Search Results (1,961)

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

This study presents a hybrid modeling framework aimed at enhancing climate mitigation strategies by evaluating the predictive power of sustainability indicators using both statistical analysis—correlation metrics, regression modeling, distribution tests—and artificial neural networks (ANNs). The analysis centers on variables critical to climate outcomes, including renewable energy use in transport and electricity, greenhouse gas emissions from production, and aggregated target completion values. The findings identify renewable energy usage in transport as the primary predictor of improved performance in the Sustainable Development Report (SDR), followed by overall target completeness, electricity-based renewables, and production-related emissions. Multidimensional interaction analyses highlight a synergetic link between transport renewables and target achievement, underscoring their strategic relevance for climate mitigation efforts. The ANN models demonstrate high predictive accuracy and minimal error, affirming the model’s suitability for scenario-based climate forecasting. Results offer actionable intelligence for policymakers and climate stakeholders to optimize resource allocation and accelerate low-carbon transitions. The study acknowledges limitations, namely, the relatively small dataset and EU-centric analysis, and recommends future extensions to more geographically diverse datasets and the incorporation of advanced econometric techniques and AI frameworks to improve generalizability and predictive potency. Full article

Desalination is an increasingly important element in the sustainable supply of potable water. To accurately predict costs, the efficiency of such systems requires accurate knowledge of seawater’s thermodynamic properties. Four models have been proposed for determining the thermophysical properties of salt water, pure water, an ideal mixture, and an aqueous sodium chloride solution, and empirical correlations, as would be expected, provide the precision necessary for accurate exergy calculations. This research began with a study of the most recent and accurate empirical investigations of the thermodynamic properties of seawater. It then employed AI techniques to develop a simpler, more accurate model for density, Gibbs free energy, specific enthalpy, and specific entropy for pressures extending up to 12 MPa, salinities from 0 to 80 g/kg, and the temperature range of 10 °C to 120 °C. The AI-based correlations achieved absolute errors of 1.5 kg/m³ for density, 0.185 kJ/kg for specific enthalpy, 0.005 kJ/kg·K for specific entropy, and 0.214 kJ/kg for Gibbs free energy. These values demonstrated at least equivalent, and even superior, accuracy to the existing state-of-the-art formulations, with the advantage of significantly reduced computational complexity, enhanced computational efficiency, and a more user-friendly implementation. Validation against experimental data demonstrated the exceptional accuracy of the predicted values for all the stated thermodynamic properties. In addition, an exergy-based assessment was conducted of the performance of a recently commissioned desalination plant in Kuwait. This was a large-scale multi-effect distillation plant with thermal vapour compression (MED-TVC), showing a second-law efficiency of 8.9%, with the primary source of exergy destruction identified as the evaporator units. Comparative assessment with a more conventional approach showed differences of less than 0.4% in total exergy destruction and less than 5% in exergetic efficiency. This is taken as a validation of the accuracy, reliability, and practical usefulness of the proposed AI framework for the performance evaluation of desalination systems. Full article

This study investigates how two widely used language-modelled generative AI tools, ChatGPT-5.1 (with DALL·E 3) and Gemini 3 (with Imagen), perform as early-stage co-design partners for biophilic interior design. Focusing on real-world use rather than theoretical capability, the research asks to what extent these systems can generate conceptually robust, visually coherent and practically feasible proposals when designers explicitly request biophilic strategies. A multiple-case design was employed across three scenarios: (1) an empty “tabula rasa” room, (2) a damaged rustic room requiring contextual renovation, and (3) a hospital staff break room to be transformed into a “cognitive restoration sanctuary.” For each case, both tools were prompted to produce a step-by-step biophilic design plan and a corresponding photorealistic image. Textual outputs were coded against the 14 Patterns of Biophilic Design and related restorative concepts, while images were evaluated by an expert panel of 15 architects with formal training in biophilic design using a structured Likert-scale instrument. Exterior and building-scale applications were not assessed. Results show that both systems can articulate broadly plausible biophilic strategies but differ in emphasis: ChatGPT tends to produce more spatially coherent, pattern-rich and functionally grounded plans, whereas Gemini excels more in visual realism and atmospheric rendering. Expert ratings indicate a consistent, though not overwhelming, preference for ChatGPT in spatial composition, human-spatial responses, contextual fit, and strategic support for cognitive restoration, with a slight advantage for Gemini in visual realism. Across all cases, however, plan-to-image fidelity is limited, particularly for non-visual and operational patterns (e.g., sound, scent, thermal variability, circadian systems, infrastructure access). The findings suggest that current generative AI tools are best positioned as fast, co-creative aides for early exploration of biophilic ideas, rather than as reliable autonomous consultants for evidence-based, cognitively targeted biophilic design. Full article

The Internet of Things (IoT) has transformed modern life through interconnected devices enabling automation across diverse environments. However, its reliance on legacy network architectures has introduced significant security vulnerabilities and efficiency challenges—for example, when Datagram Transport Layer Security (DTLS) encrypts transport-layer communications to protect IoT traffic, it simultaneously blinds intermediate proxies that need to inspect message contents for protocol translation and caching, forcing a fundamental trade-off between security and functionality. This paper presents an architectural solution based on the Recursive InterNetwork Architecture (RINA) to address these issues. We analyze current IoT network stacks, highlighting their inherent limitations—particularly how adding security at one layer often disrupts functionality at others, forcing a detrimental trade-off between security and performance. A central principle underlying our approach is the role of structural symmetry in RINA’s design. Unlike the heterogeneous, protocol-specific layers of TCP/IP, RINA exhibits recursive self-similarity: every Distributed IPC Facility (DIF), regardless of its position in the network hierarchy, instantiates identical mechanisms and offers the same interface to layers above. This architectural symmetry ensures predictable, auditable behavior while enabling policy-driven asymmetry for context-specific security enforcement. By embedding security within each layer and allowing flexible layer arrangement, RINA mitigates common IoT attacks and resolves persistent issues such as the inability of Performance Enhancing Proxies to operate on encrypted connections. We demonstrate RINA’s applicability through use cases spanning smart homes, healthcare monitoring, autonomous vehicles, and industrial edge computing, showcasing its adaptability to both RINA-native and legacy device integration. Our mixed-methods evaluation combines qualitative architectural analysis with quantitative experimental validation, providing both theoretical foundations and empirical evidence for RINA’s effectiveness. We also address emerging trends including AI-driven security and massive IoT scalability. This work establishes a conceptual foundation for leveraging recursive symmetry principles to achieve secure, efficient, and scalable IoT ecosystems. Full article

The rapid growth of renewable integration and active consumer participation has made modern power grids increasingly complex and dynamic, where maintaining balanced and efficient energy distribution remains a central challenge. This paper introduces a symmetry-aware optimized fuzzy deep reinforcement learning-gated recurrent unit (OF-DRL-GRU) model that exploits the natural symmetry and asymmetry in demand–generation behavior to achieve stable and adaptive load balancing. The proposed architecture consists of four core modules: a fuzzy logic layer that formulates symmetrically distributed membership functions for interpretable and balanced state transitions; a DRL agent that governs decision actions through a symmetry-preserving reward mechanism balancing exploration and exploitation; a GRU network that models temporal symmetries while performing controlled symmetry-breaking during dynamic fluctuations to enhance generalization; and an improved multi-objective biogeography-based optimization (IMOBBO) algorithm that optimizes fuzzy parameters and model hyper-parameters through adaptive migration alternating between symmetry preservation and deliberate asymmetry, ensuring efficient convergence and global diversity. The synergy among these modules forms a unified symmetry-aware optimization paradigm, reflecting how symmetric structures sustain stability while purposeful asymmetry enhances robustness and adaptivity. The proposed framework is evaluated using three benchmark datasets (UK-DALE, Pecan Street, and REDD) and compared against several advanced and competitive models. Experimental outcomes show that the proposed OF-DRL-GRU model achieves 99.23% accuracy, 99.69% recall, and 99.83% area under the curve (AUC), alongside faster runtime, lower variance, and improved convergence stability. These results demonstrate that incorporating symmetry–asymmetry principles within AI-driven optimization significantly enhances interpretability, resilience, and energy efficiency, paving the way for intelligent, self-adaptive load management in next-generation smart grids. Full article

Filter pruning is an effective approach for improving the inference efficiency of neural networks and is particularly attractive for on-device artificial intelligence (AI) applications. However, many existing methods fail to accurately identify redundant filters due to limited modeling of inter-filter dependencies. A filter pruning method based on nuclear norm analysis is proposed to quantify filter independence and guide structured pruning. By analyzing the layer-wise distribution of independence scores, a principled trade-off between pruning rate and accuracy preservation is achieved. In most evaluation scenarios, the proposed method achieves 75–95% parameter reduction and 70–80% FLOPs reduction, while substantially higher compression ratios (up to 99%) can be obtained for more redundant network architectures, with consistent performance trends observed across multiple accuracy-related metrics. Furthermore, deployment on an RK3588 neural processing unit (NPU) demonstrates substantial reductions in memory consumption and inference latency, confirming the practical effectiveness of the method for mobile and edge AI applications. Full article

Avian influenza (AI), particularly highly pathogenic avian influenza (HPAI), represents a serious and growing threat to global poultry production, international trade, and human health security. Control of AI is complicated by the high evolutionary rate of influenza A viruses, which drives antigenic diversity and ongoing emergence of novel strains. Effective surveillance and disease management therefore depend on timely and accurate diagnostics. While conventional methods—including virus isolation, reverse transcription-quantitative polymerase chain reaction (RT-qPCR), and enzyme-linked immunosorbent assays (ELISAs)—remain effective and widely used, they are limited by long turnaround times, the need for specialized equipment, and reliance on highly trained personnel. In addition, strict state and federal regulatory requirements restrict testing to a limited number of authorized laboratories. Although these regulations are essential for maintaining diagnostic accuracy and quality assurance, they place substantial strain on laboratory capacity during outbreaks and delay actionable results. The need for rapid, on-site decision making has driven interest in alternative diagnostic approaches, including biosensor technologies. A major limitation of current diagnostic strategies is the lack of robust DIVA (Differentiating Infected from Vaccinated Animals) capability. In countries such as the United States, where poultry vaccination against AI is not routinely practiced, the absence of DIVA-compatible diagnostics has hindered adoption of vaccination as a disease management tool, as seropositive birds and products face significant trade restrictions. Biosensor platforms capable of enabling DIVA strategies offer a potential pathway to support vaccination while preserving surveillance integrity. This review examines the current landscape of AI and HPAI diagnostics, emphasizing the limitations of traditional approaches and the opportunities presented by biosensor platforms. We evaluate electrochemical, optical, piezoelectric, and nucleic-acid-based biosensors, with particular attention to biorecognition strategies, performance metrics, field deployability, and applications supporting subtype discrimination, DIVA implementation, and One Health surveillance. Full article

In this paper, the performance of dual-hop relay transmission in modern wireless communication systems is analyzed by considering two fundamental relaying techniques, namely, Amplify-and-Forward (AF) and Decode-and-Forward (DF). The propagation conditions on the source–relay (S-R) and relay–destination (R-D) links are modeled using the κ-μ statistical distribution, which effectively captures the fading characteristics in both line-of-sight (LoS) and non-line-of-sight (NLoS) environments. The analysis focuses on key performance metrics, including the outage probability (P_out) and average bit error probability (P_e), for Binary Phase Shift Keying (BPSK) and Quadrature Phase Shift Keying (QPSK) modulation schemes, assuming transmission via a single relay without a direct S–D link. Closed-form expressions for the considered metrics are derived based on the κ-μ model and verified by numerical evaluation. In addition to classical analytical modeling, a Generative Artificial Intelligence (GenAI)-enabled workflow is incorporated as a supportive tool in order to aid in automated analysis, the interpretation of the results in the context of network management under varying channel and system parameters based on the P_out and P_e calculations with the aim to tackle the underlying complexity and cognitive load of infrastructure adaptation and re-configuration operations. The combined analytical and GenAI-assisted approach provides valuable insights for the optimization, design, and continuous evolution of robust relay-based architectures in next-generation wireless networks. Full article

Background: Artificial intelligence (AI) is becoming important in oncology, supporting risk prediction, treatment planning, and biomarker discovery. However, current evaluation practices often assume that high predictive accuracy implies reliable interpretation—a misconception that may undermine reproducibility and clinical decision-making. This study aims to reassess interpretability by introducing feature ranking order consistency as a stability-focused metric to evaluate how model explanations respond to minimal input perturbations. Methods: Using The Cancer Genome Atlas (TCGA) breast cancer multi-omics dataset, we compared supervised models—Linear Regression, Least Absolute Shrinkage and Selection Operator (LASSO), Random Forest, and Extreme Gradient Boosting (XGBoost)—with unsupervised and statistical methods, including Principal Component Analysis (PCA), Highly Variable Gene Selection, and Spearman’s rank correlation. Each method produced a Top 20 feature ranking, and stability was assessed by testing whether rankings remained consistent after removing the top-ranked feature. Predictive performance was evaluated using a Random Forest classifier with stratified 10-fold cross-validation. Results: Supervised models exhibited unstable feature importance rankings even under minimal perturbations (<0.1% feature removal), suggesting that high predictive accuracy may obscure fragile or misleading explanations. In contrast, Highly Variable Gene Selection and Spearman’s correlation consistently produced stable, biologically coherent feature sets and maintained competitive predictive performance. Conclusions: Interpretive instability is a major limitation of many machine learning models in oncology. Incorporating stability-based criteria—such as feature ranking consistency—into evaluation frameworks is essential for ensuring reproducible, trustworthy, and clinically actionable AI. As AI adoption accelerates, prioritizing interpretability alongside accuracy is critical for responsible deployment in precision oncology. Full article

Background: Heart rate variability (HRV) is a valuable indicator for assessing vascular health, but keeping clinical decision support systems (CDSSs) aligned with the rapidly evolving literature remains challenging. This study aimed to develop an LLM-assisted literature synthesis framework to support transparent HRV-based risk stratification, enabling systematic extraction and organization of HRV evidence from published studies. Methods: An LLM-driven framework was developed to extract HRV parameters from 140 medical abstracts. The system simulated step-by-step human reasoning to identify key HRV indicators and group patient data using predefined statistical thresholds derived from the literature. System performance was evaluated using ECG-derived HRV features as a feasibility evaluation of literature-guided HRV classification. Results: The proposed framework demonstrated an accuracy of 86% in literature-guided HRV classification, with a sensitivity of 81% and a specificity of 87%. Compared with traditional machine learning approaches, the LLM-assisted system provided transparent, literature-grounded reasoning and could be readily updated as new studies became available. Conclusions: Large language models can support evidence-guided parameter selection and feasibility-level HRV-based risk stratification, rather than serving as predictive classifiers. This approach reduces manual effort, enhances transparency, and addresses common “black box” concerns associated with AI-assisted CDSS development in clinical practice. Full article

Global agriculture faces unprecedented challenges, including a projected population of 10 billion by 2050, declining arable land, and the urgent need to phase out antibiotic growth promoters (AGPs) to stem antimicrobial resistance (AMR). This review evaluates fermentation technology as a sustainable solution to the “food–feed–fuel” three competing land uses. We systematically compare solid-state fermentation (SSF) and submerged fermentation (SmF), highlighting their quantitative advantages: SSF offers 2–3× higher volumetric productivity and 70–90% lower water usage for solid wastes (e.g., soybean meal, wheat bran), while SmF provides superior process control for high-value products (e.g., single-cell protein). Key molecular mechanisms are discussed, including enzymatic degradation of anti-nutritional factors (up to 95% phytate and 98.8% tannin removal), mycotoxin detoxification (60–80% reduction), and biosynthesis of bioactive compounds (e.g., vitamin B12 enrichment up to 15-fold). Fermented feeds benefit many livestock species, particularly in organic and high-density farming systems, improving growth performance, gut health, and disease resistance while reducing environmental footprints. Advanced technologies such as AI-driven digital twins, CRISPR-based strain engineering, and precision fermentation are explored to overcome bottlenecks, including heat dissipation, strain stability, and process control. Despite challenges in scale-up, economics, and divergent global regulations (EU, USA, China, Southeast Asia, and Africa), fermentation is a critical biotechnological paradigm for circularity—the circular bioeconomy—and long-term food security. Future research should prioritize cost-effective large-scale implementation and the harmonization of regulatory frameworks. Full article

Smart suspensions have significantly improved passenger comfort and vehicle stability compared to their passive counterparts. This manuscript explores the integration of artificial intelligence (AI) into hybrid suspension control systems to enhance vehicle stability and ride comfort under conditions where suspended mass changes. A one-quarter-vehicle model is employed to simulate and evaluate the performance of a hybrid control strategy, which combines skyhook and groundhook methods using a dynamic weighting factor ($\alpha$). This investigation considers an everyday situation where the sprung mass of a vehicle changes considerably when passengers enter or exit the automobile, impacting the suspension performance. Reinforcement learning techniques are utilized to optimize $\alpha$, achieving an acceptable balance between passenger comfort and vehicle stability. Simulation results show significant improvements in the dynamic response of the sprung mass compared to traditional passive systems, while keeping vehicle stability. Although improvements in road holding are incremental, simulation effort validates the AI-based hybrid system’s potential for refinement and practical application. Validation in MATLAB-Simulink demonstrates the system’s adaptability to varying road conditions and load distributions. The findings highlight the transformative role of AI in suspension control, paving the way for real-time implementation, advanced algorithms, and integration into full-vehicle models. This study contributes to the ongoing development of intelligent suspension systems toward vehicle performance advancement by improving passenger comfort and road holding. Full article

Background/Objectives: Diabetes is one of the most familiar and common diseases among people currently, and is a type of metabolic disease that is caused due to high levels of sugar in the blood for longer periods of time. If the disease is predicted at an earlier stage, the severity and risks associated with diabetes are significantly reduced, which helps to save the lifespan of people. In earlier investigations, various kinds of automated models based on artificial intelligence (AI) were developed for this purpose. However, key issues still revolve around the lack of robustness, dependability, and precise prediction. The motivation behind the proposed study is to design and develop an automated tool for the diagnosis of chronic disease with the use of novel AI methodology. Methods: For this purpose, a new detection framework is introduced, known as the Brass Optimized Learning-Based Diabetes Prediction (BOLD) model for remote healthcare applications. By using this kind of optimization-integrated deep learning technique, the overall performance and efficiency of the diabetes detection system are maximized. This framework preprocesses the input diabetes dataset before performing the data splitting, normalization, and cleaning activities. Next, the best attributes for improving the prognostic performance of the classifier are chosen using the Brassy Pelican Optimization (BPO) procedure. The Hunting Optimized Recurrent Neural Network—Long Short-Term Memory (RNN-LSTM) method is used to categorize the people into those who are diabetic and those who are not based on the chosen attributes. The approach employs a Deer Hunting Optimization (DHO) method to choose the hyperparameters needed to make an informed choice. A variety of parameters have been employed to confirm the results, which are evaluated for performance verification using the PIDD, Indonesia diabetic database, and kidney disease dataset. Results: The BOLD framework is successful to the extent that it has been able to achieve several metrics of comparably good results, such as an RMSE value of 0.015, a Cohen’s Kappa measure of 0.99, a precision of 0.991, a recall of 0.99, an accuracy equal to 0.996, and an AUC equal to 0.99. Conclusions: It is also remarkable that a very short time of 0.8 s was enough for it to deliver this kind of performance, making it a neat combination of both time and power efficiency. Full article

This study evaluates the service quality of smart campus dining services as a core element of sustainable school meal governance and health-promoting campus environments. A structured questionnaire grounded in the five SERVQUAL dimensions—tangibles, reliability, responsiveness, assurance, and empathy—was administered to 375 users of a smart campus catering platform, including students, faculty and staff, and education administrators from 20 counties and cities in Taiwan. The data were analyzed using gap analysis, confirmatory factor analysis, multiple regression, and Importance–Performance Analysis (IPA) to identify major service quality gaps and sustainability-oriented improvement priorities. The results show that tangibles, reliability, responsiveness, and assurance significantly predict overall service quality, with assurance exerting the strongest effect, while empathy is highly correlated with the other dimensions. IPA further indicates that outdated or insufficient smart facilities fall into the high-importance/low-performance area and thus represent a critical weakness. These findings provide empirical evidence for data-driven and user-centered management of school meal services, supporting more efficient resource allocation, AI-assisted menu planning, and IoT-based food safety monitoring. By linking service quality assessment with sustainable campus governance, the study contributes to efforts to promote healthy eating, reduce food waste, and strengthen localized food supply collaboration, in line with Sustainable Development Goals related to health, education, and responsible consumption. Full article