Search Results (797)

In smart cities, crowding in transportation hubs, large event venues, and commercial/tourist districts can rapidly escalate from service congestion to public-safety incidents. Real-world operations are constrained by heterogeneous sensing coverage, delayed statistics, privacy requirements, and the need for accountable multi-agency decisions. Following a rigorous PRISMA protocol, we synthesized 107 primary empirical studies (2020–2026) to systematically review context-aware crowd technologies. Moving beyond isolated algorithmic benchmarks, we organized these advances into a mathematically formalized closed-loop framework (Sensing–Prediction–Intervention–Feedback). Crowd sensing has evolved toward edge-based computer vision, passive mobile signaling, and multimodal fusion to balance operational trade-offs among density applicability, environmental robustness, privacy burdens, and end-to-end latency. Prediction architectures—converging on Spatiotemporal Graph Neural Networks (ST-GNNs) and simulation-augmented digital twins—are critically evaluated against constraints in predictive horizon, computational overhead, and explainability. To bridge theory and practical deployment, we deduce a multidimensional evaluation taxonomy and a hierarchical trigger-and-escalation matrix, tailoring control philosophies (e.g., spatiotemporal capacity synchronization and dynamic demand reshaping) to the three heterogeneous scenarios. Finally, we propose four strategic directions to chart a deployment-oriented roadmap for the integrated Urban Brain: edge-based privacy-preserving fusion, cross-scenario generalization, accountable Explainable Artificial Intelligence (XAI) with human-in-the-loop synergy, and end-to-end outcome-oriented empirical validation. Full article

Graph Neural Networks (GNNs) have achieved promising performance in various graph learning tasks. Real-world graph learning often suffers from severe label scarcity, which hinders GNNs’ feature and topology learning and causes poor generalization. Node mixup has therefore been shown to be an effective graph data augmentation paradigm for enhancing GNNs’ performance in label-scarce graph learning scenarios. However, mainstream node mixup methods adopt fixed or randomly sampled mixing ratios while neglecting inherent correlations between node pairs, thereby producing less informative or noisy augmented samples. To address the issue, we propose a novel semantic-aware adaptive node mixup method, SA-Mixup, for node classification tasks. Specifically, SA-Mixup first leverages high-confidence pseudo-labeling to enrich supervisory signals by incorporating reliable unlabeled nodes into the training set, and then it selects node pairs with the same label for node mixup. Furthermore, an adaptive node mixup mechanism is developed to dynamically learn optimal mixing ratios according to semantic relation and predictive uncertainty for each node pair. In addition, a similarity-guided neighbor connection strategy is introduced to select candidate neighbors for generated virtual nodes and establish reasonable topological connections. The experimental results show that the proposed method achieves strong performance when combined with different GNN backbones. The ablation results further show that both the adaptive node mixup mechanism and the similarity-guided neighbor connection contribute to the performance improvements. Full article

Wind power forecasting plays a crucial role in power dispatch and safety management of wind farms. However, the insufficient integration of multimodal heterogeneous data and the limitations of conventional graph construction strategies significantly restrict forecasting performance. Existing approaches either rely on simple feature aggregation, which cannot fully capture cross-modal dependencies, or adopt predefined or single-criterion graph construction methods that fail to characterize complex turbine relationships involving spatial, temporal, and nonlinear correlations. To address these challenges, this paper proposes a Multimodal Adaptive Fusion Graph Neural Network (MAF-GNN) for short-term wind power forecasting. First, a Modality-Aware Representation Learning (MARL) module is developed to extract informative multimodal representations by modeling modality-specific characteristics and cross-modal dependencies through attention-based fusion. Second, an Adaptive Graph Learning with Multi-Similarity (AGL-MS) module is introduced to parametrically integrate four complementary similarity priors—geographic distance, Dynamic Time Warping (DTW), Maximal Information Coefficient (MIC), and cosine similarity—for adaptive turbine correlation graph construction. Furthermore, a Credibility-Modulated Graph Convolutional Network (CM-GCN) is developed to reduce the influence of unreliable node information during message propagation. Extensive experiments conducted on the SDWPF dataset demonstrate that MAF-GNN reduces MAE by 14.0–21.3% compared with sequential baselines and achieves 5.3–10.2% improvement over spatiotemporal graph-based models. Ablation studies further verify the complementary effectiveness of each proposed module in improving forecasting performance. Full article

Coordinating multiple unmanned aerial vehicles (UAVs) at scale remains challenging through centralized control or fixed rule sets, particularly when vehicles must operate under intermittent communication links, incomplete observability, and constrained onboard computational resources. Graph Neural Networks (GNNs) have emerged as a promising framework for addressing these challenges; however, existing surveys do not systematically relate GNN architectural decisions to the operational constraints imposed by UAV platforms during deployment. This survey reviews 196 scholarly studies published between 1987 and 2026 to develop such a framework. A GNN architecture and deployment taxonomy is organized into six major categories—Convolutional, Attentional, Sampling-Based, Spatio-Temporal, Distributed, and Resource-Efficient—each examined through dedicated architectural subsections and evaluated in the context of UAV system constraints. Four primary application domains are examined: multi-UAV trajectory planning, cooperative target tracking, communication-aware network optimization in Flying Ad Hoc Network (FANET) environments, and spatio-temporal airspace traffic prediction. Within these domains, the analysis highlights how architectural choices influence scalability, adaptability to dynamic conditions, and computational efficiency. Several deployment challenges consistently emerge, including maintaining tractable inference as swarm size increases, adapting graph representations under high mobility, and operating within the limitations of onboard computational resources. Based on these findings, a set of architecture-selection guidelines is derived to support deployment under varying operational conditions. Emerging research directions are also discussed, particularly the integration of GNNs with reinforcement learning, federated edge computing, and next-generation wireless communication systems. Overall, this survey bridges the gap between methodological development and practical deployment, providing a structured foundation for evaluating GNN suitability in real-world multi-UAV environments. Full article

Intelligent fault diagnosis of rotating machinery is essential for ensuring the safety and reliability of industrial systems. While hypergraph neural networks (HGNNs) have recently shown promise for modeling high-order dependencies beyond pairwise graph methods, most existing variants operate in Euclidean space, which is not explicitly aligned with hierarchical fault-response structure (root cause to fault mode to observed response). To address this limitation, we propose Hyperbolic Hypergraph Neural Network (H²GNN), a framework that integrates hyperbolic geometry with hypergraph neural networks for fault diagnosis. Specifically, H²GNN constructs fault-response-aware hyperedges over diagnostic views of vibration signals and performs message passing in the Poincaré ball model, a Riemannian manifold of constant negative curvature commonly used for hierarchical representation learning. We introduce Poincaré hyperedge aggregation via an iterative Fréchet-mean solver, a learnable curvature parameter for adaptive manifold fitting, and a tangent-space classification head. Experiments are conducted on two public benchmarks, namely the Case Western Reserve University (CWRU) bearing dataset and the Machinery Failure Prevention Technology (MFPT) bearing dataset, and report mean accuracies of 99.87% and 99.75%, respectively, outperforming six competing methods, including CNN, GCN, HGNN, dynamic-HGNN, contrastive-HGNN, and spatial-temporal HGNN. Ablation studies indicate that hyperbolic geometry and the adaptive curvature mechanism both contribute to the observed performance gain. Full article

Accurate prediction of the Remaining Useful Life (RUL) of nuclear power plant systems can support more informed maintenance strategies and help reduce unplanned outages, motivating continued research into reliable, physically grounded prognostic models. This study evaluates machine-learning-based prognostic models, focusing on their ability to learn degradation patterns directly from operational data. Baseline Feedforward Neural Networks (FNNs), Long Short-Term Memory (LSTM) networks, Graph Neural Networks (GNNs), and a hybrid GNN–Transformer architecture are assessed using data generated from the ASHERAH dynamic Pressurized Water Reactor simulator, which incorporates realistic degradation mechanisms, including condenser fouling and pump head loss. The proposed hybrid model integrates graph-based representations of component interactions with Transformer-based temporal attention to capture both system-level dependencies and long-term degradation dynamics. Model performance is evaluated using error-based and tolerance-based metrics, including Prediction Within Bounds Accuracy ($PWB{A}_{20\%}$), alongside uncertainty calibration via Prediction Interval Coverage Probability (PICP), with uncertainty quantified through deep ensembles. Results show that baseline NNs exhibit limited predictive accuracy and poor uncertainty calibration, while models incorporating temporal modeling and system topology achieve substantial improvements. The GNN–Transformer attains the strongest performance, yielding the highest $PWB{A}_{20\% }$ (87.3%) and significantly improved uncertainty calibration, with an average PICP of 90.0%. These findings demonstrate the effectiveness of topology-aware, attention-based architectures for robust and reliable RUL prediction in complex nuclear systems. Full article

Non-Revenue Water (NRW) from leakage represents a major global economic and environmental challenge for urban utilities. While Deep Learning (DL) offers transformative potential for leak detection in Water Distribution Systems (WDSs) and existing reviews provide critical assessments, a consolidated, quantitative evaluation of real-world applicability and performance consistency that is actionable for engineering practice remains absent. This systematic review critically evaluates DL applications for WDS leak detection and localisation, with a focused analysis of model accuracy in relation to data types, methodological rigour, and the validation gap between controlled experiments and operational deployment. Following the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) framework, a systematic literature search was performed using Scopus, Web of Science, Google Scholar, ScienceDirect, Taylor & Francis, and MDPI databases for publications spanning the period from 2015 to 2025. From an initial 5265 records, 72 studies met the inclusion criteria for qualitative synthesis. Analysis revealed a specialisation of DL architectures by data modality: Convolutional Neural Networks (CNNs) applied to acoustic or vibration data yield the highest reported accuracy for direct leak identification; Long Short-Term Memory (LSTM) and Transformer models are predominant for temporal hydraulic data (pressure and flow); and Graph Neural Networks (GNNs) excel with topological data for state estimation. While reported accuracy is often high, performance is highly contingent on data quality and pre-processing. A significant disparity exists between results on synthetic versus real-world validation datasets, ranging from a decline of approximately 3 to 30 percentage points, with reported real-world accuracy spanning 70 to 79.7 percent. Moreover, DL demonstrates a paradigm shift in technical capability for leak management. However, transitioning to reliable field applications requires overcoming key challenges: standardising benchmarks and performance reporting, improving model generalisability and explainability, and fostering integration within practical Digital Twin (DT) frameworks to enable proactive infrastructure management. Full article

Accurately characterizing the anisotropic optoelectronic properties of crystals requires determining the band gaps at specific high-symmetry points in the Brillouin zone. Relying solely on the minimum band gap is insufficient. However, although Graph Neural Networks (GNNs) offer rapid property predictions, conventional models remain trapped in a local real-space paradigm, lacking the global symmetry information necessary to differentiate these high-symmetry energy states. To address this problem, we propose the Symmetry-Information-Enhanced Crystal Graph Neural Network (Symm-CGNN). It explicitly incorporates local atomic environments with global symmetry information, including space groups, crystal systems, material density and lattice constants. The evaluation is performed on a comprehensive dataset including 3D (Materials Project) and 2D (2DMatpedia). The results demonstrate that Symm-CGNN achieves an 18% reduction in Mean Absolute Error (MAE) for high-symmetry band gap prediction compared to the baseline Crystal Graph Convolutional Neural Networks (CGCNN). This approach bridges the representational gap between local atomic coordination and macroscopic symmetry. Consequently, it provides a robust and efficient machine-learning paradigm for the high-throughput screening of materials with anisotropic optoelectronic properties. Full article

Ultra-Wideband (UWB) positioning is a promising technique for underground mine localization, but its accuracy is strongly affected by complex tunnel topology, multipath propagation, and non-line-of-sight (NLOS) ranging errors. To address these challenges, this paper proposes a three-dimensional UWB positioning algorithm based on a Multi-head Attention Feature Fusion Graph Neural Network (MAFF-GNN). The localization problem is formulated as a graph-based node position regression task, where anchors and tags are represented as nodes and UWB ranging links are represented as edges. The proposed model integrates graph message passing, multi-head attention-based feature fusion, global skip connections, and geometry-constrained regularization to learn reliability-aware spatial representations from noisy ranging measurements. A physics-guided simulated underground mine environment is constructed by considering tunnel geometry, wall roughness, coal dust concentration, humidity attenuation, and controlled NLOS conditions. Three mine-like corridor topologies are generated, with 10,500 localization samples in total. Experimental results under controlled simulation conditions show that MAFF-GNN achieves an RMSE of 0.323 ± 0.093 m, an MAE of 0.274 ± 0.024 m, and a P90 error of 0.480 ± 0.038 m. Compared with weighted least squares and support vector regression, the proposed method reduces RMSE by 74.16% and 40.30%, respectively. Robustness tests under different simulated NLOS ratios further indicate that the proposed graph-attention framework maintains a gradual error-growth trend as NLOS severity increases. These results indicate the potential of attention-enhanced graph learning for UWB localization in a challenging mine-like environment. Full article

Wireless edge systems increasingly rely on on-device learning for tasks such as user localization and misuser detection, but centralized training is often infeasible because of privacy and bandwidth constraints. In this work, we study communication-efficient federated learning schemes that jointly train lightweight neural models across clients. We compare a multi-task shared-backbone strategy with mid-training backbone freezing against task-separate baselines. Using a wireless graph simulator, we prototype both multilayer perceptrons (MLPs) and an edge-aware graph neural network (GNN) encoder trained with FedAvg, and we evaluate localization error, misuser F1-score, and byte-level communication over training rounds. The results show that the GNN-Shared model achieves a 28.9% lower localization error than the MLP-Shared baseline and a 17.7% lower localization error than the GNN-Loc_only baseline, while improving the best misuser F1-score by 2.7% compared with GNN-Mis_only. In terms of communication, the GNN-Shared model reduces the total communication cost by 40.1% compared with MLP-Shared over 40 federated rounds. Additional experiments with FedProx, GraphSAGE, and GAT baselines show that the proposed federated graph-learning framework is flexible across different graph backbones. Multi-seed experiments over five random seeds further confirm the robustness of graph-based federated learning, with GraphSAGE-FedAvg achieving strong average localization and misuser detection performance. Scalability experiments with up to 20 clients, 150 nodes, and different Dirichlet non-IID parameters show that the framework remains stable under larger graph sizes, while communication cost grows mainly with the number of participating clients. To move the framework closer to edge deployment, this paper also introduces a lightweight hardware-oriented GNN-Lite inference prototype with multi-neighbor accumulation, finite-state-machine-based sequential computation, and read-only-memory-based coefficient storage. The prototype achieved timing closure on an Artix-7 field-programmable gate array with 213 LUTs, 111 FFs, 18 DSPs, 0 BRAMs, and 0.106 W estimated total on-chip power. The dynamic power was only 0.001 W, and the estimated total energy consumption was 10.6 nJ per inference. Overall, the results show that federated graph learning is a promising direction for wireless edge intelligence, and that different graph backbones can be selected depending on the target trade-off among localization accuracy, misuser detection performance, communication cost, robustness, and hardware deployment. Full article

Distributed acoustic sensing (DAS) systems capture complex, nonlinear wave propagation across fiber-optic cables. Conventional event classification architectures, constrained by static physical topologies or isolated spatial grids, fail to effectively adapt to the dynamic feature relationships associated with such events, particularly when modeling complex spatiotemporal interactions across sensor arrays. To resolve these structural limitations, we introduce the Spectro-Spatial Dynamic Graph Neural Network (S2-DyGNN), whose architecture couples a two-dimensional frequency–time convolutional front-end with a dual-matrix graph neural network (GNN). First, the convolutional module extracts spectro-temporal features, explicitly capturing localized acoustic dynamics independent of inter-sensor interference. Subsequently, the graph module constructs a dual-matrix topology, fusing a static physical distance prior with a data-driven adjacency matrix that recalculates spatial connections frame by frame from input signals. When evaluated on a highly skewed nine-class DAS field dataset, S2-DyGNN outperformed other conventional models by achieving a peak macro-averaged F1-score of 86.6% and an overall accuracy of 94.0%. The dual-matrix graph topology prevented dominant background features from washing out sparse transient events, improving the minority “openclose” class F1-score to 55.7% compared to the 48.0% ceiling of a static graph topology. These results demonstrate that explicitly coupling localized spectro-temporal representations with physically anchored spatial topologies consistently outperforms models that process these domains in isolation, providing a highly robust and scalable solution for real-world continuous monitoring systems. Full article

Graph Neural Networks (GNNs) have become essential for analyzing graph-structured data, yet their deployment on resource-constrained edge devices is severely limited by high computational complexity and irregular memory access patterns. Here, we introduce DynamiGraph, a specialized FPGA-based overlay accelerator engineered for ultra-low-latency GNN inference in edge computing scenarios. Unlike general-purpose accelerators that incur high resource overhead to support a broad range of operators, DynamiGraph adopts a streamlined architecture focusing exclusively on essential General Matrix Multiplication (GEMM) and Sparse–Dense Matrix Multiplication (SpDMM) kernels. We implement a hardware-native runtime optimization mechanism that dynamically exploits graph sparsity via an edge-centric execution flow, eliminating redundant computations without requiring complex static preprocessing. Experimental results on an AXU2CGA edge platform demonstrate that DynamiGraph achieves sub-millisecond inference latencies on small-scale benchmarks (e.g., Cora) and a peak throughput of 1467 inferences per second. Furthermore, our runtime sparsity exploitation yields over 2000× reductions in floating-point operations compared to dense equivalents. These findings indicate that trading off model generality for architectural specialization and runtime awareness offers an efficient architectural alternative for enabling real-time graph intelligence in power- and bandwidth-limited edge environments. Full article

This research addresses cyber risk by defending against backdoor attacks on Graph Neural Networks (GNNs). We propose the Explainable Complex System-Mitigation Triangular (ECSMT) Framework, which integrates Robust Training, Graph Regularization, and Data Sanitization into a lightweight, hardware-efficient defense layer. To evaluate structural generalizability, we conducted empirical evaluations across three distinct benchmark domains (AIDS, MUTAG, and PROTEINS) using a Graph Isomorphism Network (GIN) backbone. Under a baseline 5% backdoor subgraph trigger injection ratio, ECSMT achieves excellent utility retention, securing a Clean Accuracy (CA) of 97.33% (±0.62%) while reducing the Attack Success Rate (ASR) from 97.00% down to 69.45% on the primary AIDS benchmark. Cross-domain testing reveals that defensive efficacy is strongly constrained by dataset characteristics: small-scale datasets such as MUTAG suffer from persistent trigger concentration, while complex graph manifolds such as PROTEINS exhibit high levels of topological noise. Furthermore, mapping these technical outcomes into an enterprise asset framework yields a 61% expenditure compression at critical technological feeder locations and a 98.93% reduction in total systemic loss. This study indicates that the proposed triangular mitigation strategy offers a valuable, scalable blueprint for enhancing the technical resilience and prognostic economic modeling of critical infrastructure networks. Full article

Existing solutions for resource allocation in Zero Trust (ZT) SDN networks treat security, resilience, and efficiency separately; centralised approaches violate data privacy; deep reinforcement learning (DRL) lacks trust dynamics; and federated learning (FL) has not incorporated graph neural networks (GNNs) or failure resilience. These limitations motivate a federated GNN-DRL framework that preserves data locality while jointly optimising trust, resilience, and performance. Each domain trains a local GNN-DRL agent with a GNN encoder for topology awareness and a hybrid DRL module (Deep Q-Network for discrete failover actions and Soft Actor-Critic for continuous bandwidth tuning) under a stochastic trust evolution modelled by stochastic differential equations (SDEs) with a reflection mechanism to ensure well-posedness. Three pseudo-code algorithms detail client-side training (Algorithm 1), server-side q-fair aggregation (Algorithm 2), and local gradient updates (Algorithm 3). Extensive simulations on a 100-node topology divided into five domains demonstrate that: (i) under low-to-moderate failures (≤20%), trust violations remain below 1.8%, and even under severe failures (40%), violations stay at 4.2% (within the 5% ZT boundary); (ii) recovery time is reduced by 53%; (iii) throughput under failures improves by 32%; (iv) compromise resistance reaches seven nodes (vs. three centralised); (v) attack surface shrinks to four nodes (vs. 98 baseline); and (vi) lateral movement containment attains 98%. The federated framework approaches centralised performance while preserving data locality, offering a practical and secure solution for multi-domain ZT, SDN networks. Full article

Residual decoding algorithms repeatedly score the current elements, prune the lowest-scoring candidate, and update the residual state. Classical implementations are efficient because their scores are built from locally maintainable statistics such as degree: after a deletion, surviving scores change by local corrections, giving $O\left(n^2\right)$ total decoding cost. However, a Graph Neural Network (GNN) score does not automatically have this property. After nonlinear activations such as tanh or sigmoid, equal changes in the underlying statistic need not induce equal changes in the output, so no fixed-additive correction can generally update the score after a deletion. The standard adaptive alternative is to rerun the full network after every deletion, inflating decoding cost to $O\left(n^3\right)$. We show that, for Max-Clique, the gradient of an unsupervised clique objective provides an update-compatible residual signal. After each vertex deletion, every surviving gradient coordinate changes by a local, maintainable quantity, updatable in $O\left(n\right)$ per step. Supplying this live gradient to a lightweight learned updater recovers $O\left(n^2\right)$ decoding at state-of-the-art quality on Easy and Medium planted clique regimes. Zero-shot evaluation on six real-world graph benchmarks confirms consistent gains over score-only updating. Full article