Artificial Intelligence Review最新文献

{"title":"A review on safe reinforcement learning using Lyapunov and barrier functions","authors":"Dhruv Kushwaha,&nbsp;Zoleikha Biron","doi":"10.1007/s10462-026-11611-9","DOIUrl":"10.1007/s10462-026-11611-9","url":null,"abstract":"<div><p>Reinforcement learning (RL) has proven to be particularly effective in solving complex decision-making problems for a wide range of applications. From a control theory perspective, RL can be considered as an adaptive optimal control scheme. Lyapunov and barrier functions are the most commonly used certificates to guarantee system stability for a proposed/derived controller and constraint satisfaction guarantees, respectively, in control-theoretic approaches. However, compared to theoretical guarantees available in control-theoretic methods, RL lacks closed-loop stability of a computed policy and constraint satisfaction guarantees. Safe reinforcement learning refers to a class of constrained problems where the constraint violations lead to partial or complete system failure. The goal of this review is to provide an overview of safe RL techniques using Lyapunov and barrier functions to guarantee this notion of safety (stability of the system in terms of a computed policy and constraint satisfaction during training and deployment). Three concrete takeaways emerge from our analysis: (i) the field has shifted decisively from model-based to model-free formulations since 2017, with combined CLF–CBF approaches becoming the most active sub-area post-2022; (ii) per-class open problems are now well-defined, certificate validity under function approximation and distribution shift for Lyapunov methods, feasibility and deadlock under hard CBF–QP shielding for barrier methods, and joint CLF–CBF feasibility under model uncertainty for combined methods; and (iii) deployment to high-dimensional and partially observable settings remains the dominant scalability barrier across all three classes. The different approaches employed are discussed in detail along with their shortcomings and benefits to provide critique and possible future research directions. The review demonstrates promising scope for providing safety guarantees for complex dynamical systems with operational constraints using model-based and model-free RL.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 11","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-06-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11611-9.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149035201","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Recent advances in AI-based mobile robots for human companionship: survey","authors":"Ahmed Abdu,&nbsp;Redhwan Algabri,&nbsp;Riyadh Nazar Ali Algburi,&nbsp;Lilia Tightiz,&nbsp;Nabih Pico,&nbsp;Ammar Hawbani,&nbsp;Ali Abdo,&nbsp;Francisco Yumbla,&nbsp;Saeed Hamood Alsamhi,&nbsp;Ji-Hun Bae,&nbsp;Sungon Lee","doi":"10.1007/s10462-026-11603-9","DOIUrl":"10.1007/s10462-026-11603-9","url":null,"abstract":"<div><p>Human companionship is an essential capability for mobile robots operating in dynamic, human-centered environments. It enables robots to perform tasks such as guidance, assistance, surveillance, and service delivery across various domains, including healthcare, logistics, and public safety. The recent advances in artificial intelligence (AI), particularly in computer vision, deep learning, and sensor fusion, have significantly improved the reliability, adaptability, and contextual understanding of robotic companionship systems. This review presents a comprehensive analysis of recent advances in AI-based approaches to human companionship by mobile robots. It focuses on critical components such as sensing modalities, human detection and re-identification, trajectory prediction, motion planning, control strategies, and human-robot interactions. It presents a structured taxonomy of current methods, highlights commonly used control architectures, and explores the key challenges involved in real-world deployment. Despite significant progress, several issues remain unresolved, including occlusion handling, identity switching, socially compliant navigation, and scalability. The review concludes by identifying prominent research gaps and proposing future directions aimed at advancing the development of robust, socially aware, and context-adaptive robotic companions.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 11","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-06-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11603-9.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149035200","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Zeroing neural network for optimization: A survey of theory and applications","authors":"Ruixuan Chen,&nbsp;Zanyu Tang","doi":"10.1007/s10462-026-11628-0","DOIUrl":"10.1007/s10462-026-11628-0","url":null,"abstract":"<div><p>Zeroing neural network (ZNN) provides a dynamical-systems approach for solving time-varying optimization problems. This survey revisits ZNN from an optimization-oriented perspective. First, recent advances in ZNN are reviewed, including error construction, convergence mechanisms, robustness enhancement, and the relationship between continuous-time models and their discrete-time counterparts. These developments clarify the theoretical foundations underlying the use of error-driven dynamics for tracking time-varying solutions. Next, the survey explains how representative linear and nonlinear optimization problems can be embedded into the ZNN framework. Applications in imaging, signal processing, and robotic control further demonstrate that once inverse problems, estimation tasks, or constrained motion planning problems are formulated as optimization problems, they can be naturally integrated into the ZNN paradigm for real-time computation. Finally, the survey discusses current limitations, emerging opportunities, and future challenges for ZNN in evolving research directions. Overall, ZNN provides a unified and robust pathway linking optimization modeling with real-time dynamic computation across diverse time-varying engineering applications.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 11","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11628-0.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149035252","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Physics-informed deep learning for civil infrastructure: a review framework integrating the PIFI index and P-D-E triad","authors":"Jiahui Zhang,&nbsp;Zoia Vladimirovna Beliaeva,&nbsp;Yixiao Wang,&nbsp;Yue Huang","doi":"10.1007/s10462-026-11617-3","DOIUrl":"10.1007/s10462-026-11617-3","url":null,"abstract":"<div><p>The safety and sustainability of civil infrastructure are facing increasingly severe challenges, creating an urgent need for smarter, more reliable data-driven methods to support its lifecycle management and decision-making. Existing deep learning methods face key bottlenecks in civil engineering applications, including data scarcity, insufficient model credibility, high computational costs, and ethical compliance issues. Physics-informed deep learning, by embedding physical knowledge, offers a new paradigm for overcoming these limitations. This paper first systematically reviews the research progress of physics-informed deep learning in the field of civil engineering, followed by the proposal of an original evaluation framework. Based on a systematic analysis of 243 publications (2000–2025), it explores in depth the technological evolution, engineering adaptability, and inherent limitations of core architectures such as CNNs, RNNs, and PINNs. On this basis, the paper presents two original contributions: First, the Physics-Informed Fusion Index–which quantifies a model’s comprehensive performance by balancing physical consistency (<span>({E}_{text {physics}})</span>) and predictive accuracy (<span>({D}_{text {accuracy}})</span>) through weighting; this represents the first attempt to incorporate the trade-off between physical constraints and data fitting into a unified evaluation metric. Second, the Physics–Data–Engineering Triangular Model–a quantitative selection framework that assesses the applicability of technical solutions across three orthogonal dimensions: physicality, data dependency, and engineering adaptability. Unlike existing reviews that primarily compile methodologies, the proposed PIFI-P-D-E framework traces the trajectory of technological evolution and examines the trade-offs between physical embedding depth, data utility boundaries, and engineering implementation constraints, providing a systematic perspective and decision-making tools for the transition from technological exploration to engineering governance. Building on this foundation, this paper outlines an interdisciplinary development pathway centred on ‘mechanism–data integration’, advocating for a new paradigm in civil informatics.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 11","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11617-3.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149035253","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Trends, challenges and future directions in deep learning for citrus leaf disease recognition: a survey","authors":"Tharindu Godage,&nbsp;Nishankar Sathiyamohan,&nbsp;Selvarajah Thuseethan,&nbsp;Shanmuganathan Vasanthapriyan,&nbsp;Zhongwei Liang","doi":"10.1007/s10462-026-11629-z","DOIUrl":"10.1007/s10462-026-11629-z","url":null,"abstract":"<div><p>Citrus farming is a vital component of global agriculture that faces continuous challenges from pathogens and environmental stressors. A major difficulty for farmers is the early and accurate identification of citrus leaf diseases, which is essential to prevent yield loss and maintain fruit quality. In recent years, researchers have focused on automating the citrus disease recognition process using computer vision and machine learning techniques. More importantly, the advancements in deep learning have substantially accelerated progress in citrus leaf disease recognition by enabling robust and scalable feature extraction from complex imagery. Despite these advancements, considerable challenges remain in automatic citrus disease detection, and researchers continue to strive for improved generalisation, in-field adaptability, and suitability for low-power agricultural devices. Distinguishing itself from existing reviews, this survey provides a unique, taxonomically aligned evaluation that addresses the visual mimicry between pathogenic symptoms and pest-induced damage. This survey provides a comprehensive overview of benchmark citrus leaf disease datasets, preprocessing methods, deep learning architectures, limitations, recommendations, and future research directions. It offers a detailed comparative analysis of various deep learning models categorised by task type and acquisition environment, highlighting their performance across different datasets, feature extraction techniques, and imaging conditions. Furthermore, this study outlines a novel strategic roadmap for integrating Generative AI and Vision-Language Models (VLMs) to enhance diagnostic interpretability. The primary objective of this survey is to aid researchers in understanding state-of-the-art deep learning-based methods for recognising citrus leaf diseases, support farmers and agronomists in early disease management, address current challenges, provide recommendations and outline future directions for sustainable citrus production.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 10","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11629-z.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148987103","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Vision foundation models for engineering document intelligence and manufacturing inspection: a survey","authors":"Kishore Reddy Pagidi,&nbsp;Anusha Gardas","doi":"10.1007/s10462-026-11622-6","DOIUrl":"10.1007/s10462-026-11622-6","url":null,"abstract":"<div><p>Vision foundation models such as SAM, CLIP, DINOv2, and multimodal large language models (MLLMs) are now common starting points for general computer-vision tasks. Their use in engineering visual data, however, is still split across separate communities working on drawing understanding, document intelligence, and manufacturing inspection. We found no prior survey that studies engineering drawing understanding and manufacturing inspection together from a foundation-model perspective, although both domains face the same basic obstacle: models pretrained on natural images must be adapted to thin-line, symbol-dense engineering documents and to controlled-lighting, texture-repetitive inspection imagery. Using a PRISMA-informed protocol, this survey reviews 94 studies published from 2020 to 2026 across six databases. For engineering drawing understanding, we cover foundation-model methods for symbol detection, GD&amp;T extraction, table and BOM parsing, P&amp;ID graph construction, drawing-to-CAD reconstruction, and floor-plan analysis. For manufacturing inspection, we follow the progression from CLIP-based zero-shot anomaly detection to DINOv2 self-supervised methods and MLLM-based defect reasoning, separating benchmark performance from evidence of factory readiness. Across the two domains, we identify five SAM adaptation strategies, compare adaptation regimes from frozen prompting to LoRA fine-tuning, catalogue 16+ engineering-specific datasets, and analyse six benchmark gaps, including the lack of a standard multi-format engineering drawing benchmark. We present DrawingBench as a five-track benchmark <i>design proposal</i> (data sources, baseline lineup, and evaluation protocol for each track) and outline research directions for engineering-specific pretraining corpora, parameter-efficient adaptation libraries, edge-deployment pipelines compatible with manufacturing execution systems and programmable logic controllers, and multimodal engineering agents with metrologically traceable perception.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 11","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11622-6.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149035249","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"From controlled benchmarks to unconstrained environments: a pipeline-oriented survey of models, domains, and emerging trends on human activity recognition","authors":"Thiago Brito Cassimiro da Silva,&nbsp;Hemir da Cunha Santiago,&nbsp;Bruno José Torres Fernandes","doi":"10.1007/s10462-026-11619-1","DOIUrl":"10.1007/s10462-026-11619-1","url":null,"abstract":"<div><p>Human Activity Recognition (HAR) has become a central research topic in computer vision, machine learning, and pervasive computing, driven by its wide range of applications in intelligent surveillance, healthcare monitoring, human–computer interaction, and smart environments. Over the past two decades, the field has evolved from handcrafted feature-based approaches and classical statistical models to deep learning architectures and hybrid paradigms that combine representation learning with robust classifiers. This paper presents a comprehensive survey of HAR research from 2004 to February 2026, systematically organizing the literature according to a canonical processing pipeline that includes sensing modalities/domains, feature representations, temporal modeling strategies, learning paradigms, datasets, and evaluation protocols. Beyond a descriptive review, the survey provides a cross-analysis of the relationships between techniques, classifiers, datasets, and application domains, highlighting recurring design patterns, trade-offs, and performance implications. Particular attention is given to emerging trends such as hybrid deep–classical models, privacy-preserving HAR, device-free sensing, and edge-oriented deployments. By synthesizing methodological advances and practical considerations, this survey aims to serve both as an entry point for new researchers and as a decision-oriented reference for practitioners seeking to design robust, generalizable, and ethically responsible HAR systems.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 8","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-06-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11619-1.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148380353","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"AI-powered malware intelligence: a review of machine and deep learning paradigms in detection and classification","authors":"Ananya Redhu,&nbsp;Suraj Ranganath,&nbsp;Medhansh Rajdeo,&nbsp;Disha Agarwal,&nbsp;Raggav Subramani,&nbsp;Shikhar Prakhar Nashine,&nbsp;Jayakumar Kaliappan,&nbsp;Kathiravan Srinivasan","doi":"10.1007/s10462-026-11621-7","DOIUrl":"10.1007/s10462-026-11621-7","url":null,"abstract":"<div><p>Malware remains one of the most significant threats to modern computer and digital systems. Owing to its continuously evolving nature, preventing and detecting malware has become increasingly complex. Traditional rule-based detection techniques are often ineffective against novel or obfuscated malware variants. Consequently, recent research has focused on leveraging machine learning and deep learning methods to identify evolving malicious patterns that can detect previously unseen threats. Since 2015, there has been a substantial growth in studies applying artificial intelligence approaches for malware detection and classification. This paper presents a contemporary review of state-of-the-art machine learning and deep learning algorithms used in malware analysis. To ensure methodological rigor and transparency, the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework was adopted. The review critically evaluates the strengths and limitations of existing models, analyses their performance in detecting and classifying diverse malware types, and discusses emerging challenges such as dataset imbalance, adversarial attacks, and model generalizability. Finally, the paper outlines key research directions aimed at improving the robustness, scalability, and interpretability of AI-driven malware detection systems.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 10","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11621-7.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148927808","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Bayesian network model development: evidence synthesis using published scientific literature review for tillage adaptation as a natural flood management strategy","authors":"Qaisar Ali","doi":"10.1007/s10462-026-11594-7","DOIUrl":"10.1007/s10462-026-11594-7","url":null,"abstract":"<div><p>Bayesian belief networks BBNs effectively model uncertainty in interconnected systems, though their construction and parameterisation present significant methodological challenges. This study introduces a Published Scientific Literature (PSL) review as a novel, evidence‑synthesis‑driven method for constructing a transparent and traceable Bayesian Belief Network structure in data‑poor environmental domains. My method helps capture this resource’s knowledge systematically, resulting in the construction of a BBN structure and data curation. I amplified a systematic review, a mapping process of identifying linkages between variables and constructing the network, and its evaluation with domain experts. The goal is to base the BBN on credible data sources supported by traceable evidence to identify key variables with interdependencies. I identified and coded narrative themes (as parameters) and subthemes (as parametric states) and extracted data frequency (as cases for parametrization) during a systematic review (SR). For illustration, I built a prototype BBN structure and parametrised the model using a case file originating through SR, synthesised evidence for tillage adaptation as a natural flood management strategy for sustainable farming and flood alleviation. This approach introduces the PSL review technique with a data curation framework as a sound foundation for model developers to build a basic BBN model for multi-disciplinary exploration.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 10","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11594-7.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148927908","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Beyond image generation: visual data analysis with diffusion models—a comprehensive survey","authors":"Bartłomiej Olechno,&nbsp;Jacek Mańdziuk","doi":"10.1007/s10462-026-11615-5","DOIUrl":"10.1007/s10462-026-11615-5","url":null,"abstract":"<div><p>This survey examines recent non-generative applications of diffusion models to visual data, systematically reviewing research papers from top-tier Artificial Intelligence / Machine Learning conferences, leading journals, and arXiv preprints. Our analysis focuses on a discriminative utilization of vision diffusion models (DMs) beyond image generation. We provide a novel taxonomy that divides existing applications of DMs into four groups: content detection, action understanding, spatiotemporal view estimation, and representation learning, with further more detailed division within each category. Our systematic analysis reveals that diffusion models achieve superior performance in uncertainty-critical discriminative tasks including pose estimation, anomaly detection, semantic correspondence, and depth estimation, but universally face computational overhead challenges with 10-100 times slower inference times than their discriminative alternatives. The survey identifies key advantages of diffusion-based data analysis, including a better handling of ambiguous ground truth, inherent uncertainty quantification, and rich representations of foundational characteristics. We also highlight promising hybrid approaches that combine diffusion and discriminative methods that maintain high performance while addressing computational constraints. The paper provides machine learning practitioners with systematic guidelines for leveraging diffusion models in visual analysis tasks and identifies critical research gaps in efficiency optimization and cross-domain generalization.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 10","pages":""},"PeriodicalIF":18.8,"publicationDate":"2026-06-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11615-5.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148927810","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}