Search Results (743)

Accurate prediction of a ship’s Estimated Time of Arrival (ETA) is of great significance for port scheduling, logistics management, and navigation safety. Traditional ETA prediction approaches often rely on manual experience for parameter tuning, which tends to be inefficient and susceptible to subjective factors. To address this issue and improve prediction accuracy, this study proposes a hybrid modeling framework, integrating Bayesian Optimization (BO), Convolutional Neural Networks (CNNs), and Long Short-Term Memory (LSTM) networks. In this approach, Automatic Identification System (AIS) data is leveraged to predict the total voyage duration before departure, thereby deriving the vessel’s ETA. The model, referred to as BO-CNN-LSTM, utilizes BO for automatic hyperparameter tuning, employs CNN for extracting local features, and applies LSTM network to capture temporal dependencies. The model is developed using a dataset of 32,972 distinct voyage records, among which 23,947 are retained as valid samples after data cleaning. Pearson correlation analysis is conducted to select key input variables, including navigation speed, ship type, sailing distance, and deadweight tonnage. Additionally, sailing distance is processed using the Ramer–Douglas–Peucker algorithm. Experimental evaluation indicates that the BO-CNN-LSTM model achieves a coefficient of determination of 0.987, along with a mean absolute error and root mean square error of 6.078 and 8.730, respectively. These results significantly outperform comparison models such as CNN, LSTM, CNN-LSTM, random forest, AdaBoost, and Elman neural networks. Overall, this study validates the effectiveness and superiority of the proposed BO-CNN-LSTM model in ship ETA prediction, providing an efficient and effective prediction solution for intelligent maritime transportation systems. Full article

In the field of engineering, the advancement of automated scoring systems for laboratory reports has been significantly hampered by three persistent challenges: scarcity of high-quality annotated data, high domain-specific complexity, and insufficient model interpretability. To address these limitations, this study proposes an AdaBoost regression model based on multi-level feature engineering and threshold control, denoted as MFTC-ABR. This method constructs a multi-dimensional feature set using a lightweight neural network, which evaluates laboratory reports across four core dimensions: comprehension of experimental principles, completion of experimental procedures, depth of result analysis, and plagiarism detection. At the scoring algorithm level, a dynamic threshold adjustment mechanism is integrated into the AdaBoostReg ensemble learning framework. By redesigning the sample weight update rule, the prediction errors of samples are divided into three intervals: the acceptable region, the stable learning range, and the focus range. Accordingly, a differentiated weight update strategy is implemented, and a history-aware mechanism is introduced to further regulate the attention allocated to individual samples. Finally, experimental results on the power electronics laboratory report dataset show that MFTC-ABR model achieves a mean absolute error (MAE) of 3.09 and a scoring consistency rate of 82% within a five-point error tolerance. These findings validate the effectiveness and practicability of the proposed method for automatic assessment in specialized domains with limited data availability. Full article

The present paper proposes an adaptive steering weight allocation strategy based on a non-cooperative Stackelberg game and Model Predictive Control (MPC) for dynamic steering authority allocation in human–machine shared control of intelligent vehicles. First, the human–machine steering interaction is modelled as a Stackelberg game, and the steering control problem is formulated as an MPC optimization problem. The optimal control sequences of the driver and the Advanced Driver Assistance System (ADAS) under game equilibrium are then derived through backward induction. Subsequently, driver behaviour is classified as aggressive, moderate, or conservative according to lateral preview error and lateral acceleration, and the driver state is quantified using parametric indicators. Furthermore, by integrating potential field-based driving risk assessment with human–machine conflict intensity, a fuzzy logic-based dynamic weight adjustment mechanism is constructed. Simulation results show that when the steering intentions of the driver and the ADAS are highly consistent, the proposed strategy can effectively reduce driver workload and improve driving safety. In high-risk driving situations, the strategy automatically transfers more steering authority to the ADAS to enhance safety, whereas under low-risk conditions with strong human–machine steering conflict, greater driver authority is preserved to ensure that the vehicle follows the intended path. Hardware-in-the-loop experiments in lane-changing assistance scenarios further verify the effectiveness of the proposed strategy under different driving styles. Quantitative results show that, compared with manual driving, the proposed strategy reduces the maximum lateral overshoot by 98.75%, 85.54%, and 98.58% for aggressive, moderate, and conservative drivers, respectively. In addition, the peak yaw rate and driver control effort are significantly reduced, indicating smoother vehicle dynamic response and lower steering workload. These results demonstrate that the proposed strategy can effectively improve lane-change stability, reduce driver burden, and maintain safe and coordinated human–machine shared control. Full article

Advanced Driver Assistance Systems (ADAS) are rapidly evolving to improve driving safety and comfort, in line with the European Community’s Vision Zero objectives. However, current validation methods mainly focus on steady-state conditions and regulatory compliance, while standardized KPIs capable of capturing vehicle dynamics during transient maneuvers and relating them to human driving behavior are still lacking. This study proposes a KPI-based methodology to evaluate lane centering systems in both steady-state and transient conditions, using a human driver as a reference. New parameters—aggressiveness and smoothness indices—are introduced to quantify the dynamic response. Simulator and constant-speed vehicle tests highlight the differences between ADAS strategies and human behavior, providing insights into comfort and acceptance. Full article

In highly automated driving systems (ADSs), drivers’ ability to resume manual driving remains a road safety issue. However, to the best of our knowledge, there is no existing computational model to predict which drivers require more than the 4 seconds mandated by United Nations Regulation No. 157 to regain manual control. To address this challenge, we developed a Random Forest model that predicts takeover time using measurable human factors. Three controlled driving simulator experiments were conducted in which participants engaged in distinct tasks—texting, drinking, and traffic monitoring—before responding to a takeover request. During the experiments, we collected human factor features, including gaze behavior, age, and scores, from the self-reported driving behavior questionnaire (K-DBQ). The Random Forest classifier achieved 77% accuracy. Recursive feature elimination selected 10 dominant predictors; notably, engaging in non-driving-related tasks, reduced on-road gaze, and older age were significantly associated with longer takeover times. Although K-DBQ scores were not directly correlated with takeover time, their inclusion improved model robustness, consistent with ensemble learning from weak yet complementary signals. The proposed model can be integrated into advanced driver assistance systems (ADASs) to proactively identify drivers likely to exceed the 4-second takeover window, support targeted interventions, and enhance human-centered transition safety in ADSs. Full article

The cryptocurrency market has attracted significant attention from global investors, with Cardano (ADA) ranking among the top cryptocurrencies by market capitalization. However, predicting ADA returns remains challenging due to the complex, multi-scale dynamics influenced by Federal Reserve policies, geopolitical events, and high-frequency trading. This study proposes a “Sliding EMD–Multi Variables” framework for cryptocurrency return prediction, leveraging Empirical Mode Decomposition’s multi-scale fractal properties to capture nonlinear dynamics at different time scales. The sliding window decomposition method addresses data leakage issues while incorporating key economic and policy variables at the component level. The empirical results demonstrate that the Sliding EMD system significantly outperforms univariate and multivariate benchmarks. Compared to the univariate system, it improves MSE, RMSE, SMAPE, and DSTAT by 0.83%, 0.42%, 5.23%, and 0.43%, respectively, while enhancing investment metrics (maximum drawdown, Sharpe ratio, Sortino ratio, Calmar ratio) by 0.19, 0.36, 0.95, and 0.15. Against the multivariate system, improvements reach 5.52%, 3.14%, 5.74%, and 17.62% in prediction accuracy, with investment performance gains of 0.47, 1.69, 4.27, and 0.31. Incorporating economic variables at the component level yields additional improvements of 0.94%, 0.47%, and 0.78% in MSE, RMSE, and MAE. These findings offer valuable insights for cryptocurrency portfolio optimization using fractal-based decomposition methods. Full article

Graph Neural Networks (GNNs) have emerged as the standard for learning representations from graph-structured data. While traditional architectures relying on message-passing mechanisms excel in homophilic settings, they essentially function as fixed low-pass filters. However, this smoothing operation limits their ability to generalize to heterophilic graphs, where connected nodes often exhibit dissimilar labels and high-frequency signals are crucial for discrimination. Furthermore, existing Mixture-of-Experts (MoE) methods for graphs often suffer from local-view routing, failing to capture global structural context during expert selection. To address these challenges, this paper proposes SS-AdaMoE, a novel Spatio-Spectral Adaptive Mixture of Experts framework designed for robust node classification across diverse graph patterns. Specifically, a Dual-Domain Expert System is constructed, integrating heterogeneous spatial aggregators with learnable spectral filters based on Bernstein polynomials. This allows the model to adaptively capture arbitrary frequency responses—including high-pass and band-pass signals—which are overlooked by standard GNNs. To resolve the locality bias, a Hierarchical Global-Prior Gating Network augmented by a Linear Graph Transformer is introduced, ensuring that expert selection is guided by both local node features and global topological awareness. Extensive experiments are conducted on five benchmark datasets spanning both homophilic and heterophilic networks. The results demonstrate that SS-AdaMoE consistently outperforms baselines, achieving accuracy improvements of up to 2.65% on Chameleon and 1.41% on Roman-empire over the strongest MoE baseline, while surpassing traditional GCN architectures by margins exceeding 28% on heterophilic datasets such as Texas. These findings validate that the synergy of learnable spectral priors and global gating effectively bridges the gap between spatial aggregation and spectral filtering. Full article

Precise regulation of the postharvest storage environment is critical for reducing losses and maintaining potato quality. Semi-underground storage facilities are widely used in major potato-producing regions of northern China; however, pronounced spatiotemporal heterogeneity in the internal temperature field often leads to localized quality deterioration. To enable accurate sensing and proactive prediction of temperature dynamics in such facilities, this study investigated a typical semi-underground potato storage cellar in Wuchuan County, Inner Mongolia. A high-density sensor network was deployed to collect temperature data, and the spatiotemporal variation patterns of the internal temperature field were systematically analyzed. The results indicate that, at the same vertical height, spatial temperature gradually increases from the entrance toward the interior of the cellar. Both the maximum and minimum temperatures in the entrance zone are lower than those in other regions, while the highest temperatures are observed near the rear wall. Based on the collected data, hierarchical clustering was employed to partition the internal temperature field into three spatiotemporal pattern clusters with significant differences. Key representative monitoring locations were then identified using the Spearman correlation coefficient. An AdaBoost-based prediction model was subsequently developed to estimate the temperatures at other test locations within each cluster using measurements from the representative points. The results demonstrate that the proposed model maintains high prediction accuracy while substantially reducing dependence on a dense sensor network. The overall MAE ranges from 0.075 to 0.373 °C, and the sensor reduction ratio reaches 87%. This approach provides a paradigm for low-cost intelligent monitoring and offers theoretical support and decision-making guidance for the smart regulation of potato storage environments. By optimizing the monitoring of potato storage environments, this study can reduce monitoring system costs and resource consumption, providing technical support for building a sustainable potato supply chain and delivering significant economic benefits in promoting the development of a resource-conserving potato industry. Full article

This study evaluates the feasibility of using two affordable thermal cameras (UNI-T UTi260M and UTi260T), which are not designed as automotive sensors, for observing pedestrians and warm objects during night-time driving under low-illumination conditions. The experimental setup includes mounting the camera on the vehicle body (e.g., side mirror area/roof), recording road scenes in urban and rural environments, and selecting representative frames for qualitative and quantitative analysis. The study assesses: (i) observable pedestrian detectability in unlit road sections and under oncoming headlight glare, where visible cameras often lose contrast; (ii) the influence of low ambient temperature and strong cold wind on image appearance (including “whitening”/contrast shifts); and (iii) workflow differences, where UTi260M relies on a smartphone application for streaming/recording, while UTi260T supports PC-based image analysis and temperature-profile visualization. In addition, a calibration-based geometric method is proposed for approximate pedestrian distance estimation from single frames using silhouette pixel height and a regression model based on 1$/h_{px}$, valid for a specific mounting configuration and a known subject height. Results indicate that both cameras can highlight warm objects relative to the background and support visual pedestrian identification at low illumination, including in the presence of oncoming headlights, with UTi260M showing more stable behavior in parts of the tests. This work is a feasibility study and does not claim Advanced Driver Assist Systems (ADAS) functionality; it outlines limitations, repeatability considerations, and a minimal set of metrics and procedures for future extension. All quantitative indicators derived from exported frames are explicitly treated as image-level proxy metrics, not as physical sensor characteristics. Full article

This paper evaluates the current maturity of automatic code-generation workflows for deploying modern CNN-based object detectors on embedded GPU platforms. We compare a native pipeline against a code generation pipeline through a Model-Based Engineering (MBE) approach, using YOLOv8/YOLOv9 inference on NVIDIA Jetson Orin Nano and Jetson AGX Orin as representative edge-GPU workloads. We report detection-quality metrics (mAP, PR curves) and system-level metrics (latency distribution and initialization overhead) under a controlled single-class scenario based on a CARLA-generated sequence with frame-level annotations. Absolute accuracy and latency values are scenario-dependent and may vary under different camera optics, illumination, motion blur, sensor noise, occlusion patterns, and multi-class scene. Results quantify the performance gap between code generation and native pipelines and show that, for the evaluated workloads, the automated pipeline remains less competitive in both latency and accuracy. We discuss the implications of this gap for deployment workflows in safety-oriented domains, and we outline bottlenecks that should be addressed. The study is intended as a controlled traffic-light detection micro-benchmark and does not aim to validate full ADAS perception stacks. Full article

Colonial historic city centers represent a paradigmatic challenge for modern road safety, as they are characterized by narrow streets originally designed for carriage and pedestrian traffic. This research presents ADAS-TSR, a deep learning-based advanced driver assistance system for vertical traffic sign detection with voice alerts, specifically designed for the Historic Center of Ayacucho, Peru, which is located at 2761 m a.s.l. An original dataset comprising 2250 images with 2450 instances corresponding to 14 sign classes according to Peruvian regulations was constructed. The dataset was captured under real operational conditions, including deteriorated, partially occluded, and vehicle impact-deformed signage. A comprehensive multi-model benchmark experiment was conducted, comparing four CNN-based detectors (YOLOv8m, YOLO11n, YOLO26n, YOLO26s) and one transformer-based detector (RT-DETR-l) spanning both classical and state-of-the-art architectures released through January 2026. YOLO26s achieved the best overall performance, with an mAP@0.5 of 0.994 and mAP@0.5:0.95 of 0.989 while using only 9.5 M parameters. YOLO11n matched the performance of YOLOv8m with 10× fewer parameters (2.6 M vs. 25.9 M). Uncertainty analysis revealed that modern architectures exhibit significantly higher prediction confidence (mean > 0.90) compared to YOLOv8m (0.82), and fairness analysis confirmed equitable detection across all 14 classes (Gini < 0.002). A voice alert system with five priority levels and rule-based temporal filtering for detection stabilization was implemented. Validation across five urban circuits spanning 14.11 km demonstrated a detection rate of 94.7% with a 73% reduction in redundant alerts. Full article

The integration of renewable energy and power electronics has intensified the occurrence of complex power quality disturbances (PQDs), which increasingly threaten grid stability. To address the challenges of multi-class PQD identification under noisy conditions, this paper proposes a novel framework that combines an enhanced time–frequency analysis method with an optimized AdaBoost decision tree. The main contributions are three-fold: (1) We develop an instantaneous frequency adaptive Fourier synchrosqueezing transform (IFAFSST) equipped with a custom adaptive operator that aligns closely with the frequency modulation patterns in PQD signals, thereby improving time–frequency energy localization. (2) The IFAFSST outputs are decomposed into low-frequency and high-frequency components, from each of which a set of 16 discriminative features is extracted. (3) An improved AdaBoost classifier is introduced, incorporating forward feature selection and Hyperband-based hyperparameter optimization to enhance classification performance. Hyperband accelerates the optimization process by dynamically allocating computing resources and iteratively eliminating suboptimal configurations, thereby enabling efficient determination of the optimal hyperparameters. The method proposed in this paper achieved an accuracy rate of 99.50% on simulated data containing 30 dB white noise and 98.30% on hardware platform data. This framework can effectively handle 23 types of interference, including seven types of single interference, 12 types of double compound interference, three types of triple compound interference, and one type of quadruple compound interference. It performs particularly well in identifying composite interference scenarios. This research has made a significant contribution to power quality analysis, providing a powerful solution with high accuracy and practical applicability, and offering great potential for the implementation of smart grid monitoring systems and the integration of renewable energy. Full article

Background and Objectives: Diabetes mellitus represents one of the most prevalent chronic metabolic disorders worldwide, necessitating precise insulin dose management to prevent both acute and long-term complications. The optimization of insulin dosing remains a significant clinical challenge, as inappropriate dosing can lead to hypoglycemia or hyperglycemia, each carrying substantial morbidity risks. Machine learning approaches have emerged as promising tools for developing clinical decision support systems; however, their practical implementation requires both high predictive accuracy and model interpretability. This study aimed to develop and evaluate an explainable machine learning framework for predicting insulin dose adjustments in diabetic patients. We sought to compare multiple ensemble learning approaches and identify the optimal model configuration that balances predictive performance with clinical interpretability through comprehensive SHAP and LIME analyses. Materials and Methods: A comprehensive dataset comprising 10,000 patient records with 12 clinical and demographic features was utilized. We implemented and compared nine machine learning models, including gradient boosting variants (XGBoost, LightGBM, CatBoost, GradientBoosting), AdaBoost, and four ensemble strategies (Voting, Stacking, Blending, and Meta-Learning). Model interpretability was achieved through SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) analyses. Performance was evaluated using accuracy, weighted F1-score, area under the receiver operating characteristic curve (AUC-ROC), precision-recall AUC (PR-AUC), sensitivity, specificity, and cross-entropy loss. Results: The Meta-Learning Ensemble achieved superior performance across all evaluation metrics, attaining an accuracy of 81.35%, weighted F1-score of 0.8121, macro-averaged AUC-ROC of 0.9637, and PR-AUC of 0.9317. The model demonstrated exceptional sensitivity (86.61%) and specificity (91.79%), with particularly high performance in detecting dose reduction requirements (100% sensitivity for the ‘down’ class). SHAP analysis revealed insulin sensitivity, previous medications, sleep hours, weight, and body mass index as the most influential predictors across different insulin adjustment categories. The meta-model feature importance analysis indicated that LightGBM probability estimates contributed most significantly to the ensemble predictions. Conclusions: The proposed explainable Meta-Learning Ensemble framework demonstrates robust predictive capability for insulin dose adjustment recommendations while maintaining clinical interpretability. The integration of SHAP-based explanations facilitates clinician understanding of model predictions, supporting transparent and informed decision-making in diabetes management. This approach represents a significant advancement toward the clinical implementation of artificial intelligence in personalized insulin therapy. Full article

A support vector machine based on AdaBoost algorithm (SVM-AB) is proposed for complicated underground mine tunnel modeling. This method accurately predicts the nonlinear propagation characteristics of electromagnetic waves in complex environments in the case of small samples. Firstly, an electromagnetic wave propagation loss model is established by analyzing complex factors including tunnel geometry, wall roughness, tilt, dielectric properties, and multipath effects. Secondly, the complex factors and measured signal strength serve as inputs of the SVM model to establish a nonlinear mapping for preliminary prediction. Furthermore, the AdaBoost algorithm is applied to dynamically correct the SVM prediction errors, further enhancing accuracy. Finally, the measured experiments are carried out in complex underground mine tunnels to verify the proposed theoretical model. The experimental results demonstrate that the proposed SVM-AB model achieves a fitting accuracy of over 99.92%. In addition, compared with the traditional support vector machine, its Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) are reduced by about 84.76% and 92.61%, respectively. The proposed tunnel model has important application value for optimizing the layout of communication system of underground mine tunnel. Full article

The development of AI-driven automated driving functions requires vast amounts of diverse, high-quality data to ensure road safety and reliability. However, both the manual collection of real-world data and creation of 3D environments are costly, time-consuming, and hard to scale. Most automatic environment generation methods still rely heavily on manual effort, and only a few are tailored for Advanced Driver Assistance Systems (ADAS) and Automated Driving Systems (ADS) training and validation. We propose an automated generative framework that learns ground-truth features to reconstruct 3D environments from a road definition and two simple parameters for country and area type. Environment generation is structured into three modules—map-based data generation, semantic city generation, and final detailing. The overall framework is validated by training a perception network on a mixed set of real and synthetic data, validating it solely on real data, and comparing performance to assess the practical value of the environments we generated. By constructing a Pareto front over combinations of training set sizes and real-to-synthetic data ratios, we show that our synthetic data can replace up to $85\%$ of real data without significant quality degradation. Our results demonstrate how multi-layered environment generation frameworks enable flexible and scalable data generation for perception tasks while incorporating ground-truth 3D environment data. This reduces reliance on costly field data and supports automated rapid scenario exploration for finding safety-critical edge cases. Full article