A Glovebox is a sealed enclosure that allows handling materials while keeping them isolated from the external environment. It is widely used in scientific laboratories, industry, and various technical fields. Usually, a glovebox is a standard, monolithic structure purchased from a supplier, with little customization. This work presents a novel approach to the design of gloveboxes. It acts as a design methodology based on using off-the-shelf, lightweight yet robust structural parts, employing a modular architecture. Key advantages include the possibility of being assembled "on site" with simple tools (no welding required), with the possibility of integrating robotic manipulators. The approach enables the design of solutions to be adapted to existing laboratory environments without modifying the surrounding infrastructure. To illustrate the advantages, a representative albeit challenging use case has been identified: the repair of Carbon Fiber Reinforced Plastics (CFRP), where the Glovebox encloses and executes the scarfing (sanding) process. More broadly, the approach can be evaluated in any situation requiring controlled interaction between isolated internal environments and external laboratory or industrial settings. The CFRP Glovebox prototype has been built in our laboratory and is ready for use by expert operators under continuous supervision, and eventually in a manufacturing context. This work is licensed under CC BY-SA 4.0. (https://creativecommons.org/licenses/by-sa/4.0).

Bridges are critical infrastructure assets whose seismic performance is increasingly compromised by aging and corrosion of reinforced concrete elements. Existing risk assessment methods are often computationally demanding and neglect degradation effects, underestimating seismic vulnerability. This study presents a novel AI-based framework that leverages computer vision to quantify corrosion severity from inspection images and integrates time-dependent corrosion evolution laws to forecast the Mean Annual Frequency of Exceedance over the remaining service life. A custom convolutional neural network with attention mechanisms translates observed damage into probabilistic degradation parameters for steel and concrete, incorporated into nonlinear seismic analyses and fragility assessment. By considering the expected evolution of the observed corrosion state, the framework identifies the time at which reliability thresholds are exceeded. The approach provides a fast, cost-effective tool for estimating time-dependent seismic performance and residual service life of bridges characterized by corroded piers, supporting life-cycle–oriented maintenance planning and risk-informed decision-making.

This study presents a novel probabilistic methodology for the rapid seismic assessment of reinforced concrete bridge piers affected by end-corrosion. The approach integrates visual inspection supported by computer vision and stochastic structural analysis to capture the degradation effects caused by corrosion on seismic performance. Central to the procedure is an image-based classification system that leverages a customized convolutional neural network (CNN), designed with attention mechanisms and color space preprocessing, to automatically assess corrosion severity from photographs. The classified severity levels are then linked to a probabilistic model of material deterioration, enabling condition-informed adjustments to structural parameters such as steel and confined concrete strength and ductility. The assessment workflow includes geometric characterization, artificial intelligence-driven visual inspection, stochastic modeling of degraded materials, nonlinear analysis, seismic fragility evaluation and loss assessment. Application to a representative case study demonstrates significant impacts of corrosion on the seismic fragility, emphasizing the added value of integrating CV with probabilistic analysis for data-informed risk assessment of aging infrastructure. The results, in terms of expected annual losses, support the use of image-based methods as effective tools for prioritizing maintenance and optimizing resource allocation in bridge management systems.

Defect detection in Carbon Fiber Reinforced Polymer panels represents a crucial step in aircraft production processes, to improve security and reduce maintenance costs. One of the most common strategies for inspecting this material is the use of Ultrasonic Testing, which enables the non-destructive inspection of the internal structure of objects. However, the evaluation of the ultrasonic data currently relies on domain experts, making the process few scalable and potentially inaccurate in ambiguous situations. In this study, an automated method based on Artificial Intelligence is proposed to support human operators in detecting defects on CFRP panels. The proposed solution is characterized by training an Autoencoder to perform the anomaly detection task directly on raw A-scan. Two architectures have been implemented: a Vanilla Autoencoder, which acted as the baseline, and a Convolutional-1D Autoencoder. Results show that using 1-D convolutions yielded very satisfactory performance, achieving a top F1-score equal to 67.68% for a 27plies thick panel. Additionally, the relatively lightweight architectures and the use of raw data enabled us to achieve very low inference times, specifically in the order of milliseconds, making the proposed solution applicable to real industrial scenarios.

Autonomous navigation has gained increasing attention with the rise of self‐driving robots in applications ranging from logistics to space exploration. The robot's capabilities, context, and objectives significantly influence the design and selection of efficient navigation methods. However, traditional navigation methods, reliant on rule‐based algorithms and deterministic reasoning, struggle with adaptability and robustness in complex scenarios. Artificial Intelligence (AI) offers a solution, enhancing navigation through techniques like deep learning (DL), semantic understanding, and real‐time anomaly detection. This review explores AI's role in overcoming the limitations of classical approaches, focusing on adaptability, safety, and collaboration in navigation tasks. Analyzing diverse techniques and their integration with sensors highlights the potential of AI to enable reliable, efficient, and secure operation in real‐world environments, guiding future research in intelligent robotics.

Accurate fall detection is increasingly important for continuous monitoring in home and workplace environments. Recent advances have focused on vision-based systems and skeleton-based representations, exploiting Transformer-based architectures to model complex spatio-temporal relationships. However, the lack of transparency of these models limits trust and practical deployment. To address this challenge, this paper employs ShaTS, a variant of the SHAP technique, as a post-hoc, model-agnostic explainability method to interpret and understand the predictions of the Transformer-based model. ShaTS is used to quantify the global contributions of joints and temporal segments on the model’s predictions. The results show that explainability can reveal meaningful motion cues, support domain-level validation, and guide future model refinement, ultimately enhancing confidence in the model’s decision-making process.

Manual drilling of Carbon Fiber Reinforced Polymer (CFRP) is a pivotal aerospace process where tool wear directly compromises the productivity of aircraft manufacturing. To the best of our knowledge, no previous study has yet systematically analyzed acoustic features to identify the most influential ones for tool wear monitoring in manual drilling. This study proposes a data-driven framework for tool-wear monitoring using acoustic emissions captured during manual drilling. Using a real-world dataset labeled by operator perception of effort, we extracted time and frequency domain features to train and compare different machine learning algorithms, including Logistic Regression, SVM, Random Forest, and XGBoost. Results indicate that aggressive outlier removal degrades performance, implying that outliers hold critical information in data-scarce environments. While the baseline Logistic Regression achieved an F1 score of 69.45%, the optimized XGBoost model outperformed all architectures with an F1 score of 75.81%. Furthermore, experiments revealed that using signal duration as a sole predictor yielded a notable F1 score of 66.52%. This work demonstrates that acoustic monitoring and ensemble learning are effective tools for decision support in manual drilling.

Insects feeding on xylem sap, such as adult Aphrophoridae spittlebugs, are vectors of the plant pathogenic xylem-limited bacterium Xylella fastidiosa (Xf), a causal agent of a number of severe diseases, including the Olive Quick Decline Syndrome (OQDS), which has decimated olive trees in the Mediterranean region. The Aphrophoridae life cycle and behaviour feature a weak stage, known as the juvenile stage, in which the insects live solitary on stems covered in a self-produced foamy fluid (froth) that protects them from dehydration and temperature stress. Juvenile vectors are ideal targets for a control intervention aimed at reducing transmission by adults. This paper presents the first, to the best of our knowledge, image dataset framing spittlebug froth samples in the field for the purpose of automated Aphrophoridae nymph identification. Images were captured using different devices including a consumer-grade RGB-D sensor, a digital reflex camera, and a smartphone camera. The dataset comprises 365 colour images, focusing on spittlebug foam. 211 of these images were captured in April 2024 during a two-day campaign. For these 211 images, a manual semantic annotation was performed, generating PNG binary masks that precisely distinguish spittlebug foam pixels from the background. To further enhance usability, labels are also provided in YOLO (You Only Look Once) format as text files, both for segmentation and object detection. The remaining 154 images were collected during a separate two-day campaign in 2025. These images are unannotated and are intended for further testing purposes. Overall, the dataset enables the development of both semantic segmentation models and object detectors for automated froth detection in natural images, thus facilitating the early identification of potentially harmful insects in sustainable pest management and control systems.

Multimodal industrial anomaly detection benefits from integrating RGB appearance with 3D surface geometry, yet existing \emph{unsupervised} approaches commonly rely on memory banks, teacher-student architectures, or fragile fusion schemes, limiting robustness under noisy depth, weak texture, or missing modalities. This paper introduces \textbf{CMDR-IAD}, a lightweight and modality-flexible unsupervised framework for reliable anomaly detection in 2D+3D multimodal as well as single-modality (2D-only or 3D-only) settings. \textbf{CMDR-IAD} combines bidirectional 2D$\leftrightarrow$3D cross-modal mapping to model appearance-geometry consistency with dual-branch reconstruction that independently captures normal texture and geometric structure. A two-part fusion strategy integrates these cues: a reliability-gated mapping anomaly highlights spatially consistent texture-geometry discrepancies, while a confidence-weighted reconstruction anomaly adaptively balances appearance and geometric deviations, yielding stable and precise anomaly localization even in depth-sparse or low-texture regions. On the MVTec 3D-AD benchmark, CMDR-IAD achieves state-of-the-art performance while operating without memory banks, reaching 97.3\% image-level AUROC (I-AUROC), 99.6\% pixel-level AUROC (P-AUROC), and 97.6\% AUPRO. On a real-world polyurethane cutting dataset, the 3D-only variant attains 92.6\% I-AUROC and 92.5\% P-AUROC, demonstrating strong effectiveness under practical industrial conditions. These results highlight the framework's robustness, modality flexibility, and the effectiveness of the proposed fusion strategies for industrial visual inspection. Our source code is available at https://github.com/ECGAI-Research/CMDR-IAD/

The paper presents a framework to automatically identify crack patterns and the related features in existing reinforced concrete (RC) bridges. The challenge of this work is to define a tool for detecting the focused defect and highlighting the number and the orientation of cracks, allowing for correct interpretation and driving further evaluations on the residual life of the structure. The study is framed within the increasing interest in monitoring the structural health of existing bridges through automated tools, able to support engineers in the phase of visual assessment and interpretation of structural defects. When dealing with periodic inspection of large bridge portfolios, the support provided by automated tools can be fundamental for planning further strategies aimed at ensuring the structural safety and preventing future disasters. Given a stack of photos of a bridge structural element, an image stitching procedure is proposed to produce a near-complete image of the entire element. On the latter, a pipeline of deep-learning (DL) algorithms is employed to automatically detect and identify cracks (as a combination of object detection and segmentation algorithms). Finally, the proposed tool extracts cracks for counting and defines their orientation (i.e., vertical, horizontal, diagonal), in order to provide near-complete information about the crack pattern for the structural element. A full description of the methodology and the proposed algorithms is reported throughout the manuscript, showing the main pros and cons and assessing the effectiveness of the tool on a real-life case study.

Manufacturing applications increasingly integrate visually aided robotic systems. Such systems must rely on excellent kinematic parameter calibration and a hand - eye matrix estimation to perform according to standards. The latter is as precise as the camera pose estimation capability and the robotic forward kinematic precision. To enhance the overall system's precision, one must simultaneously act and improve the robot's kinematic parameters and hand - eye transformation due to mutual inference. This work exploits standard 2D camera systems to simultaneously estimate the kinematic parameters and the hand - eye transformation matrix through a method based on the Unscented Kalman Filter (UKF) and the parameters uncertainty transportation through the robot's kinematic. The method employs data gathered during the robot movements and camera readings and iteratively improves the system parameters' estimate. The method is applied to industrial mobile manipulators and tested on both synthetic data and real experiment data, showing a great improvement in the kinematic parameters estimation.

Composite structures are commonly used in complex applications such as automotive and aerospace due to their high strength-to-weight ratio. Although strictly supervised and inspected, they are often subject to dynamic events during their useful life that can cause invisible failures that extend and severely compromise their performance over time. Detecting these defects preventively and repairing them could avoid dramatic accidents. Here, we present a deep learning-based method for the non-destructive detection of defects in composite samples based on a laser ultrasonic system (LUT). Laser ultrasonic technology is a promising non-destructive testing (NDT) method for detecting inner defects in a non-contact way, as it does not require liquid coupling media. We investigated a composite laminate specimen containing six programmed defects as a test sample. We show that training deep learning-based models as autoencoders makes it possible to extract features that can be used to discern defective areas from non-defective ones in the US C-scan maps. The results demonstrate high detection accuracies (above 90% balanced accuracy and 75% F-1-score), indicating a promising and effective approach to NDT on composite materials.

This paper explores innovation in cognitive robotics to automate complex mission-critical processes. By autonomously generating its models of the surrounding environment, the system localizes its target, detects any obstacle, calculates cycle time optimization and collision-free path planning and makes decisions that intuitively refine its movements and actions. All this technical complexity is hidden behind a simple and effective user interface based on low-code programming. Comau and CNR have embedded AI-powered methodologies in an innovative model-driven intuitive programming framework, applied to a self-adaptive system that autonomously inspects helicopter blades measuring up to 7 m. The focus will be given to the system's versatility while performing tapping tests and multispectral surface inspection along the helicopters' blades.

This study proposes a framework for the rapid assessment of seismic fragility and risk of reinforced concrete (RC) circular bridge piers affected by corrosion. The methodology integrates a novel computer vision (CV) algorithm to enhance visual inspections for corrosion level identification, combined with a probabilistic approach to seismic fragility analysis. The aim of the methodology is to quantify the impact of corrosion-induced deterioration on structural performance, expressed as an increment in terms of seismic risk. The first part of the framework consists of defining a custom convolutional neural network able to automatically predict the corrosion severity class starting from a metric-photographic survey. The proposed network incorporates attention mechanisms and color space transformations to ensure robust performance under varying image conditions. The output is used within a probabilistic-based structural modelling and analysis framework, which allows to derive seismic performance of the considered bridge pier typology. On the modelling side, a specific fiber-based approach was employed, in order to account for non-uniform cross-sectional corrosion and current deterioration condition. The results are returned in terms of seismic fragility and risk metrics for quantifying the reduction of seismic performance with respect to the initial conditions. The framework was tested on a real-life case-study exhibiting non-uniform cross-sectional base corrosion, and subsequently, additional scenarios considering full-section base corrosion at varying severity levels were investigated. The outcomes of this study demonstrate the potentialities of artificial intelligence in improving the current practices in the field of seismic assessment of aging RC infrastructures.

Ensuring crop yields in a world with a changing climate and increasing population is of utmost importance. To achieve this goal, genotypical and phenotypical techniques can be exploited to evaluate whether specific mutations prove to be more resilient to drought stress or extreme climate phenomena. This process can be further improved by providing domain experts with automated analytics tools to assist them in the analysis of a large amount of data, therefore providing more statistical reliability and robustness. To this end, this work proposes an image processing pipeline for helping practitioners in the automatic evaluation of the response of tomato plant mutations to drought stress. The pipeline is based on an efficient implementation of a colour space conversion, aiming at reducing the dependency on illumination of the a and b components of the CIELab colour space and providing statistical uniformity, and anisotropic Gaussian mixture model clustering for improved feature extraction and analysis. The results were evaluated over a total of 388 mutations in three time instants, highlighting the effectiveness of the workflow in supporting domain experts in assessing the most resilient tomato variants.

Machine learning (ML) offers promising capabilities for predicting rail infrastructure failure and enabling a shift from diagnostic to prognostic railway maintenance. However, the real-world adoption of high-performing ML models in safety-critical domains such as railway systems hinges on their trustworthiness, particularly their interpretability and transparency. This study, based on a case study in track geometry management, explores the trade-off between accuracy and interpretability in predicting track alignment failures by comparing six ML classifiers: Logistic Regression, Random Forest, Gradient Boosting, XGBoost, Support Vector Machine (SVM), and a Neural Network (NN). The models were trained on railway defect datasets using features such as operating speed, train traffic, total gross tonnage, and defect length. Performance was evaluated using recall as the primary metric, given the high cost of false negatives in rail safety contexts. Results showed that SVM and NN models achieved the highest recall (0.704 and 0.734, respectively), but at the cost of lower interpretability. To address this, post-hoc Explainable AI (XAI) techniques, including SHAP and LIME, were applied. These methods collectively enhance both local and global interpretability and support model transparency, stakeholder trust, and the bridging of the gap between predictive performance and decision-making needs. While XAI is increasingly applied in other sectors, its use in asset management and particularly railway predictive maintenance remains limited. This work fills that gap by demonstrating how XAI can foster more informed and confident adoption of ML models in rail infrastructure management. These explainability techniques help domain experts and end users understand why a model produced a specific result and what key factors influenced that decision, while also supporting data scientists and developers in refining model performance. For instance, feature refinement guided by SHAP improved SVM recall from 0.704 to 0.716.