Model predictive control (MPC) is a powerful technique that can be used for various applications, including trajectory tracking. However, MPC requires an accurate model of the system and a long enough prediction horizon to achieve good performance. Although variations of MPC like stochastic MPC and robust MPC can alleviate the problem to a certain degree, they still more or less rely on the knowledge of system models, disturbance and noise. MPC can also suffer from high computational cost or loss of recursive feasibility, particularly in the case of the existence of model mismatch (actual system model is different from the model used by the controller). Iterative learning control (ILC) is another technique that can be used for improving system performance (reducing tracking error) for repetitive tasks. ILC is known for its capability of learning from previous history (iterations of tasks) and less reliance on the model accuracy. In this paper, both MPC and different ILC algorithms are employed for a trajectory tracking problem with model mismatch. The performances of different algorithms are then analyzed and compared. A criterion on the degree of model mismatch to guarantee error convergence is also discussed for model-based ILC.

Clinical decision-making often requires physicians to iteratively generate hypotheses, gather multimodal evidence, and refine diagnoses under conditions of time pressure and information overload. While medical vision–language models (VLMs) have shown promise in visual question answering, report generation, and image-based diagnosis, most remain standalone tools without integration into broader diagnostic workflows. In this work, we introduce Casidence, an agentic medical VLM framework that unifies evidence retrieval, similar case retrieval, and conversational AI into a single decision-support platform. At its core, Casidence incorporates a fine-tuned 3D medical VLM trained on the CT-RATE dataset to achieve state-of-the-art performance on volumetric imaging tasks. To enable robust similar case retrieval, we propose a novel Query Auto Encoder (QAE) that disentangles semantic medical content from surface linguistic variation, producing compact embeddings aligned across paraphrased reports. Together, these components allow Casidence to operationalize evidence-based reasoning: planning and executing tool-augmented workflows, curating structured evidences, and generating auditable diagnostic outputs. Quantitative and qualitative evaluations demonstrate that Casidence improves planning transparency, retrieval fidelity, and report quality over strong baselines. By grounding model outputs in clinical evidence and supporting iterative human–AI collaboration, Casidence represents a step toward trustworthy, workflow integrated medical agentic systems.

A hot topic in current computer vision research is image generation, which essentially refers to the use of computer algorithms and models to generate artistic and creative images. With the development of artificial intelligence and deep learning, numerous image generation models have emerged, and image generation technology is gradually becoming an important research direction in the fields of computer graphics and art. This has driven the widespread application of image generation in areas such as film, gaming, and painting. This paper provides a preliminary introduction to the basic principles of image generation models, including popular models, their evolution and variants. It then discusses some of the application areas in which image generation can be used. Finally, by analysing the challenges and issues faced by image generation models, this paper looks ahead to some of the work that future researchers will undertake and the direction in which image generation models will develop.

This paper presents a smart trash-can prototype that tackles the inefficiency of manual waste separation by embedding low-cost infrared sensors above a rotating, multi-compartment drum. Each discarded object is first scanned for its unique thermal signature; a micro-controller then identifies the material, spins the drum to the correct partition and drops the item without human contact. Colored slight-emitting diode (LED) and short audio cues give users real-time feedback, turning sorting into an intuitive game. Controlled trials with common household samples—plastic bottles, banana peels, aluminium cans—achieved high classification accuracy and rapid response time, outperforming both conventional bins and untrained volunteers. Power draw remains modest, allowing long operation on a small battery. The study demonstrates that infrared-based automation can deliver reliable, low-maintenance sorting at the point of disposal, offering municipalities a practical, scalable route to cleaner streets, higher recycling rates and more sustainable cities. Its open blueprint lets any community replicate and refines the loop, turning every alley into a quiet driver of the circular economy.

To address the limitations of existing Olympic medal prediction models—which over-rely on macroeconomic indicators and lack interpretability—this study proposes a hybrid machine learning framework integrating multi-dimensional feature engineering and model interpretability. This article construct a 19-feature system incorporating historical medals, athlete gender ratios, elite coach attributes (simulated via Monte Carlo), and host-nation effects. Three synergistic models are developed: (1) A multiple linear regression predicting 2028 medal distributions (RMSE=0.25); (2) A LASSO-Logistic regression identifying first-time medal-winning nations (e.g., UAE and Samoa, probability >0.5); (3) A Gradient Boosting Tree with SHAP interpretability quantifying elite coaches’ contribution (SHAP mean=0.217, *p*<0.001). Hyperparameter optimization via RandomizedSearchCV achieves high accuracy (gold medal MAE=0.09, total medals MAE=0.25). Key innovations include dynamic feature fusion and micro-level coach impact quantification, providing actionable insights for Olympic resource allocation.

In recent years, 3D printing technology has developed rapidly, with applications spanning traditional manufacturing, aerospace, and medical fields. Mainstream technologies include Binder Jetting, DED, and Power Bed Fusion. This paper discusses the principles, characteristics, and application cases of these three technologies. For example, Binder Jetting can be used to form silk powder combined with a binder, DED can achieve rapid manufacturing and repair of large metal components, and Power Bed Fusion can generate internal lattice structures that cannot be achieved with traditional processes. Additionally, the paper introduces applications of 3D printing in medical surgery planning, aerospace component lightweighting, and industrial material waste reduction. It highlights challenges such as quality stability, high certification costs, and insufficient material strength, emphasizing the trend toward transitioning from rapid prototyping to full-scale manufacturing technology. It concludes that as precision and material performance improve, 3D printing technology will leverage material innovation and process optimization to demonstrate breakthrough potential beyond traditional manufacturing across multiple fields, establishing a core position in high-end manufacturing and personalized applications.

In the field of high-dimensional dynamic system modeling, achieving high-precision prediction of target variables, interpretability of model structures, and quantification of output uncertainty has always been a key focus of intelligent system research. Traditional methods encounter bottlenecks in terms of feature heterogeneity, multi-dimensional interactions, and prediction credibility. This paper proposes a multi-objective sequence modeling framework that integrates generative adversarial learning, simulated annealing optimization, and a game-theoretic interpretation mechanism. This framework first cleans, standardizes, and imputes missing values in the original structured data, and then reconstructs the input tensor using a BP neural network. Subsequently, a prediction model based on GAN is constructed, and SA is used for global optimization of key hyperparameters. During the prediction phase, a Monte Carlo perturbation strategy is introduced to quantify uncertainty. Finally, the importance of features is analyzed based on the SHAP method to achieve causal transparency. This method outperforms mainstream baseline models such as LSTM and Transformer in multiple metrics, demonstrating good generalization ability and interpretability. It is suitable for prediction and auxiliary decision-making in complex engineering systems in fields such as finance and industry, and has broad application prospects and significant practical value.

This study proposes an integrated ARIMA–XGBoost–LightGBM model to predict both medal counts and medal-winning probabilities for the 2028 Los Angeles Olympic Games. The framework begins by applying the ARIMA model to forecast time-series features, such as the number of athletes and participating countries. These features are then fed into an XGBoost regressor to estimate gold, silver, bronze, and total medals per country. Additionally, a LightGBM classifier is utilized to predict the probability that a nation will win at least one medal. Model performance was evaluated using ten-fold cross-validation, with R² values exceeding 0.84 across all medal categories, demonstrating high accuracy and robust generalization ability. Notably, six countries are projected to win their first Olympic medals based on the predicted probabilities. This ensemble approach effectively combines time series forecasting with machine learning algorithms, showcasing its potential in supporting strategic sports planning and medal outcome prediction. It provides valuable forecasting capabilities for complex, dynamic events such as the Olympics.

High-angle deviated wells have a larger contact area with the reservoir compared to conventional vertical wells. As extraction and injection wells, they can significantly enhance extraction efficiency. Additionally, an integrated geological engineering design allows for large platform implementation, reducing surface construction costs and increasing overall project returns. However, due to the inter-layer and intra-layer heterogeneity of sandstone types thermal storage, high-angle deviated wells are prone to issues such as uneven vertical utilization and cold breakthrough in high-permeability layers. Moreover, predicting water extraction temperatures becomes more challenging. This paper is based on implemented project examples to optimize the design of high-angle deviated geothermal wells and screen casing. The integrated geological engineering approach aims to reduce costs and increase efficiency by optimizing well trajectories considering inter-layer heterogeneity, ensuring effective water extraction deeply. Additionally, we establish a QS flow equation and use the harmonic level KH method for temperature prediction, achieving a 10% increase in thermal storage thickness utilization with measurement and temperature errors within 3%. A modeling system for thermal storage temperature fields throughout the lifecycle is developed, providing a solid foundation for parameter optimization in project planning. Real-time monitoring of field operations is conducted to provide early warnings for cold breakthrough risks, ensuring stable operation of public welfare projects.

With the deepening of the "dual carbon" goals (carbon peak and carbon neutrality), ensuring inter-provincial carbon emission equity alongside improving carbon emission reduction efficiency has become crucial. This paper uses panel data from 30 provinces in China from 2006 to 2023. Employing the multi-period difference-in-differences (DID) method and the spatial Durbin model, we investigate the impact of carbon trading markets on inter-provincial carbon emission equity. The results show that the implementation of carbon trading pilot policies significantly enhances inter-provincial carbon emission equity. Mechanism analysis further reveals that carbon trading pilot policies significantly improve inter-provincial carbon emission equity through industrial structure adjustment. Furthermore, heterogeneous effects on inter-provincial carbon emission equity exist across different regions and levels of income distribution. Further analysis indicates that the impact of carbon trading pilot policies on inter-provincial carbon emission equity exhibits a positive spatial spillover effect, with a negative "siphon effect" observed among pilot regions. Finally, policy recommendations are proposed based on the research findings. This study is of significant importance for enriching research in the field of carbon emission equity and promoting the establishment of a nationwide unified carbon market.