Search Results (508)

With the continuous advancement of intelligence in power systems, traditional unimodal fault diagnosis methods can no longer satisfy the demand for precise monitoring of complex power equipment. To address the challenges of multimodal data fusion and fault diagnosis in intelligent sensing scenarios, this paper proposes a multimodal intelligent diagnosis model for power transformers based on causal dynamic fusion reasoning. By introducing a causal reasoning mechanism, the proposed model overcomes the limitations of conventional multimodal fusion approaches that rely solely on statistical correlations. A four-layer architecture is constructed, consisting of a Causal Dynamic Fusion layer, a Graph Reasoning layer, a State Prediction layer, and a Meta-Reinforcement Learning Optimizer, thereby forming a complete closed-loop framework from multimodal feature extraction to intelligent diagnostic decision-making. This study focuses on key issues including causal discovery and dynamic fusion in multimodal data, cross-sample contextual enhancement, equipment state prediction, and early warning. Performance evaluation experiments are conducted on a large-scale synchronized dataset containing image, audio, and time-series modalities. Experimental results demonstrate that the proposed CDFMD model outperforms conventional methods in diagnostic accuracy and real-time performance, providing a novel technical pathway for intelligent operation and maintenance of power equipment. Full article

Knowledge-based visual question answering (KB-VQA) requires leveraging external knowledge relevant to the image to assist reasoning. Existing methods typically convert images into a single textual description for knowledge retrieval or directly rely on the implicit knowledge within large language models to generate answers. However, a single textual description struggles to preserve fine-grained visual information such as object attributes and scene text, limiting retrieval quality. Meanwhile, naively fusing multi-source information tends to introduce modality noise, undermining reasoning accuracy. To address these issues, we propose a unified framework that constructs multi-source semantic anchors to bridge the cross-modal semantic gaps among vision, questions, and external knowledge. Specifically, we unify image captions, object tags, and optical character recognition (OCR) text as semantic anchors. These anchors serve as shared intermediaries to pre-align visual and textual features, avoiding direct interaction between heterogeneous modalities. During cross-modal fusion, a cross-residual gating mechanism adaptively suppresses modality noise by leveraging the semantic anchors as stable references. The framework further integrates contrastive learning to strengthen cross-modal alignment and employs a retrieve-then-read pipeline for open-domain answer reasoning. Experiments on the OK-VQA, FVQA, and A-OKVQA datasets demonstrate that the proposed framework outperforms state-of-the-art methods across multiple metrics, validating the effectiveness and robustness of the proposed framework. Full article

The imbalance of urban public transport networks often leads to traffic congestion. Traditional planning prioritizes system optimization and single-mode travel, neglecting interactions between different modes. From an economic perspective and based on passenger travel behavior, this paper constructs a reasonable path set for multimodal networks. Using information entropy, it establishes multidimensional indicators including site path diversity entropy, destination regional entropy vectors, and weighted comprehensive entropy. Regional aggregation and coefficient of variation analyze internal balance, while scatter plots and the Gini coefficient measure global resource allocation equity. ArcGIS Pro 3.4.3 is employed for spatial analysis and visualization. An empirical study of Beijing’s six central districts reveals significant spatial heterogeneity in path distribution across functional zones: working areas exhibit concentric patterns, commercial areas form corridor agglomerations, residential areas have the highest entropy values, and transport hubs are relatively balanced. Cluster analysis based on entropy vectors effectively identifies commuter, residential, and hub station types. Some hubs show an ideal “high richness, low imbalance” state, while areas like Beijing Railway Station exhibit “low richness, high imbalance.” The Gini coefficient of 0.1864 indicates relatively balanced public transport resources overall. The “route-region-demand” collaborative analysis framework constructed in this study achieves a paradigm shift from static network structure to dynamic human-oriented evaluation, providing methodological support for equity assessment, network optimization, and resource allocation in multimodal public transport networks, and can contribute to the equitable and balanced sustainable development of public transport. Full article

To address the technical challenges of difficult deduction, limited field measurement, and ambiguous instability determination of roof separation critical values in mining roadways within the weakly cemented coal-bearing strata of Xinjiang, this paper proposes a discrete element method that integrates the fracture of anchor bolt and anchor cable support materials with the damage degree of the surrounding rock. Taking a specific mine in the Hosh Tolgay coalfield as the research object, a systematic study was conducted. The research process was as follows. (1) Model parameter calibration was performed. Intact rock parameters were obtained through laboratory basic mechanical tests, and rock mass parameters were corrected based on reduction empirical formulas and the Hoek–Brown criterion. Numerical model verification showed that the errors between the simulated and theoretical values of the elastic modulus, compressive strength, and tensile strength of the rock mass were all less than 10%, indicating that the corrected parameters are reasonable. (2) The critical damage values of the rock mass considering a non-constant confining pressure environment were proposed. Through triaxial compression simulations, the differential evolution patterns of rapid damage increase in sandy mudstone under low confining pressure and stable damage accumulation in coal were revealed, thereby clarifying the damage thresholds for rock mass instability under different confining pressures. (3) A large-scale model was established to analyze the evolution laws of the fracture field, support field, and displacement field of the roadway surrounding rock. A comprehensive determination method for the instability of the roof anchored bearing structure was proposed. By comparing the damage thresholds of the scaled rock mass and the roadway surrounding rock and analyzing the fracture conditions of the roadway support system, a dual-criterion consisting of surrounding rock damage and support material fracture was constructed. Based on this criterion theory, the critical values for deep and shallow separation were obtained. The research results indicate that the evolution patterns of damage in coal and sandy mudstone differ with confining pressure. The sandy mudstone layers in the shallow part of the roof are more sensitive to mining-induced unloading disturbances. Consequently, the surrounding rock damage and support fracture of the mine roof exhibit distinct distribution characteristics: the dominant failure of the roadway is shear failure, with wide-range coalescence of shallow fractures and gradual development of deep fractures, alongside the concentrated failure of shallow anchor bolts and partial failure of deep anchor cables. Based on the instability state of the roof monitoring zones, the critical value for shallow separation was determined to be 90.7 mm, and the critical value for deep separation was 129.03 mm. These results are very close to the field measured values, verifying the engineering applicability of the method. This paper reveals the damage characteristics of the rock mass and surrounding rock in weakly cemented strata, as well as the mechanism of roof separation initiation and evolution. The proposed method for determining critical values provides a scientific and feasible practical reference for the support optimization and monitoring and early warning of roadway roofs in weakly cemented strata, possessing significant engineering value for ensuring safe and efficient mine production. Full article

Autonomous traversal flight in unknown 3D environments remains challenging due to mapping bottlenecks and computational latency. Inspired by pigeons navigating cluttered forests through instantaneous visual perception rather than constructing global metric maps, this paper presents a pigeon-inspired depth-reasoning-driven decision framework for agile quadrotor traversal in unmapped spaces without explicit map construction. To ensure feasibility, we leverage a robust state estimation backbone enhanced by deep-learning-based feature matching, providing stable pose feedback under aggressive maneuvers. The core contribution is a pigeon-inspired depth-reasoning framework that translates raw sensory depth data into a hybrid optimization framework, integrating both hard safety constraints and soft geometric smoothness constraints, directly emulating the three avian mechanisms: gap selection via instantaneous depth gradients, path selection that minimizes posture changes, and a safety field driven by the looming effect. By bypassing time-consuming mapping and spatial discretization processes, the framework significantly reduces perception-to-control latency. Finally, validated via simulations and real-world experiments on a resource-constrained quadrotor platform, our map-less approach achieves superior decision frequencies and comparable safety margins to those of state-of-the-art map-based planners. This framework offers a practical, high-frequency solution for autonomous flight where computational resources and environmental knowledge are strictly limited. Full article

Large language models (LLMs) are increasingly used in applied intelligent systems, but mid-sized models still lag on mathematical reasoning, partly because reliable step-level supervision is scarce. Many existing remedies rely on costly human annotation, stronger teacher models, or heavy training pipelines, which limits practical adoption. We propose VGPO-MCTS (Value-Guided Group-wise Policy Optimization over Monte Carlo Tree Search), a search-and-distillation framework that constructs reusable step-level supervision from datasets that provide only problems and final answers. VGPO-MCTS augments a frozen backbone with (i) a lightweight value model that scores candidate reasoning states formed by a reasoning prefix and its candidate next step, and (ii) a policy updated with parameter-efficient adaptation. During search, the value model guides tree expansion and selection, while verified outcomes are propagated backward to correct node utilities. The corrected search trees are then distilled into two complementary datasets: a value regression dataset for value learning and group-wise sibling candidate sets for GRPO-style policy optimization. Experiments on GSM8K and the MATH dataset with ChatGLM3-6B and SciGLM-6B show stable round-wise improvements in final-answer exact match under a lightweight adaptation setting. After three rounds of self-training, the proposed framework improves performance by about 6.3 percentage points on GSM8K and about 3.9 percentage points on MATH across the two backbones. Full article

Agentic artificial intelligence—systems that reason, plan, and act autonomously within governed workflows—is converging with autonomous electric mobility and urban robotics to reshape how cities govern, move, and manage physical space. We argue that the simultaneous arrival of these three vectors is triggering a transformation comparable in scope to the Industrial Revolution. Cities that deploy across all three domains are becoming the new hubs of innovation: they concentrate talent, accelerate knowledge circulation, enable cross-fertilisation, and generate hybrid proposals that no single vector could produce alone. Just as Manchester, Birmingham, and the Ruhr became the defining centres of industrialisation because steam, textiles, iron, and coal recombined through the proximity of the engineers and entrepreneurs who moved between them, a small number of cities today are pulling ahead because they host the shared talent pool around which agentic governance, autonomous mobility, and urban robotics co-evolve. Conceptually, we extend the mirroring hypothesis in two directions: dynamically, arguing that organisations and urban ecosystems converge toward the configurations new technologies make possible; and ontologically, arguing that agentic AI introduces non-human agents into organisational architectures, requiring hybrid human–AI coordination. We formalise this dynamic as five propositions (P1–P5) of cumulative recursive hybridisation (CRH), operating through four reinforcing feedback loops—data, regulation, infrastructure, and talent. Together, these loops explain why the emerging urban order is path-dependent: early movers accumulate compounding advantages, while latecomers face exponentially rising costs of entry. We demarcate CRH from adjacent frameworks—general-purpose technologies, organisational complementarities, and complex adaptive systems—and test it against counterfactual evidence from failed, stalled, and Global South trajectories (Sidewalk Toronto, the Cruise rollback, Songdo, Bengaluru). We also examine its political-economy, equity, and surveillance limits. Drawing on comparative evidence from public-sector chatbot deployments, autonomous mobility ecosystems in the United States and China, and emerging urban robotics cases, we conclude that what is at stake is not incremental modernisation but the construction of a new urban order. The cities that act as innovation hubs for the agentic AI era will shape global standards, attract global talent, and define the institutional templates that others eventually adopt—much as the industrial cities of the eighteenth and nineteenth centuries did. Full article

Accurate modeling of learners’ evolving cognitive states is essential for intelligent educational systems, yet many existing knowledge tracing and graph-based approaches rely on static structures or purely sequential representations that inadequately capture dynamic structural changes in learning processes. This study proposes a Learner Cognitive Graph (LCG) framework that integrates dynamic heterogeneous graph modeling, structured behavioral data acquisition, and reinforcement learning-based intervention optimization. A Dynamic Cognition Graph (DCG) is formally defined as a sequence of temporally evolving graph snapshots representing interactions among learners, knowledge concepts, and exercises. A reverse Turing test-based agent with structured prompting is introduced to collect reasoning-oriented behavioral evidence, improving data reliability for cognitive modeling. Temporal message passing, multi-scale memory updating, and self-supervised learning objectives are employed to construct dynamic cognitive representations. Personalized intervention is formulated as a Markov decision process to optimize long-term learning outcomes. Experiments conducted on real-world and simulated educational datasets demonstrate improved knowledge mastery prediction accuracy, cognitive state transition modeling, and intervention efficiency compared with representative baselines. The proposed framework provides a systematic and scalable approach for dynamic cognitive modeling and adaptive educational support. Full article

Against the background of carbon neutrality, the government’s carbon regulations have had a profound impact on the distributed energy sharing behavior of industrial parks. To deeply explore the interactive relationship between distributed energy sharing in industrial parks and government regulation, this paper constructs a three-party evolutionary game model composed of the government, core enterprises and supporting enterprises; endogenizes government behavior; and integrates inter-enterprise contractual mechanisms into the evolutionary framework. By establishing a revenue payment matrix and a replication dynamic equation, the stability conditions and system evolution paths of the strategy choices of each subject are analyzed, and numerical simulations are conducted. The results show that there are multiple evolutionary stable equilibria in the system, among which the equilibrium where core enterprises actively share, supporting enterprises actively share, and the government actively regulates carbon is the ideal state. Cost-sharing contracts and cooperative penalty contracts play a significant role in promoting the participation of supporting enterprises in sharing and curbing “free-riding” behavior, respectively. The changes in government subsidy rates and carbon tax rates have a crucial impact on the evolution of corporate strategies. Quantitatively, the carbon tax rate exhibits a threshold effect; enterprises shift to positive energy sharing when the tax rate exceeds 0.8, while a subsidy rate above 0.4 leads the government to withdraw from regulation. This indicates that a reasonable design of carbon regulations can help achieve coordinated energy emission reduction between the government and enterprises. The findings provide theoretical support for optimizing carbon regulations and designing cooperation strategies. Full article

Urban waste management remains a significant challenge for achieving environmental sustainability and advancing smart city infrastructures. This study proposes a multimodal vision–language framework that integrates real-time object detection with automated semantic interpretation and structured semantic analysis for intelligent urban waste monitoring. A custom dataset including 2247 manually annotated images was constructed from publicly available sources (TrashNet and TACO), enabling robust multi-class detection across six waste categories. Two state-of-the-art object detection models, YOLOv8m and YOLOv10m, were trained and evaluated using a fixed 70/15/15 train–validation–test split. Under this configuration, YOLOv8m achieved a mAP@50 of 90.5% and a mAP@50–95 of 87.1%, slightly outperforming YOLOv10m (89.5% and 86.0%, respectively). Moreover, YOLOv8m demonstrated superior inference efficiency, reaching 120 FPS compared to 105 FPS for YOLOv10m. To obtain a more reliable estimate of performance stability across data partitions, stratified 5-Fold Cross-Validation was conducted. YOLOv8m achieved an average Precision of 0.9324 and an average mAP@50–95 of 0.9315 ± 0.0575 across folds, suggesting generally stable performance across data partitions, while also revealing variability associated with dataset heterogeneity. Beyond object detection, the framework integrates MiniGPT-4 to generate context-aware textual descriptions of detected waste items, thereby enhancing semantic interpretability and user engagement. Furthermore, GPT-5 Vision is incorporated as a structured auxiliary semantic classification and category-suggestion module that analyzes object crops and multi-class scenes, producing constrained JSON-formatted outputs that include category labels, concise descriptions, and recyclability indicators. Overall, the proposed YOLOv8–MiniGPT-4–GPT-5 Vision pipeline shows that combining accurate real-time detection with multimodal semantic reasoning can improve interpretability and support interactive, semantically enriched waste analysis in smart-city and environmental monitoring scenarios. Full article

The knowledge graph plays an important role in the construction of artificial intelligence applications. However, the incompleteness of the knowledge graph seriously affects the performance of downstream applications. The problem has fueled a lot of researches on knowledge graph completion (also known as the tasks of link prediction). Reinforcement learning-based multi-hop reasoning that formulates link prediction as a sequential decision problem has also become an interesting and promising approach. Nevertheless, in an incomplete knowledge graph environment, the policy-based agent might travel a large number of low-quality or spurious search trajectories, which inhibits the model performance. Therefore, in this paper, we propose a path and structural features-enhanced reinforcement learning model (referred as PGATRL). First, we leverage the path constraint resource allocation algorithm to mine high-quality inference paths, which can be employed to pre-train the LSTM path encoder module in the reinforcement learning architecture, and thus play a role in guiding the agent’s action decision-making. Second, we exploit the adapted graph attention networks to encode the local structural features of an entity, which can provide more evidence for the agent to find a more suitable reasoning path. With extensive experiments on several benchmark datasets, our proposed approach gains significant improvements compared with the state-of-the-art baselines. Full article

Accurate classification of breast cancer histopathology images is essential for early diagnosis and effective clinical management. However, conventional deep learning models often exhibit performance degradation under limited labeled data and lack interpretability, which restricts their clinical applicability. To address these challenges, we propose GAFR-Net, a robust and interpretable Graph Attention and Fuzzy-Rule Network designed for histopathology image classification under scarce supervision (defined here as less than 10% labeled data). GAFR-Net constructs a similarity-driven graph to model inter-sample relationships and employs a multi-head graph attention mechanism to capture complex relational representations among heterogeneous tissue structures. Meanwhile, a differentiable fuzzy-rule module integrates intrinsic topological descriptors—such as node degree, clustering coefficient, and label consistency—into explicit and human-readable diagnostic rules. This architecture establishes transparent IF–THEN inference mappings that emulate the heuristic reasoning process of clinical experts, thereby enhancing model interpretability without relying on post-hoc explanation techniques. Extensive experiments conducted on three public benchmark datasets—BreakHis, Mini-DDSM, and ICIAR2018—demonstrate that GAFR-Net consistently surpasses state-of-the-art methods across multiple magnifications and classification settings. These results highlight the strong generalization capability and practical potential of GAFR-Net as a trustworthy decision-support framework for weakly supervised medical image analysis. Full article

Temporal Knowledge Graph Inference (TKGI) is a cornerstone for intelligent decision-making in dynamic scenarios, but existing models face critical bottlenecks, including inadequate complex-context modeling, a lack of entity importance quantification, insufficient novel-event reasoning accuracy, and weak domain adaptability. To address these issues, this study proposes a semantics-enhanced model (LLM-DSaR) integrating Large Language Models (LLMs), temporal attention networks, and optimized contrastive learning. Specifically, a two-stage LLM semantic enhancement (LLM1 + LLM2) framework first generates structured semantic analysis reports via adaptive prompt engineering, and then extracts domain-specific semantic embeddings from the last-layer hidden states through pooling and linear projection, which are further fused with TransE-based structural embeddings; meanwhile, LLM2 mitigates data sparsity in novel-event reasoning; a dynamic weight fusion (DWF) framework adaptively assigns feature weights to achieve deep feature synergy; an LLM-enhanced contrastive-learning module strengthens event clustering and discrimination. Experiments on five public datasets and a self-constructed Robotics Temporal Knowledge Graph (RTKG) show LLM-DSaR outperforms 16 baselines: on RTKG, its MRR is 10.35 percentage points higher than GCR, and Hits@10 reaches 88.87%. Ablation experiments validate core modules’ effectiveness, confirming LLM-DSaR adapts to professional scenarios like robot maintenance prediction, providing a novel technical paradigm for complex-domain TKG reasoning. Full article

This study explores the multi-rate sampling-based $H_{\infty }$ load frequency control (LFC) problem for networked power systems by using an area-information-fusion method. This problem is addressed for two reasons: (1) most of networked control methods for LFC are focused on the one-rate sampling scheme and (2) the previous looped function cannot be directly applied within the multi-rate sampling scheme. Here, the multi-rate sampling scheme involves each area sampling rate being reliant on its own sensor. Namely, all area sampling rates are different from each other. In the presence of a multi-rate sampling scheme, a new sampling instants sequence is established by using an area-information-fusion method. It contributes to constructing a corresponding closed-loop model by adding virtual state variables. In addition, a new looped-function approach is devised to capture the sampling information from diverse area sensors. Based on Lyapunov stability theory, less conservative LMI conditions are derived to guarantee the $H_{\infty }$ performance of the multi-rate LFC system. Additionally, a co-designed method for determining the control gain and maximum sampling frequency is established. Finally, simulation studies are conducted to validate the efficacy and features of the proposed control strategy. Full article

The rapid progress of large language models (LLMs) has fostered the development of domain-specific variants in law, medicine, and finance. However, existing legal LLMs still struggle to generate contextually grounded and regulation-compliant responses in complex scenarios of fair competition review. To address this, we present FairAgent, a collaborative multi-agent framework that unifies data refinement and reinforcement learning for legal reasoning. FairAgent integrates two core modules: (1) EchoCourt, a closed-loop data generation and refinement pipeline that constructs high-quality question–answer pairs through generation, critique, and optimization guided by a hierarchical Fairness Knowledge Forest; and (2) a two-stage outcome-based reinforcement learning mechanism that progressively teaches the model to invoke and integrate external retrieval in reasoning. We further enhance learning stability through a RAG-based rollout and retrieval-mask loss. Extensive evaluations demonstrate that FairAgent significantly improves reasoning accuracy, interpretability, and compliance in fair competition review compared with state-of-the-art baselines, establishing a scalable framework for retrieval-augmented legal intelligence. Full article