics and pathologists, due to the high potential for detecting diseases in a fast and accurate way. Deep Learning (DL) is especially emerging as a standard paradigm to classify Whole Slide Images (WSI) and thus differentiate between tissue types and possible diseases. In this field, the use of DL-based methods such as Convolutional Neural Networks and Vision Transformers enables high-accuracy classification of the images; however, drawbacks such as the limited dimensionality of the databases and the extremely high size of microscope-scanned WSIs limit a straightforward adoption of such methods. This tutorial will describe how to use DL for achieving an effective processing of WSI, leveraging the capabilities of CNNs to detect diseases with high accuracy: in particular, the tutorial will use as a case study the detection of Acute Lymphoblastic Leukemia, a disease that affects blood cells and can be fatal if not treated in a timely manner. Hands-on activities will show how to deploy state-of-the-art approaches for Leukemia detection, by downloading pre-trained models for histopathological processing and tuning them to differentiate between healthy and cancerous blood cells.

There has been an astronomical increase in the number of technical paper submissions in the past decade.  Some of the reasons include: 

• pressure to publish, as the success indicator, for promotion and professional advancement,
• universities moving away from the traditional Theses and Dissertations compilations and instead use peer-reviewed journal papers,
• creation of new journals, and
• the open-access publishing “economy”.

Effective journal paper writing requires consideration of many implicit and explicit issues. This involves deciding on the proper, most effective and relevant publications venue; having the knowledge of (and experience) how to compile the content that goes into the paper; having a clear knowledge of the publications’ guidelines and rules for the chosen venue; etc. 

The review process is the most crucial aspect of a proper publishing process.  It involves the authors, the editors, and the reviewers, and how each performs detrimentally impacts the outcome and the “Quality”.

Inclusion of AI-assisted and AI-generated content has become an important issue. To this end, IEEE has specific guidelines to which authors and reviewers must adhere and ignoring them could have significant implications for both.

The acquisition of information about physical quantities through sensors has historically encouraged the interpretation of measurement as a purely experimental task. In contrast, measurement should be recognized as a fundamentally model-based and conceptually rich endeavor, often involving tasks far more complex than merely connecting an instrument and reading its output.

Before any empirical procedure is meaningfully executed, measurements demand a set of descriptive and conceptual efforts to ensure both the validity of the performed actions and a meaningful interpretation of the obtained information.

This tutorial outlines the foundational principles of measurement, highlighting their universality across disciplines and their growing relevance in the rapidly evolving landscape of artificial intelligence.

Rail systems are vital infrastructures in urban mobility and intelligent transportation, where effective health monitoring prevents unexpected downtime and ensures passenger safety. Defects detection and classification from the monitoring also provide substantial economic benefits by supporting optimized operations and management. Current methods mainly rely on vibration measurement or visual inspection, yet both face limitations, with visual techniques particularly vulnerable to lighting and weather conditions. This tutorial introduces two emerging non-contact measurement approaches for quantitative rail condition assessment: inductive coupling and electromagnetic acoustic emission (EMAE). The inductive coupling method employs probe setups to inject radio-frequency signals into the system under test and capture responses, offering a reliable means to evaluate the structural properties of vehicle components or track elements. EMAE applies electromagnetic induction to generate external excitation and detect transient elastic waves within the material, offering high sensitivity, wide frequency bandwidth, and suitability for harsh or high-speed environments. The tutorial will outline the measurement principles, sensor configurations, and data acquisition schemes of these methods, followed by case demonstrations and practical considerations. Future perspectives on integrating these techniques into intelligent measurement systems will also be discussed to advance precise and scalable solutions for railway applications.

• Course 1: Condition Monitoring of Urban Rail System Based on Inductive Coupling Setup
• Course 2: Electromagnetic Acoustic Emission Technique for Rail System Fault Diagnosis

The emergence of new sensors and robust data analysis promises a sea change in the dynamic understanding of the electric distribution grid. During its operation, the distribution grid produces a wide range of signals, which provide invaluable information and indications about its state of health, and by proper analysis can provide indications on its reliability, different “conventional” faults, and less conventional faults, such as High-Impedance Faults. Examples are indications of vegetation interactions that could initiate wildfires, or insulation breakdown. Timely  detection, identification, and response, can prevent costly failures. However, since these signals are short and contain high frequencies, they are often missed by the conventional sensors at the substations.

Advances in robust electronics and computing are making it now possible to measure and react to variations in the distribution grid. Commercial instrumentation addresses part of the issue and research into sensor design, operation and signal analysis are promising unprecedented insight into the grid status. This tutorial addresses the potential opportunity, the design opportunities and constraints, experience in extracting novel information from operating distribution grids, and the breadth and depth of the signal processing research available to augment understanding.

This tutorial is designed to introduce participants to fundamental compensation methods used in impedance spectroscopy. Attendees will gain both theoretical knowledge and practical experience.

The theoretical component will be presented in a lecture by Konstantin Weise. This session will provide a basic understanding of impedance spectroscopy, including the mathematical foundations of common compensation methods.

In the second part of the tutorial, the theory will be put into practice during a hands-on lab session conducted in collaboration with Sciospec Scientific Instruments GmbH (Bennewitz 04828, Germany). Participants will collect their own measurement data using the provided equipment and independently explore various compensation methods.

Space is limited to 30 participants. Ensure to add this to your registration if planning to attend.

Magnetic localization has emerged as a highly valuable technique, particularly for biomedical applications where line-of-sight constraints make common imaging impractical. Magnetic fields are not ionizing and unlike optical or acoustic approaches, they can penetrate biological tissues without significant attenuation, enabling accurate tracking of objects or devices inside the human body. A magnetic localization system typically consists of magnetic field generators and magnetic sensors that detect the resulting field distribution. Various magnetic sensing technologies—such as Hall effect sensors, magnetoresistive sensors, fluxgate, or SQUID—offer different trade-offs in terms of sensitivity, bandwidth, cost, and integration possibilities. Selecting the most suitable technology is therefore essential to ensure optimal system performance for a given application. The localization process itself relies on solving an inverse problem, which involves estimating the position and orientation of an object based on sensor measurements. While the algorithms can be straightforward on ideal data, they are highly sensitive to noise, which can significantly degrade performance. Choosing an appropriate algorithm—analytical, numerical, or data-driven—is therefore essential to ensure robustness and accuracy.

• Part 1: Hardware
• Part 2: Software

The reliable operation of autonomous systems in dynamic, real-world environments — from industrial logistics to surveillance — builds upon the accuracy and integrity of measurement acquisition. This tutorial establishes the central role of measurement science in addressing the critical challenges of localization, mapping, and state estimation. We begin by defining the foundational concepts of system state and motion, positioning system observability as the primary metric for validating state estimation feasibility. Our core focus is the practical metrological analysis of sensors, examining operational principles through rigorous uncertainty modeling. Emphasis is placed on characterizing how systematic and random measurement errors propagate through raw sensor data, directly influencing the performance of state estimation algorithms. Furthermore, we explore advanced instrumentation topics, including sensor fusion methodologies, radio-frequency ranging systems, and robust techniques for quantifying measurement uncertainty. The goal is to equip attendees with the theoretical and practical expertise to select appropriate instrumentation, characterize sensor performance, design measurement-driven estimation pipelines, and validate the overall metrological reliability of autonomous systems for autonomous-systems applications.

Device-free Wi-Fi sensing has gained much attention due to its simplicity, low cost, and no requirement for additional hardware sensors. The main advantages of Wi-Fi sensing are that it is unobtrusive, can operate through walls, work without lighting, is ubiquitous, and does not require users to carry any additional wearable devices. CISCO estimates that there will be 543 million Wi-Fi hotspots in the world by the end of 2022, which makes the Wi-Fi signal availability almost omnipresent. The traditional methods of video and sensor-based systems suffer from many shortcomings, like acceptability, availability, affordability, and, moreover, privacy concerns. The received Wi-Fi signal characteristics change with a change in the dielectric constant of the medium, and other reflections and scattering. These changes in the Wi-Fi signal patterns can be exploited to detect various events, environmental conditions in the wireless zone, identify the materials in the wall, and also detect faults. Recently, Wi-Fi sensing techniques have also been used for measuring physiological parameters like heartbeat, breathing rate, and monitoring a few other things.

In modern measurement sciences, compound-oriented measurement technology has emerged as a specialized domain within measurement sciences that requires a more comprehensive and systematic approach than standard measurements. These methodologies typically involve multiple stages such as sampling, sample preparation, transpiration, reformatting, measurement, and data processing. Given this complexity, thorough validation of all subprocesses within the overall workflow is essential to ensure accurate and reliable results. 

Unlike most physical measurements, compound-oriented measurements not only quantify but also qualify the measurand of interest. Frequently, the analyte must be identified within complex multi-component mixtures embedded in biological, chemical, medical, or environmental matrices. Robotic systems can significantly enhance such workflows by increasing throughput, improving accuracy, and ensuring operator safety.