Search Results (8,942)

Motorcyclists face a disproportionately high risk of severe injury or death compared to other road users, highlighting the need for intelligent rider assistance technologies. This paper presents an initial, modular, and interpretable AI pipeline that generates context-aware safety advice from first-person motorcycle videos with practical inference latency suitable for on-device deployment, framing large language models as interpretable cognitive support agents for motorcycle safety. The system integrates lightweight perception and reasoning components to emulate the function of an Advanced Rider Assistance System (ARAS). Video frames are processed at 1 FPS using Pixtral, a Mistral-based multimodal large language model (MLLM), to produce descriptive scene captions, while YOLOv8 identifies key objects such as vehicles, pedestrians, and road hazards. A Mistral-small language model then fuses this information to generate concise, imperative safety tips. Preliminary evaluations on publicly available motorcycle POV datasets demonstrate promising performance in terms of contextual accuracy, interpretability, and scalability, suggesting potential for real-world deployment in low-resource or embedded environments. The proposed framework offers interpretable, context-aware safety assistance that is particularly valuable for young and newly licensed riders during the transition from supervised training to independent riding, where real-time hazard interpretation support is most needed. Full article

As populations age, the gap between care needs and available support systems is widening, leading to critical vulnerabilities in staffing, infrastructure, and funding. The need for accessible, human-centred, and ethically grounded care technologies is growing. However, the development of digital health tools often lacks inclusivity and practical guidance. Existing ethical frameworks tend to remain abstract, which limits their real-world application. This study examines how such frameworks support the responsible development and implementation of caring technologies in older adult care. To achieve this, in-depth interviews were conducted with care providers, technology developers, and policymakers from partner organisations of the EMPOWERCARE project in the four participating countries: the UK, the Netherlands, Belgium and France. A core challenge was the limited applicability of abstract ethical principles in daily care settings. While existing initiatives often define ethical domains, few offer a structured, actionable process to guide implementation in practice. The proposed guide responds with a step-by-step structure, practical examples, and participatory tools to support inclusive, value-driven technology adoption. It is envisioned both as an implementation aid and a quality label to align stakeholders. Future research should validate the guide’s usability, explore its role across care contexts, and examine how ethics can be more firmly embedded in innovation governance. Full article

Artificial Intelligence is increasingly embedded in national security, defense, and critical infrastructure systems, yet the security of these systems remains insufficiently addressed in traditional cybersecurity education. National initiatives led by the National Security Agency and the National Science Foundation have identified the Security of Artificial Intelligence (SecureAI) as a distinct educational priority supported by formal knowledge units and program validation requirements. Concurrently, workforce data and federal reporting reveal persistent shortages of qualified cybersecurity professionals, particularly in defense and government sectors. This paper presents Illinois Institute of Technology as a case study in the design of a SecureAI applied concentration aligned with NSA-style knowledge units and Center of Academic Excellence principles. The paper demonstrates how a four-course SecureAI program, anchored by a shared undergraduate and graduate cybersecurity foundation, addresses emerging AI security risks while strengthening the national cybersecurity workforce pipeline. Full article

The widespread connectivity of the Industrial Internet of Things (IIoT) improves the efficiency and functionality of connected devices. However, it also raises serious concerns about cybersecurity threats. Implementing an effective intrusion detection system (IDS) for IIoT is challenging due to heterogeneous data, high feature dimensionality, class imbalance, and the risk of data leakage during evaluation. This paper presents a leakage-safe hybrid intrusion detection framework that combines text-based and numerical network flow features in an IIoT environment. Each network flow is converted into a short text description and encoded using a frozen Large Language Model (LLM) called the Bidirectional Encoder Representations from Transformers (BERT) model to obtain fixed semantic embeddings, while numerical traffic features are standardized in parallel. To improve class separation, class prototypes are computed in Principal Component Analysis (PCA) space, and cosine similarity scores for these prototypes are added to the feature set. Class imbalance is handled only in the training data using the Synthetic Minority Over-sampling Technique (SMOTE). A Random Forest (RF) is used to select the top features, followed by a Histogram-based Gradient Boosting (HGB) classifier for final prediction. The proposed framework is evaluated on the Edge-IIoTset and ToN_IoT datasets and achieves promising results. Empirically, the framework attains 98.19% accuracy on Edge-IIoTset and 99.15% accuracy on ToN_IoT, indicating robust, leakage-safe performance. Full article

Engineering education is increasingly expected to prepare graduates capable of addressing sustainability challenges, public safety concerns, and ethical responsibilities. However, in many civil and environmental engineering curricula, sustainability and ethics are still treated as supplementary topics rather than being systematically embedded in core technical courses. This study reports a sustainability-oriented curriculum reform implemented in a Building Water Supply and Drainage Engineering course, integrating Education for Sustainable Development (ESD) principles into CDIO-aligned project-based learning activities. A single-group pre–post quasi-experimental design was adopted with 100 undergraduate students. Quantitative data were collected using a competency-based questionnaire, and paired-sample t-tests, effect sizes, and 95% confidence intervals were applied to examine changes in students’ self-reported competencies. Qualitative data were obtained from reflective learning reports and analyzed through thematic analysis. The results indicate statistically significant improvements in sustainability awareness, ethical and professional responsibility, human-centered design, and systems thinking, with large effect sizes. These findings provide context-specific descriptive evidence supporting the feasibility of embedding sustainability and ethical responsibility within discipline-specific technical engineering courses. Nevertheless, the absence of a control group and the reliance on self-reported measures limit causal interpretation. Future research is recommended to adopt comparative or longitudinal designs and incorporate more objective performance-based assessments. Full article

Electric vehicles have strengthened the coupling between transportation systems and power distribution networks, giving rise to new challenges in the coordinated management of traffic flow and charging demand. Monetary incentives, such as tariffs and subsidies, have been widely adopted to influence drivers’ route and charging decisions and to improve system-level performance. This paper proposes a user-centric incentive framework in which a system operator allocates rewards to guide drivers’ behavior, thereby enabling coordinated operation of power–traffic networks. A reward scheme is developed to provide joint subscription-based and path-based incentives that account for drivers’ behavioral responses through a logit choice model for scheme adoption embedded within a traffic assignment model. The resulting interaction is formulated as a bilevel optimization problem, in which a coupled power–traffic system operator determines incentive schemes to achieve system optimality within a given budget constraint, while individual drivers respond by selecting routes and charging strategies to minimize their perceived travel costs. A single-level Karush–Kuhn–Tucker (KKT) reformulation is developed, and linearization techniques are employed to compute the resulting equilibrium, yielding a tractable mixed-integer second-order cone program (MISOCP). Numerical experiments demonstrate the effectiveness of the subscription-based and path-based reward schemes in improving network performance and budget saving. Full article

Off-grid integrated energy systems offer a practical solution for remote regions lacking access to the main power grid; however, their planning and design are challenged by the inherent uncertainty of renewable energy resources. To address this issue, this paper proposes a stochastic optimization framework for off-grid integrated energy systems that explicitly accounts for wind speed and solar irradiance variability. Continuous probability distributions combined with Monte Carlo sampling are employed to generate stochastic scenarios, which are embedded into a bi-objective optimization model minimizing total system cost and pollutant emissions under power balance and device operational constraints. Unlike existing studies that primarily focus on cost–reliability trade-offs, this work introduces the Renewable Energy Penetration Rate (REPR) as a quantitative, planning-oriented indicator and systematically investigates its interactions with economic performance, pollutant emissions, and renewable uncertainty. The REPR is not only used to characterize renewable utilization levels, but also to support investment decision-making and the comparative assessment of Pareto-optimal configurations. A real-world off-grid service area is adopted as a case study. The results show that increasing the REPR leads to a significant reduction in carbon emissions while exhibiting a nonlinear impact on total system cost. Specifically, the proposed framework identifies a Pareto-optimal solution set in which the total system cost varies within 40–92 million ¥, carbon emissions are reduced by 86% compared with diesel-dominated configurations, and the REPR increases from 70% to 96.4% as renewable capacity expands. In addition, the analysis reveals that higher renewable volatility requires a larger stochastic sample size to ensure solution stability. These findings demonstrate that the proposed framework provides a more comprehensive and decision-relevant assessment of off-grid integrated energy systems under renewable uncertainty, thereby offering practical insights for low-carbon and economically viable system planning. Full article

Background/Objectives: Disruptive behavior problems are common in early childhood, yet access to evidence-based parent training remains limited in many communities due to workforce shortages and service delivery barriers. Behavioral Skills Training for Families (BSF) is a Parent–Child Interaction Therapy (PCIT)-informed, home-based behavioral skills practice model designed to be delivered by bachelor’s-level paraprofessionals under close supervision. This pilot study evaluated the feasibility and preliminary caregiver and child outcomes associated with the Child-Directed Interaction (CDI) module of BSF to inform refinement of training and implementation protocols and guide future evaluation. Methods: Using a non-randomized pre–post design embedded within routine services, caregiver–child dyads (children ages 2–10 years) receiving BSF CDI across community-based agencies in Minnesota were included. Outcomes were assessed using observational coding of caregiver skills (Dyadic Parent–Child Interaction Coding System; DPICS) and caregiver-reported child behavior measures (Eyberg Child Behavior Inventory [ECBI]; Weekly Assessment of Child Behavior–Positive [WACB-P]). Paired-sample t-tests with intent-to-treat analyses examined changes from the baseline to the last attended CDI session. Results: Caregivers demonstrated statistically significant and large increases in observed positive parenting skills and reductions in negative verbalizations during child-led play. Children showed significant reductions in disruptive behavior intensity and problem scores on the ECBI, reflecting movement toward clinically meaningful improvement. No significant change was observed in caregiver-reported positive child behaviors on the WACB-P. Post hoc analyses were conducted to further explore these differences and found consistent changes in the ECBI for cases, regardless of no reported changes in positive child behaviors on the WACB. Conclusions: The results provide preliminary evidence that a structured, PCIT-informed CDI skills practice model can be feasibly implemented by paraprofessionals and is associated with meaningful improvements in caregiver behavior and child behavior outcomes in the first 2–3 months following service initiation. The findings support BSF as a promising workforce-embedded approach and inform future controlled studies examining effectiveness, sustainability, and broader implementation outcomes. Full article

International accreditation has become a pivotal mechanism through which universities outside Europe seek legitimacy and alignment with global quality regimes, particularly regarding sustainable development goals (SDGs). This study investigates how non-EU universities adapt to ASIIN accreditation, focusing on its role in developing a sustainable quality culture that supports long-term educational excellence and social responsibility. Drawing on new institutionalism, the analysis views accreditation as a process of institutional change under isomorphic pressures necessary for the sustainability of quality assurance (QA). Data were derived from a triangulated dataset, including 78 publicly available final accreditation reports via the DEQAR database and expert on-site observations across multiple non-EU universities. The analysis identifies systemic challenges, such as ‘facade conformity’ in learning outcomes and fragmented QA loops, which reveal an ‘adaptive lag’ impeding the sustainable implementation of quality standards. The study concludes by proposing an “Expert-Facilitated, Institutionally-Embedded Evidence Loop” framework to bridge external compliance and internal quality enhancement, thereby ensuring the long-term viability and global relevance of engineering education in alignment with SDGs. Full article

Unsafe behavior remains a dominant contributor to accidents in complex socio-technical systems (STSs), yet it is still frequently interpreted as an individual-level information failure. This study argues that unsafe behavior is more accurately understood as a systemic outcome shaped by multi-level technological, organizational, and environmental conditions. To address this gap, an integrated human factor risk analysis framework is proposed by combining the STS perspective with safety information cognition (SIC) theory. The framework conceptualizes unsafe behavior as the result of risk transmission through safety information flows, linking system-level risk sources to individual perception, cognition, decision-making, and action. Within this perspective, human factor risk does not arise directly from individual error, but from deficiencies and asymmetries in the generation, transmission, and utilization of safety-related information embedded in the STS. Based on this conceptualization, a system-oriented human factor risk analysis (HRFA) approach is developed to support the identification, assessment, and control of unsafe behaviors across both accident scenarios and operational contexts. The framework is applied to road transportation of dangerous goods in China, a typical high-risk STS. The application results demonstrate that the proposed approach can effectively distinguish the comprehensive risk characteristics of different unsafe behaviors and reveal their underlying systemic causes. This study contributes to systems thinking in safety governance by shifting the analytical focus from individual behavior correction to upstream system conditions and information processes. The proposed framework provides a transferable approach for understanding and managing human factor risk in complex STSs and offers practical implications for proactive, system-oriented safety governance. Full article

Accurate state estimation is essential for the performance of uncalibrated visual servoing systems, yet it is frequently undermined by non-Gaussian disturbances—such as impulse noise, motion blur, and occlusions—whose heavy-tailed statistical characteristics are not adequately represented by conventional Gaussian models. To address this issue, this paper presents an evolutionary neural-enhanced intelligent controller designed for robotic visual servoing under such noise conditions. The controller architecture incorporates a hybrid estimation core that integrates $\alpha$-stable distribution modeling for principled noise characterization with an Interacting Multiple Model Kalman filter (IMM-KF) to address system dynamics and uncertainties. A multi-layer perceptron (MLP), optimized globally via the Stochastic Fractal Search (SFS) algorithm, is embedded to provide adaptive compensation for residual estimation errors. This integration of statistical modeling, adaptive filtering, and evolutionary optimization constitutes a coherent learning-based control framework. Simulations and physical experiments reveal that the proposed method enhances improvements in estimation accuracy and tracking performance relative to conventional approaches. The outcomes indicate that the framework offers a functional solution for vision-based robotic systems operating under realistic conditions where non-Gaussian sensor noise is present. Full article

This paper presents a systematic comparative study of optical character recognition (OCR) architectures for Korean license plate recognition under identical detection conditions. Although recent automatic license plate recognition (ALPR) systems increasingly adopt Transformer-based decoders, it remains unclear whether performance differences arise primarily from sequence modeling strategies or from backbone feature representations. To address this issue, we employ a unified YOLOv12-based license plate detector and evaluate multiple OCR configurations, including a CNN with an Attention-LSTM decoder and a MobileNetV3 with a Transformer decoder. To ensure a fair comparison, a controlled ablation study is conducted in which the CNN backbone is fixed to ResNet-18 while varying only the sequence decoder. Experiments are performed on both static image datasets and tracking-based sequential datasets, assessing recognition accuracy, error characteristics, and processing speed across GPU and embedded platforms. The results demonstrate that the effectiveness of sequence decoders is highly dataset-dependent and strongly influenced by feature quality and region-of-interest (ROI) stability. Quantitative analysis further shows that tracking-induced error accumulation dominates OCR performance in sequential recognition scenarios. Moreover, Korean license plate–specific error patterns reveal failure modes not captured by generic OCR benchmarks. Finally, experiments on embedded platforms indicate that Transformer-based OCR models introduce significant computational and memory overhead, limiting their suitability for real-time deployment. These findings suggest that robust license plate recognition requires joint consideration of detection, tracking, and recognition rather than isolated optimization of OCR architectures. Full article

Canine and feline fibrosarcomas are malignant tumors of mesenchymal origin showing histological, molecular, and clinical structures similar to their human equivalent. In human medicine, cellular myelocytomatosis (c-Myc) has already been proposed as a vital gene that regulates aggressive cell growth and is a predictive marker and possible therapeutic target in fibrosarcomas. The goal of this study was to evaluate c-Myc expression in canine and feline fibrosarcomas by gene and protein analyses and correlate its expression to the histological grading system of the examined cancers. Nineteen archival formalin-fixed and paraffin-embedded canine and feline sarcoma samples were recruited. Sixteen of them were determined to be canine and feline fibrosarcomas. The samples consisted of eight cutaneous samples, four eyelid samples, three oral mucosa samples, and one vulva sample. Histopathological examinations were performed to estimate histological type, differentiation, cellularity, and mitotic count. On each sample, immunohistochemical assessment and mRNA expression were performed using an anti-c-Myc antibody and a c-Myc primer. The differences in the immunoreactivity staining values and the relative folds of the c-Myc mRNAs were categorized using already recognized grading criteria and a statistical analysis was performed. Among all the tested canine and feline fibrosarcomas, 75% of the grade II fibrosarcoma cases and all of the grade III fibrosarcoma cases were positive for the c-Myc protein. Furthermore, the relative values of the c-Myc gene were up-regulated in grade III compare to grade I and II fibrosarcomas. The results demonstrated that c-Myc expression was significantly higher in grade III than in grade II and I (p ≤ 0.05) fibrosarcomas regarding protein and gene levels. In summary, this work provides evidence for the high expression of c-Myc in canine and feline fibrosarcomas and proposes a supposed relationship with poor prognoses as determined by grading systems. Full article

Tomato leaf diseases exhibit subtle inter-class differences and substantial intra-class variability, making accurate identification challenging for conventional deep learning models, especially under real-world conditions with diverse lighting, occlusion, and growth stages. Moreover, most existing approaches rely solely on visual features and lack the ability to incorporate semantic descriptions or expert knowledge, limiting their robustness and interpretability. To address these issues, we propose LPDiag, a multimodal prototype-attention diagnostic framework that integrates large language models (LLMs) for fine-grained recognition of tomato diseases. The framework first employs an LLM-driven semantic understanding module to encode symptom-aware textual embeddings from disease descriptions. These embeddings are then aligned with multi-scale visual features extracted by an enhanced Res2Net backbone, enabling cross-modal representation learning. A set of learnable prototype vectors, combined with a knowledge-enhanced attention mechanism, further strengthens the interaction between visual patterns and LLM prior knowledge, resulting in more discriminative and interpretable representations. Additionally, we develop an interactive diagnostic system that supports natural-language querying and image-based identification, facilitating practical deployment in heterogeneous agricultural environments. Extensive experiments on three widely used datasets demonstrate that LPDiag achieves a mean accuracy of 98.83%, outperforming state-of-the-art models while offering improved explanatory capability. The proposed framework offers a promising direction for integrating LLM-based semantic reasoning with visual perception to enhance intelligent and trustworthy plant disease diagnostics. Full article

Background: Salmonella enterica remains a leading cause of foodborne illness worldwide despite decades of advances in surveillance and control. Traditional interventions have targeted specific points in the food chain, yet recurrent outbreaks show that Salmonella exploits system-wide gaps and inconsistencies. Methods: This review synthesises recent evidence from epidemiology, experimental microbiology, and regulatory practice to evaluate how management decisions, from farm through processing, influence Salmonella risk in livestock-derived foods. Results: Poultry, pig, and cattle farms employ targeted measures, including rodent control, litter management, batch rearing, and secure feed storage, to reduce contamination. The greatest reductions in Salmonella prevalence occur when these measures are embedded in coherent farm-to-fork programmes. Future gains are likely to come less from novel interventions and more from rigorous implementation, integration, and the validation of existing tools, supported by high-resolution surveillance (including whole-genome sequencing) and prevention-focused management systems. Artificial intelligence can enhance control through real-time surveillance, predictive risk modelling, and targeted interventions informed by diverse farm data. Conclusions: Sustained progress in Salmonella control will depend on rigorously applying existing interventions, supported by high-resolution surveillance and prevention-focused management. Carefully governed AI can enhance real-time monitoring and risk prediction, but its value hinges on addressing data, cost, and regulatory challenges. Full article