Search Results (1,893)

Insulator types and materials vary substantially across power system inspection scenarios, while damage defects occur infrequently; consequently, defect images that match a target insulator type and operating environment are often difficult to obtain. Existing open-source insulator image datasets provide limited coverage of equipment types, scene backgrounds, and defect morphologies. Their direct use for detector training may therefore cause domain mismatch and poor generalization. To address these limitations, this study proposes a scene-domain-adaptive sample expansion method for few-shot damaged-insulator detection. The method adapts a general-purpose pretrained diffusion model to the insulator inspection domain and incorporates three-dimensional (3D) structural constraints to generate targeted samples of damaged insulators. First, low-rank adaptation (LoRA) is used for scene-domain adaptation, enabling the generation model to learn the characteristic geometry, appearance, and material properties of insulators. Second, a 3D model of the target insulator is constructed, and physical damage simulation and edge extraction are applied to obtain geometric guidance maps containing shed boundaries and fracture contours. These maps constrain the locations and shapes of the generated defects. Finally, the geometric guidance is injected into the diffusion process to synthesize damaged-insulator images, which are combined with limited real samples to train detectors for damaged-insulator instances, which were evaluated exclusively on real validation images. Experimental results show that adding a moderate number of generated samples enriches the scarce defect features in the real dataset and improves detector performance. In the mixture-ratio experiment using YOLOv8, mAP@0.5 increased by 8.2 percentage points. Additional experiments with multiple detectors yielded performance gains of varying magnitudes, demonstrating that the generated samples provide an effective supplement to limited, real-world data. The proposed method alleviates the scarcity of insulator defect samples and offers a practical data-augmentation strategy for intelligent inspection of power equipment. Full article

Background: Attention-Deficit/Hyperactivity Disorder (ADHD) affects an estimated 6% of adults in the United States and contributes to a significant economic burden. Telehealth has emerged as a vital tool in the management of ADHD, offering improved access to care, especially for individuals in underserved communities. Despite its growing role, there remain critical gaps in understanding how social determinants of health (SDOH) are associated with disparities in telehealth utilization for ADHD treatment. Objectives and Methods: This study analyzed data from the National Center for Health Statistics (NCHS) Rapid Surveys System (RSS) Round 2: ADHD (October–November 2023), a nationally fielded survey of U.S. adults. Respondents were classified into three groups: never diagnosed, previously diagnosed, and currently diagnosed with ADHD. The study aimed to (1) compare the distribution of SDOH across ADHD status groups and the general adult population to identify factors associated with ADHD diagnosis; (2) assess the homogeneity of SDOH distributions across ADHD groups; (3) evaluate telehealth utilization among adults currently diagnosed with ADHD; and (4) examine the relationship between SDOH and telehealth use for ADHD treatment. Multivariable logistic regression (MVLR) served as a benchmark model, while machine learning (ML) models—including regularized linear regression, support vector machine (SVM), random forest (RF), LightGBM, multilayer perceptron (MLP), and Few-Shot Learning (FSL)—were trained to identify key predictors. Results: A total of 7009 survey responses were analyzed: 124 had a past diagnosis, 444 were currently diagnosed, and the remainder had never been diagnosed with ADHD, corresponding to a current ADHD prevalence of 6.3%. Adults with current ADHD were more likely to be male, single, younger, white, non-homeowners, and frequent users of online health resources. They also reported lower education, income, and financial security. About 70% used telehealth for counseling and prescriptions; insurance covered telehealth visits for 82.32% of users, yet 38.76% reported no coverage of ADHD-related diagnostic or treatment costs. Nineteen SDOH elements across four domains—demographic, socioeconomic, neighborhood/built environment, and healthcare access—were identified as predictors. ML models outperformed MVLR, with SVM and FSL achieving the highest F1 (both 0.63), and FSL the highest recall (0.69). Age, race, marital status, difficulty paying bills, home ownership, education, and household size were the most consistently important variables. Limitations: This study is limited by a cross-sectional design, reliance on self-reported ADHD diagnoses, and a lack of genetic or family-history measures. Additionally, the omission of complex sampling weights limits the national representativeness of these findings. Finally, the small effective sample size poses risks of model overfitting, and the generalizability of the models could not be externally validated due to the unavailability of comparable independent datasets. Conclusions: Despite widespread internet access, disparities in telehealth use for ADHD persist. Among 19 SDOH predictors, age (aOR = 0.56), difficulty paying medical bills (aOR = 2.52), and race (aOR = 1.37) were significantly associated with telehealth use, and all ML models outperformed the MVLR benchmark, though bootstrap CIs overlapped. Future research should incorporate inclusive data collection and stratified modeling to better represent disadvantaged populations and inform equitable access strategies. Full article

Accurate State-of-Health (SOH) prediction is essential for deploying retired electric vehicle batteries into reliable second-life energy storage systems. However, this task is challenged by sparse and noisy operational data from onboard Battery Management Systems (BMS). This study systematically evaluates a spectrum of deep learning architectures for SOH forecasting under BMS-style data constraints derived from laboratory cycling data: a BiLSTM on aggregated cycle statistics (Model A), preliminary zero-shot transfer to a single unseen cell (Model B), a waveform BiLSTM with full intra-cycle voltage, current, and temperature trajectories (Model C), a baseline TCN (Model D) and a probabilistic TCN-GPR hybrid (Model E). All models are constrained to identical low-fidelity BMS-style variables extracted from the NASA battery aging dataset. Model C achieves the lowest point accuracy error of 0.46% ± 0.18% MAE across five random seeds, demonstrating that high-resolution waveform inputs capture degradation signatures, notably voltage plateau morphology, transient dynamics, and implicit SOC information, that aggregated features irreversibly lose. Model D using the same waveform inputs and evaluation protocol as Model C, achieves a MAE of 2.99% at a single seed, providing direct architectural comparison evidence that the BiLSTM’s position-sensitive temporal summarization outperforms GlobalAveragePooling1D under these conditions. Model E achieves a higher MAE of 2.12% ± 0.33% but uniquely provides calibrated predictive distributions of 99.4% ± 1.2% coverage, NLL = −1.877 ± 0.038, with approximately uniform 95% predictive intervals (mean width 19.83% SOH across 34 test cycles at seed = 42), reflecting the near-constant posterior variance produced by the large optimized GPR length-scale under the frozen two-stage training design. A paired t-test confirms that Model C statistically significantly outperforms Model E on point accuracy (p < 0.01). Isotonic regression recalibration reduces mean calibration error from 0.138 to 0.010, demonstrating that shape-level miscalibration is correctable post hoc. The central implication for second-life battery deployment is a clear accuracy–uncertainty trade-off: Model C is preferred when point estimates suffice, while Model E is essential for risk-aware decisions requiring confidence intervals. Full article

Background: Emergency aeromedical services, including helicopter emergency medical service (HEMS) and medical emergency evacuation (MEDEVAC), are critical to time-sensitive care. Dispatch decisions are complex and consequential, determining whether, which, and under what conditions to deploy aeromedical resources. This study reviews modelling approaches for emergency aeromedical dispatch. Methods: Following PRISMA, studies between 2003 and 2026 (June) were screened, yielding 42 studies. Models were classified as predictive and learning-based, sequential decision, and prescriptive optimisation-based, with solution techniques, operational applications, and policy contexts analysed. Results: Markov decision process and approximate dynamic programming models dominate the sequential decision literature, particularly in military MEDEVAC. Prescriptive models support resource allocation, base location, coverage planning, and dispatch optimisation, while predictive and AI/ML-based approaches remain limited but emerging. Key challenges include computational complexity, data uncertainty, policy fragmentation, and ethical concerns. Sequential models reflect dispatch’s dynamic, stochastic nature, where current deployments constrain future resource availability. Priority-aware policies outperform closest-unit rules, but limited real-world validation hinders adoption. Conclusions: This review maps methods and provides evidence-based guidance for researchers and dispatch organisations selecting decision support models. Future opportunities include AI-assisted dispatch, hybrid predictive–prescriptive modelling, real-time adaptive algorithms, sustainability-oriented optimisation, improved helicopter landing zone identification, and standardised ethical and regulatory frameworks. Full article

Development of new agricultural fungicides requires reliable computational tools to screen candidate molecules before investing time, money, and effort in experimental assays. In this study, curated datasets of fungicidal and non-fungicidal compounds were employed in the construction of possible predictive models using diverse machine-learning algorithms and molecular descriptors. The best performance was obtained using balanced datasets and topological descriptors. Two classification models based on different kinds of negative class instances were selected, achieving results of accuracy, sensitivity, and specificity for the test set of 0.806, 0.778, and 0.828 for the first model (RF_BF_14), and 0.937, 0.943, and 0.933 for the second model (RF_BF_17). Differences in performance were consistent with the chemical nature of the negative class. Both models showed excellent applicability domain coverage (>99.6%). Predictions of fungicidal likeness were applied to a curated database of about 1.2 million molecules from the ChEMBL database by using an in-house developed Python3 tool: FLiP-Z (Fungicide Likeness Predictor by Zimeck). Roughly 22% of molecules were predicted by positive consensus as fungicidal candidates. A subsequent screening based on Acute Oral Toxicity reduced the set to 295 molecules with low-toxicity and positive fungicide-likeness predictions. Proprietary rights for FLiP-Z are held by Zimeck C.L.; the tool is available by permission. Full article

This study evaluates the potential and limitations of open/model-derived hydroclimatic data for multi-horizon river discharge forecasting in Babahoyo, Ecuador. A quantitative, applied, retrospective longitudinal design used daily data for 2017–2026. The discharge target is the GloFAS v4 seamless product—reanalysis until July 2022, archived operational forecast thereafter; meteorological predictors are ERA5/ERA5-Land/IFS reanalysis, not forecasts. The framework combined leakage-aware feature engineering, temporal validation, rolling-origin backtesting, naïve baselines, machine-learning regression, conformal prediction intervals, and high-flow classification. Performance was strongest at one day, where models reproduced the signal closely (R² = 0.909), although persistence remained highly competitive. Skill deteriorated at t + 7 and t + 14, where peak timing and magnitude became unreliable. Interval coverage was near-nominal at t + 1 but unreliable at longer horizons. The high-flow classifier identified most q90 cases, yet moderate precision and the absence of gauge validation prevent operational warning claims. Because the target is a 5 km grid simulation of a channel 100–150 m wide, the metrics quantify agreement with GloFAS, not with the physical river, and are reported to three significant digits. Overall, the study is a conservative benchmark for open hydroclimatic data in data-limited tropical floodplains: useful for exploratory monitoring and uncertainty diagnosis, but not a substitute for local hydrometric validation. Full article

Quantifying the impacts of climate change and extreme climatic events on crop yields is essential for safeguarding global food security. The rapid growth of data availability and advances in computational capacity have established machine learning (ML) as a critical tool for unraveling the complex, nonlinear relationships between climatic factors and agricultural productivity. This systematic review synthesizes evidence from 137 peer-reviewed studies published between 2015 and 2025 that applied ML models to assess the effects of both long-term climate trends and discrete extreme events on crop yields worldwide. Bibliometric and thematic analyses reveal a rapidly evolving field, with over 85% of studies published since 2020, and a strong concentration on staple cereals—wheat, maize, and rice—in major agricultural regions including China, the United States, and India. Random Forest (RF) was the most commonly used algorithm; ensemble and deep-learning models achieved high predictive accuracy within well-resourced study contexts. Temperature and precipitation extremes emerged as the most frequently examined stressors, with distinct methodological patterns: studies focusing on climate change trends predominantly employed RF and LSTM models, whereas those investigating extreme events increasingly adopted hybrid approaches that integrate ML with process-based crop models. This review highlights the transformative potential of ML while identifying persistent challenges, such as geographical imbalances in research coverage, the need for enhanced interpretability in extreme event attribution, and the critical importance of modeling compound extremes. Future research should prioritize the development of explainable, causally informed, and transferable ML frameworks to support equitable climate adaptation strategies in global agriculture. Full article

Emotion recognition in children’s drawings is difficult because affect is carried by sparse strokes, symbolic objects, and overall composition rather than by the stable appearance statistics of photographs. We built a reproducible four-class benchmark (Angry, Fear, Happy, Sad) on a single public corpus of 818 children’s drawings and compared three transfer-learning regimes under identical stratified five-fold splits with nested model selection: ResNet-50, ViT-B/16, and an end-to-end fine-tuned SigLIP image encoder (SigLIP-FT). SigLIP-FT reached the highest macro-F1 (0.773 ± 0.028), ahead of ViT-B/16 (0.700 ± 0.040) and ResNet-50 (0.598 ± 0.038), and was the best calibrated (ECE 0.119). Frozen-feature linear probes preserve this ordering (0.541, 0.648, 0.731), locating the advantage in the pretrained representations rather than in fine-tuning, while zero-shot SigLIP reaches only 0.510, so task-specific supervision remains necessary. Margin-based abstention raised retained-set macro-F1 to 0.810 at 78.0% coverage and 0.844 at 61.4% coverage—post hoc operating points computed on the pooled out-of-fold predictions; deployment thresholds must be fixed on independent data—and SigLIP-FT attains the lowest area under the risk–coverage curve (AURC 0.126 versus 0.199 and 0.278). Residual errors are highly structured: 87.0% lie within the negative-emotion triad, and Fear is the hardest category for all three architectures. All findings are established on this single corpus; their transfer to other collections is an open question. Coverage–performance behavior, rather than a single full-coverage score, is the appropriate reporting standard for ambiguous visual domains of this kind. Full article

Rapid urban population aging and increasing land constraints pose significant challenges for improving both service coverage and spatial equity in elderly care facility planning. This study develops an integrated optimization framework that simultaneously addresses accessibility, efficiency, and equity in the spatial allocation of elderly care facilities. First, an improved Gaussian Two-Step Floating Catchment Area (G2SFCA) method is employed to evaluate accessibility patterns across multiple facility types under both walking and driving scenarios. Second, resource allocation equity is quantified using Lorenz curves and spatial Gini coefficients to identify mismatches between elderly care supply and population demand. Building upon these analyses, a fairness-oriented maximum covering location model—termed the Equity Maximum Covering Location Problem (EMCLP)—is formulated and further transformed into a Markov Decision Process. A deep reinforcement learning-based algorithm is subsequently designed to solve the EMCLP under complex spatial constraints. Experiment results demonstrate that the proposed approach achieves improved computational efficiency while maintaining robust solution quality, and effectively enhances service provision in underserved areas through differentiated functional allocation strategies. Full article

The Internet of Things (IoT) market is currently undergoing a period of unprecedented, rapid evolution, leading to the enabling of novel and diverse applications spanning both the civilian and industrial sectors. A significant proportion of these emerging use cases, particularly those in domains such as maritime communications and forestry management, require service continuity and connectivity within geographically remote regions, where conventional terrestrial infrastructure is often absent or economically unfeasible. To bridge this coverage gap and achieve truly ubiquitous connectivity, the recent 3GPP initiative to extend 5G services into Non-Terrestrial Segments (NTNs) holds substantial promise. This expansion is crucial for ensuring that massive Machine-Type Communication (mMTC) services can be reliably provisioned globally. This paper aims to detail the progress in standardization and academic activities towards the design and deployment of NTN-based Narrowband IoT (NB-IoT) systems, which are the leading NTN mMTC enabler in the 3GPP portfolio. We will specify the challenges faced by these systems and outline the solutions proposed thus far. We conclude the paper with a discussion on already operational systems and lessons learned from their deployment and operation. Full article

Accurate air conditioning (AC) load prediction under extreme weather conditions is critical for power grid stability and energy management. While tree-based ensemble methods such as XGBoost, LightGBM, and Gradient Boosting have become the dominant paradigm in short-term load forecasting, their effectiveness for weather-sensitive AC load prediction—particularly during extreme weather events—remains insufficiently examined. This study presents a systematic comparison of four machine learning models—Support Vector Regression with RBF kernel (SVR-RBF), LightGBM, XGBoost, and Gradient Boosting—for daily AC load estimation conditioned on measured same-day meteorological and calendar features. Based on five years of processed load and weather data from a major city in Southwest China, we construct 31 features and evaluate model performance across four scenarios: normal days, weather-extreme days, high-load P85, and high-load P90. Hyperparameters are selected within the first four years by expanding-window validation, and the fifth year is held out for testing. Our results reveal that SVR-RBF achieves an overall $R^2$ of 0.9772, substantially outperforming LightGBM (0.9091), XGBoost (0.8972), and Gradient Boosting (0.9094); paired moving-block bootstrap intervals for the tree-minus-SVR MAE differences exclude zero. The advantage of SVR-RBF is most pronounced under extreme conditions: on weather-extreme days, SVR-RBF attains $R^2=0.8316$ versus $R^2=0.39$ for the best tree model. This pattern is consistent with the smooth U-shaped temperature–load relationship captured by the RBF kernel. Additional trend-sensitivity analysis shows that annual load growth and target-level extrapolation also explain a substantial part of the tree-model degradation. Furthermore, we quantify the performance limitation of the restricted feature set: while normal-day estimation achieves $R^2=0.984$, high-load P90 days reach $R^2=0.843$. These findings support SVR-RBF as a strong baseline for measured-weather conditional AC load estimation while emphasizing that temporal shift and training-domain coverage must be considered when interpreting model differences. For energy-system applications, the lower errors under high-load and weather-extreme conditions are relevant to peak-demand assessment, reserve planning, and demand-side management. Full article

Microplastics are a pervasive contaminant of aquatic, terrestrial, and atmospheric systems, with accumulating evidence of ecological harm and human exposure. Conventional workflows, manual microscopy, FTIR, and Raman spectroscopy, are labor-intensive and operator-dependent, motivating artificial intelligence (AI) and machine learning (ML) as scalable alternatives. Following PRISMA guidelines, five databases were searched; 936 records were screened and 113 primary studies met the eligibility criteria, analyzed through dual-reviewer extraction across seven dimensions. Contrary to the assumption that deep learning dominates, classical and chemometric estimators (46.0% of studies) were at least as prevalent as deep learning (36.3%), with support-vector machines the single most used family (35.4%). Method choice tracked input modality rather than recency: chemometric classifiers such as SVM and PLS dominate spectroscopic data (FTIR, Raman; 53.1% of studies), whereas deep learning concentrates in the image-based minority, where representation learning outperforms manual feature engineering. Detection/identification remained the principal task (65.5%), although 19.5% addressed predictive modeling of sorption, toxicity, distribution, and remediation. Characterization coverage was uneven: origin and polymer type were reported in 88.5% and 72.6% of studies, but size and shape in only 50.4% and 28.3%. Reported accuracies, often exceeding 90%, derive from heterogeneous, non-comparable datasets. Future progress depends on open benchmarks, standardized reporting, and explainable methods. Full article

Reviews of artificial intelligence (AI) for mobile robots usually cover one competence—perception, SLAM, path planning, control, or reinforcement learning—and rarely show how these combine into a working system. We take the opposite view and treat autonomy as one pipeline: sensing, perception, localization and mapping, prediction, planning, visual servoing and control, high-level decision-making, and continual learning. We survey how AI has reshaped each stage for industrial and service robots across Industry 4.0, 5.0, and the emerging Industry 6.0. Using a structured, PRISMA-informed protocol with explicit search strings, inclusion criteria, and cross-embodiment transfer rules, we screen the literature, analyze a corpus drawn mainly from the last five years, and position it against prior surveys with a coverage matrix that exposes their single-block focus. Four findings stand out. Perception and localization approach engineering maturity through multimodal fusion and foundation vision models. Planning and control stay effective but computationally demanding. Decision-making, now driven by large language and vision–language–action models, is powerful yet unverifiable and fails under safety constraints. Lifelong learning is almost absent from deployed systems. The decisive weaknesses sit at the interfaces: at the perception–planning, planning–control, and control–decision handoffs the sim-to-real gap, limited on-robot compute, and scarce industrial data compound. We compare AI families by technology readiness, catalog datasets and benchmarks, examine the safety-certification barrier, and consolidate cross-cutting gaps. We close with a staged roadmap toward Industry 6.0 and a next-generation architecture coupling a foundation perception backbone, a world model and digital twin, a continual-learning memory, and a reasoning core wrapped by a safety monitor. The aim is to move from cataloging algorithms to engineering integrated autonomy. Full article

Large-scale maritime activity, persistent shipping incidents, and complex marine environments continue to place substantial demands on maritime search and rescue (MSAR). Current MSAR systems do not fully capitalize on the complementary strengths of unmanned aerial vehicles (UAVs) and satellites for collaborative tracking and rescue support. Existing air-space collaboration technologies suffer from two critical limitations: (1) rigid processes, including fixed task allocation, pre-determined path planning without real-time environmental adaptation, and isolated satellite–UAV decision-making, and (2) long task completion cycles, mainly because many methods are adapted to wide-area, long-duration military tracking scenarios. They therefore provide limited support for the dynamic flexibility required in MSAR. This study proposes a Proximal Policy Optimization (PPO)-based air-space collaborative tracking method for maritime moving targets to address these shortcomings and enhance air-space cooperation in MSAR operations. The core implementation of the method includes: (1) integration of target drift forecasting, satellite orbit prediction, UAV task allocation, and path planning into a unified reinforcement learning framework to reduce isolated single-platform decision-making; (2) the adoption of PPO to generate dynamic and flexible air-space collaborative tracking strategies that adjust satellite observation angles and scanning ranges, as well as UAV altitude, speed, and heading according to real-time target, environmental, and platform states; and (3) the design of a multi-dimensional reward function that balances target proximity, energy efficiency, coverage overlap, and inter-platform cooperation to guide strategy optimization. Simulation experiments include system-feasibility verification, baseline-controller comparison, PPO hyperparameter screening, and cross-scenario evaluation. Under idealized communication and payload-matching assumptions, the method enables coordinated tracking of maritime moving targets in simulated MSAR scenarios. In the standardized evaluation, PPO achieved an 11.9% higher mean evaluation episode return, 11.2% lower aggregate UAV energy consumption, and a 9.92-percentage-point greater endurance margin than DDPG. Hyperparameter screening compared candidate learning rates, discount factors, and training budgets, informing the PPO configuration for the subsequent six-scenario evaluation. Across the six controlled scenarios, rewards stabilized after approximately 1400 steps, while action magnitudes varied among regions. These results indicate that the proposed method has potential to enhance air-space collaborative tracking for MSAR decision support. Full article

The popularity of electric vehicles (EVs) is growing at a fast pace, creating a need for the strategic deployment of charging stations (CSs) to provide enough coverage, cost effectiveness, and compliance with grid and urban planning regulations. The deployment of large-scale infrastructure under multiple, often conflicting constraints remains a challenging engineering decision-making problem. In this paper, we propose a hybrid optimization framework that combines greedy initialization with reinforcement learning to efficiently explore the charging station deployment problem. The proposed approach employs Q-learning and Deep Q-Network (DQN) agents to iteratively refine the initial deployment while simultaneously optimizing deployment cost, charging demand coverage, and operational utility under practical planning constraints. The constraints include grid capacity limitations, renewable energy utilization, and fairness considerations. The proposed framework is evaluated in realistic urban scenarios. The experimental results demonstrate that the reinforcement learning (RL) approach achieves superior trade-offs among competing objectives compared to baseline heuristic strategies, while maintaining computational scalability for large candidate location sets. The proposed framework demonstrates stable performance across three evaluated deployment scenarios, indicating its potential applicability to increasingly complex charging infrastructure planning problems. The proposed methodology is scalable to other complex engineering planning and resource allocation problems characterized by multi-objective trade-offs and dynamic constraints. Beyond improving optimization performance, the proposed framework contributes to sustainable transportation planning by supporting the efficient deployment of electric vehicle charging infrastructure. Optimized charging station placement promotes greater accessibility to charging services, encourages electric vehicle adoption, reduces unnecessary travel associated with charging activities, and contributes to lower greenhouse gas emissions. Consequently, the proposed methodology provides decision-makers with a scalable and intelligent planning tool that supports the transition toward more sustainable and energy-efficient urban mobility systems. Full article