Pre-Conference Tutorials

Tomographic imaging based e-skin sensors offer a unique  solutions for  sensing for many applications. This includes artificial skins in robotics, medical application for interfacing with human skin.  Additionally this type of e-skin will have a key role in ever growing and emerging field of  flexible and  soft robots.  

Sensor transform processes and converts raw data across different sensor modalities restructuring inputs into meaningful representations to enhance machine perception and decision-making. From remote sensing to autonomous systems, sensor transform has the potential to break down barriers between different sensor modalities, reshaping how we design and integrate intelligent systems for a more seamless and efficient future.

The illustrious history of terahertz (THz) imaging and sensing is nearly 50 years long. During this time, photonic and electronic THz technology has developed a lot, but it has not been able yet to bridge the famous THz gap between electronic and photonic devices. In the THz range of frequencies, a low photon energy (smaller than the room temperature thermal energy) makes the development of efficient THz lasers to be a challenge. And the cutoff frequency and maximum frequency of operation of field effect and bipolar transistors struggles to reach one THz. 

Field-effect transistor (FET)-based sensors offer high sensitivity, rapid detection and compatibility with large-scale semiconductor fabrication, making them ideal for point-of-care diagnostics. Recent advances leverage one- and two-dimensional materials with high surface area-to-volume ratios and excellent charge carrier mobility, greatly enhancing sensing performance. Notable examples include MoS₂, metal dichalcogenides, graphene and MXenes, which have demonstrated superior sensitivity. These nanostructured materials are driving the next generation of electrochemical FET sensors with improved efficiency and scalability. 

Biosensors can be broadly classified into different types based on the method used for signal transduction, including electrochemical, optical, thermal, piezoelectric and magnetic biosensors. Due to a vast area of possibilities, in this tutorial, I will focus my attention on electrochemical biosensors and I will discuss the underlying physical and electrical principles of operations and their areas of applications. 

Foundation models are rapidly transforming the analysis of wearable sensor data by enabling robust, generalizable learning from massive unlabeled datasets. Originally developed for language and vision, these models are now being adapted to time-series data from wearable devices—offering new capabilities in health monitoring, activity recognition, and digital phenotyping.

MEMS micromirrors have been proposed and adopted in many different laser-beam-scanning (LBS) applications , starting from barcode reading, and ranging to both medical, consumer and automotive applications, such as miniaturized projection, augmented-reality (AR) headsets and LiDAR/3D sensing. The main architecture adopted in these fields consists in two-dimensional scanning of a laser source using either a bi-axial device or a couple of mono-axial ones.

Contemporary gas-monitoring demands push existing gas sensor designs to their fundamental limits in “3S” requirements of Sensitivity, Selectivity, and Stability. The origin of these limits is in the single-output (e.g., resistance, light intensity) sensor designs, also known as zero-order sensors. Any zero-order sensor is affected by chemical background and sensor drift that cannot be distinguished from the response to an analyte. 

Despite outstanding performance in laboratory environments, only a limited number of novel biosensor concepts transition into practical point-of-care (POC) or on-site detection devices. Key obstacles include limited robustness, cost constraints, and biofouling in complex biological samples. A major unmet challenge lies in engineering antifouling biofunctional surfaces that simultaneously support high biorecognition efficiency, resist biofouling, and allow scalable, reproducible sensing performance. 

This introductory tutorial addresses problems of storing and/or multilinear processing of very large multidimensional  data  arrays  (tensors)  that could possibly result from multi-sensor measurements. After  introducing  basic  tensor  operations,  we  will  cover  both  theoretical and computational aspects of two classic tensor decompositions: Canonical Polyadic Decomposition (CPD) and Tucker Decomposition (TD). We will also introduce the concept of Tensor Networks, with particular emphasis on the relatively recent Tensor Train Decomposition (TTD).