Search Results (1,931)

Persistent open-vocabulary navigation enables robots to search for sequential language-specified objects using reusable visual semantic evidence. On edge GPUs, the pipeline is constrained by foundation-model inference, semantic projection, persistent map movement, frontier processing, and repeated target detection. Using OneMap as a controlled reference, we reorganize this workload through compiled mixed-precision perception, direct-to-map confidence-weighted feature aggregation, accelerator-resident semantic maps, compiled navigation operations, and language-conditioned detector scheduling. The resulting dataflow removes the dense image resolution feature intermediate and schedules target confirmation detection from current-view target similarity. On the complete Habitat-Matterport 3D benchmarks, our implementation retains 92.6% and 96.1% of the reproduced OneMap single- and multi-object success rates. Across 36 paired hardware-in-the-loop comparisons on a Jetson AGX Orin at four power modes, it achieves geometric mean speedups of 4.64× for on-device computation and 3.55× for the complete episode duration. It also reduces peak system random access memory (RAM) by 51.21% and a summed board-rail energy proxy by 75.32% on average while preserving every paired outcome. These results establish the coordinated dataflow and current-view target-conditioned execution as effective mechanisms for efficient persistent foundation-model navigation on power-constrained edge robots. Full article

In railway operations and safety oversight, vast amounts of accident-related data are recorded in unstructured textual formats, posing challenges for efficient information extraction and analysis. To address this, we constructed a knowledge graph from railway accident profile texts and integrated it with a Graph Retrieval-Augmented Generation (GraphRAG)-enhanced retrieval framework to support both basic and composite queries on railway accident information. First, railway accident profile texts were preprocessed, and Easy Data Augmentation (EDA) was used to expand samples of different types, thereby constructing a railway accident profile text dataset for subsequent knowledge extraction. We then introduced a deep learning model, RoBERTa-CNN-BiLSTM-CRF (RCBC), for automatic extraction of seven categories of entities, including accident IDs and causes. To extract semantic relations, we designed a prompt-based template leveraging large language models (LLMs). To mitigate information loss during entity extraction, a GCN-Attention-LLM (GA-LLM) model was further designed for knowledge graph completion. Experimental results show that RCBC achieves MicroF scores above 85% across entity extraction tasks, while GA-LLM attains an average Hits@3 of 82.84% in knowledge completion. In tests on basic and composite questions, LLM-GraphRAG outperformed both LLM and LLM-RAG in faithfulness, semantic similarity, context precision, and context recall. The resulting knowledge graph contains 1493 entities and 1832 relations. Combined with the retrieval framework, the system enables access to key railway accident information and offers technical support for intelligent railway safety management. Full article

Operational redundancies in Nigeria’s airport logistics, including duplicated clearance procedures, fragmented data systems, and repeated physical inspections, generate substantial inefficiencies that hamper service delivery, increase costs, and raise sustainability concerns. Despite growing interest in blockchain technology (BCT) as a logistics innovation, empirical evidence from Sub-Saharan African aviation contexts is largely limited. This study examines how stakeholders perceive BCT as a potential mechanism for reducing procedural, informational, and structural redundancies across Nigerian airport operations. A qualitative research design was employed, drawing on 45 semi-structured interviews with airport managers, airline officials, regulatory bodies, and aviation experts. Data were analyzed using Leximancer, yielding eight thematic clusters across 115 concepts. To augment thematic rigor, three quantitative measures—Normalized Pointwise Mutual Information (NPMI), Jaccard similarity, and degree centrality—were applied to the exported Concept Co-occurrence Matrix (CCMatrix; 6555 pairwise relationships). The blockchain theme exhibited the highest intra-theme cohesion (mean NPMI = 0.615) and the second highest normalized centrality (0.638), suggesting that it functions as a semantically tight, cross-cutting concept in participants’ discourse. The strongest concept pair, inefficiency–reliability (NPMI = 0.989; Jaccard = 0.909), is consistent with participants’ recurring association of service failures with procedurally generated redundancies. Stakeholders perceived BCT as potentially addressing these redundancies through shared ledger architectures, smart contract automation, and consolidated identity management, with anticipated sustainability co-benefits across energy, documentation and processing domains. The study offers an exploratory, contextualized application of the Technology–Organization–Environment (TOE) framework, the Unified Theory of Acceptance and Use of Technology (UTAUT), and the Dynamic Capabilities Framework (DCF) to a resource-constrained developing-country aviation context and proposes a stakeholder-informed, phased BCT adoption framework—not yet tested through implementation—with direct policy and managerial implications for the Nigerian airport administration. Full article

HERA-GM separates an offline learned tactical proposal from the authority to execute it. A Dueling Double DQN trained with Conservative Q-Learning proposes one of five tactical actions. The runtime process then uses the action-value margin, semantic feature availability, annotation density, a Mahalanobis diagnostic, and fixed hard-rule conditions to assign ACCEPT, DEFER, or RECOVER. The study separately examined proposer agreement, authority changes, closed-loop outcomes, and held-out-family discrimination. The original frozen evaluation used 113 nuPlan Mini scenarios from 39 logs and 1970 closed-loop runs. An additional exploratory behavior-cloning block added 339 runs. Behavior cloning had slightly higher offline macro-F1 than CQL, whereas DDQN without CQL had much lower agreement under the tested configurations. The main comparison between M1 and the simpler B3 gate showed no supported primary safety difference, indicating limited added endpoint effect from Mahalanobis and hard-rule evidence in this cohort. M1 also showed lower safety-failure and drivable-area violation rates than behavior cloning, but with lower conditional progress; the primary result did not remain below 0.05 after pooled Holm adjustment across the five clean comparisons. The Mahalanobis score did not distinguish held-out semantic families reliably. The findings describe the operating trade-offs and limits of conditional execution authority rather than a safety guarantee. Full article

Early and accurate cancer detection from medical imaging remains challenging because clinically relevant evidence is distributed across local image appearance, structural relationships between suspicious regions, and ordered imaging context. This study proposes a graph-aware and sequence-aware deep learning framework that combines a convolutional neural network (CNN) backbone for spatial feature extraction, a Graph Attention Network (GAT) for lesion-structure modelling, and a Bidirectional Long Short-Term Memory (BiLSTM) module for ordered-view or slice-sequence representation learning. Cross-attention-based multimodal fusion is evaluated exclusively for the RSNA mammography task, where the metadata branch is restricted to patient age and implant status, both available before diagnosis. In contrast, the primary LIDC-IDRI experiment is conducted as an image-only analysis because radiologist malignancy scores and semantic nodule attributes are annotation-derived variables and are not treated as independent clinical predictors. The framework is evaluated on the RSNA Breast Cancer Detection dataset and the LIDC-IDRI lung CT dataset using accuracy, precision, recall, F1-score, specificity, and area under the receiver operating characteristic curve (AUC). Additional ablation experiments assess the contribution of graph learning, sequence-aware modelling, and leakage-safe RSNA metadata fusion, while SHAP analysis quantifies the influence of the included RSNA metadata variables on multimodal predictions. For the RSNA multimodal experiment, the best held-out run achieved 95.2% accuracy and an AUC of 0.978, while three repeated runs yielded 94.0 ± 0.3% accuracy and an AUC of 0.970 ± 0.007 under the internal patient-wise benchmark protocol. These results should be interpreted as public-dataset benchmark outcomes rather than evidence of real-world clinical performance; external multi-centre and prospective validation is required before clinical deployment. Full article

Spray application in spatially heterogeneous three-dimensional crop canopies is constrained by inadequate within-canopy deposition, off-target losses, and poor matching between operating parameters and target structure. This review synthesizes the technological evolution of intelligent agricultural spraying equipment from the perspectives of airflow-assisted transport, demand-based dose allocation, target sensing, actuation control, system integration, and performance evaluation. Air-assisted spraying has expanded transport regulation from hydraulic output alone to coordinated airflow–droplet–canopy interactions. Variable-rate and prescription approaches subsequently linked dose allocation to measurable spatial differences in canopy structure and target distribution. Ultrasonic sensing, light detection and ranging (LiDAR), red–green–blue depth (RGB-D) imaging, machine vision, and multimodal perception have increased the spatial and semantic resolution of target representation. Pulse width modulation (PWM), fuzzy control, and data-driven methods have improved nozzle-level actuation and compensation for nonlinear hydraulic dynamics. However, closed-loop actuator or navigation should not be conflated with closed-loop control of application quality: deposition, drift, and biological efficacy are still evaluated mainly after operation and are rarely used as real-time feedback variables. Current limitations arise less from the absence of individual sensing or control technologies than from weak coupling among canopy representation, spray transport, multivariable actuation, and performance feedback. Future development should prioritize transferable multiscale models and explicit coordination of liquid flow, airflow, droplet size, speed, and equipment state. Scenario-specific architectures must connect sensing, decision-making, actuation, and evaluation without sacrificing field robustness. Full article

Regression testing in continuous integration (CI) helps preserve delivered software value by exposing regressions early, but limited feedback windows make test order consequential. This study evaluates a leakage-safe dynamic test case prioritization framework that reconstructs job–build–commit provenance, combines strictly prior-build history with change–test similarity, applies class-conditional conformal calibration, and evaluates an operator-level deep Q-network (DQN). The primary held-out evaluation contains 273 failure-bearing jobs and 45,774 job–test rows from four Java projects. Pretrained history–semantic fusion achieved an equal-project macro failing-test-entity APFD surrogate (FTE-APFD surrogate) of 0.8840, compared with 0.8689 for history-only and 0.7023 for pretrained semantic-only ranking; the fusion–history difference did not survive Holm correction. A train-only TF–IDF control reached 0.7502 for semantic-only ranking and 0.8871 when fused with history in a seed-averaged diagnostic, so the benchmark does not establish unique superiority of the pretrained representation. A protocol-aligned RETECS reimplementation achieved 0.8296 ± 0.0111 FTE-APFD surrogate across five seeds. Adaptive conformal calibration reduced the candidate fraction from 0.3171 to 0.2598 while retaining 0.9058 failing-test coverage. The candidate-informed gated DQN remained below strong deterministic fusion rankings at 10%, 25%, and 50% budgets; five-seed, short-budget, hyperparameter, pre-execution-budget, and project-exclusion sensitivities did not establish a consistent DQN advantage. An additional three-project operational analysis containing 444 passing test jobs showed high failing-row coverage but strongly project-dependent candidate burden. Overall, the evidence favors strong leakage-safe deterministic rankings with explicit uncertainty control, while the evaluated reinforcement-learning design remains a local negative result rather than a general conclusion about reinforcement learning. Full article

Underground coal-mine equipment operates under broadband noise, high dust, humidity, and methane. Acoustic sensing is uniquely suited to this environment: it captures vibration, friction, and airflow signatures without physical contact, incurs low sensor-deployment cost, responds at millisecond speed, and remains effective in low-light, high-dust conditions where optical and vibration alternatives fail. Acoustic fault perception is therefore critically important in underground coal-mine operations. Three unresolved challenges remain: (i) motor whine, material-collision impacts, and ventilation-fan roar compound into a low-SNR soundscape where conventional models lose noise robustness; (ii) acoustic signatures vary widely across equipment types and fault-development stages; and (iii) existing supervised classifiers, trained on imbalanced data, exhibit limited generalization and output only binary judgments, lacking the semantic descriptions that maintenance crews actually need. To address all three, we propose a domain-adaptive audio LLM coupling a BEATs encoder (frozen during Stage II, adapted via Low-Rank Adaptation (LoRA) during Stage I), a Querying Transformer (Q-Former) alignment layer with Dynamic Acoustic Token Compression (DATC), a LLaMA-3.1-8B decoder adapted via LoRA, and Constrained Decoding for Structured Fault Description (CD-SFD) enforcing a three-slot output of fault type, danger level, and handling recommendation. DATC allocates query budget by signal energy to suppress noise-dominated frames; CD-SFD is a state-machine decoder that guarantees the three-slot schema. We release CMEASD: 1200 recordings comprising 24 physical machines (4 per equipment type, 12/6/6 machine-disjoint split). Under machine-disjoint evaluation, the model reaches Macro Accuracy 84.6 ± 1.5% and Macro F1 82.9 ± 1.6%, outperforming the strongest discriminative baseline (PANNs-Transformer, 81.7%) by +2.9 pp and SALMONN-LoRA by +2.5 pp. Ablations attribute +2.4/+1.4/+0.9 pp to DATC, CD-SFD, and the domain prompt. A 30-day mine trial achieves 7.0 s end-to-end latency with 21/30 days of stable, zero-false-shutdown operation. Full article

Resource misallocation in data governance is an overlooked source of carbon emissions, since most prioritization methods disregard the environmental cost of computing and human effort. This study proposes a four-layer hybrid intelligence framework (FLHIF) that unifies cloud model uncertainty quantification, social network analysis of data lineage graphs, DeepSeek-V4-driven multi-agent semantic reasoning, and adaptive weight learning via proximal policy optimization (PPO). On a benchmark of 439 data asset instances, FLHIF achieves an NDCG@10 of 0.9500, outperforming all baselines. Beyond ranking accuracy, expert evaluation yields an interpretability score of 4.2/5.0 (ICC > 0.80), indicating that FLHIF’s recommendations are auditable as well as accurate, a requirement of growing importance for AI adoption in regulated governance settings. Holdout stability tests, sensitivity analysis, and zero-shot generalization across finance, healthcare, manufacturing, and government sectors further demonstrate the framework’s robustness across the tested benchmarks. Preliminary simulations indicate that FLHIF can reduce redundant governance operations by 15–22% in a typical enterprise setting, corresponding to an estimated annual reduction of 120–180 kg CO₂-equivalent emissions per medium-sized data platform. FLHIF thus offers a systematic, resource-efficient, explainable, and adaptive approach to sustainable data governance, directly supporting SDG 9.1, SDG 11.3, and SDG 12.2/12.5. Full article

The rapid evolution of Large Language Models (LLMs) has opened new frontiers in cybersecurity automation, enabling intelligent agents capable of multi-step reasoning, tool invocation, and autonomous decision-making across complex security tasks. While individual applications have emerged across threat intelligence, vulnerability assessment, penetration testing, and security operations center (SOC) automation, a systematic understanding of the LLM-based agent paradigm in cybersecurity—encompassing both single-agent and multi-agent architectures—remains lacking. This paper presents a systematic literature review following PRISMA guidelines, identifying records through 59 structured web-search queries whose results resolve predominantly to arXiv, Semantic Scholar, the ACM Digital Library, IEEE Xplore, USENIX, MDPI, SpringerLink, and Elsevier ScienceDirect, supplemented by citation chaining, for works published between January 2022 and April 2026; the full query record is published with the paper. We applied structured inclusion and exclusion criteria and classified 59 primary studies along five dimensions: security function, agent architecture pattern, knowledge augmentation strategy, human-in-the-loop posture, and evaluation rigor. Our analysis reveals that penetration testing and threat intelligence are the most extensively studied domains, while incident response and compliance verification remain critically underrepresented. Penetration testing alone accounts for over half the corpus (50.8%). Single-agent tool-calling remains the most prevalent architecture (30.5% of studies), whereas centralized multi-agent orchestration—present in 18.6%—yields the strongest reported performance gains, up to 4.3× on zero-day exploitation; prevalence and performance therefore point in opposite directions. No included study achieves production-grade (E4) evaluation: the entire field currently rests on controlled laboratory assessments. An independent search of six bibliographic databases recovers 86.3% of the studies the primary search had surfaced (79.7% of the full corpus) while indicating a total eligible literature of roughly 400 studies, so the corpus is reported as a documented subset rather than an exhaustive census. We propose a unifying taxonomy, identify cross-cutting challenges including hallucination, prompt injection, and benchmark fragmentation, and outline open research directions with particular emphasis on multi-agent orchestration design. Financial sector applicability under DORA and the EU AI Act is treated as a documented evidence gap rather than a synthesis: the corpus’s only compliance and risk assessment study is also its only banking-specific system. Full article

Automated surface defect detection in hot-rolled steel is a prerequisite for real-time quality control, yet most deployed inspection systems operate in isolation from Industrial Internet of Things (IIoT) infrastructure. This paper reports a laboratory-scale proof of concept with two contributions. The first is a segmentation study on the Severstal dataset using a leakage-free, defect-stratified split of 1886 test images. Because a trivial all-background predictor already attains 96.66% pixel accuracy, performance is reported through Dice, IoU, precision, recall, and F1 with 95% confidence intervals. A compact from-scratch U-Net (0.49 M parameters) reaches a Dice of 0.416 at 38.6 ms per image, an ImageNet-pretrained DeepLabV3+ model reaches 0.677 at 46.3 ms and 37 times the parameters, and a classical Otsu baseline reaches 0.060, bracketing an explicit accuracy-versus-footprint design space rather than a single recommended model. The second contribution is architectural: a three-layer IIoT architecture whose messaging layer is empirically characterized on a Raspberry Pi broker over 158,500 messages. A factorial experiment isolates the transport configuration of the broker, rather than that of the publisher, as the determinant of end-to-end latency, yielding a seventeen-fold reduction. The layer sustains 1920 messages per second without loss, and a deliberate broker outage shows that MQTT delivery guarantees are semantic rather than temporal, motivating an application-level message-expiry policy. Embedded inference deployment is identified as the primary next step. Full article

Reliable localization and mapping are critical for intelligent robotic systems operating in dynamic indoor environments, where pedestrians and other moving objects can lead to erroneous feature associations, map contamination, and accumulated trajectory drift. To address these challenges, this study proposes the Boundary-Protected Semantic-Geometric Dynamic-Probability (Boundary-SGDP) framework, an enhanced red–green–blue-depth (RGB-D) visual simultaneous localization and mapping (SLAM) system based on boundary-protected semantic–geometric dynamic-probability estimation. The proposed method combines instance-level semantic priors generated by the YOLO26n-seg detector, a segmentation-oriented model in the You Only Look Once (YOLO) family, and the Segment Anything Model 2 (SAM2) with morphological region decomposition and RGB-D depth-edge detection. Potentially dynamic regions are further divided into dynamic interiors, semantic boundary protection bands, and geometrically informative depth-edge regions. Semantic and geometric cues are integrated to estimate a dynamic score for each feature, which is subsequently propagated to the MapPoint level as a dynamic probability. During pose optimization, these probabilities are used to adaptively adjust the weights of reprojection constraints, thereby reducing the influence of motion-contaminated observations while preserving geometrically valuable features around object boundaries and occlusion regions. Unlike conventional hard semantic masking strategies, Boundary-SGDP provides a soft and adaptive mechanism for handling dynamic observations. Experiments conducted on four dynamic walking sequences from the TUM RGB-D benchmark demonstrate that the proposed method achieves lower absolute and relative trajectory errors than the original ORB-SLAM3 system, while retaining substantially more boundary-related features. The results confirm the effectiveness of semantic–geometric fusion and boundary protection for robust visual localization and mapping in dynamic indoor scenes, and demonstrate the potential of the proposed framework for practical autonomous navigation and intelligent perception applications. Full article

The rapid advancement of autonomous agricultural systems (AASs) is transforming modern agriculture, where labor shortages, sustainability imperatives, and demands for precision farming are driving the adoption of intelligent agricultural platforms. Agricultural production environments present uniquely challenging conditions for autonomous agricultural systems, including unstructured and dynamically changing terrain, biologically variable targets, unpredictable illumination and weather conditions, and safe human–machine coexistence. This review systematically investigates three cornerstone technologies: multi-source perception, intelligent decision-making, and precision control. Furthermore, typical agricultural operations, including soil tillage, planting, irrigation and drainage, fertilization, plant protection, harvesting, and agricultural product processing, are reviewed to illustrate their applications. Based on representative operational scenarios, the research progress and application characteristics of intelligent equipment in environmental perception, operational optimization, and control execution are summarized. Specifically, multi-source perception is evolving from isolated sensor-based acquisition toward multimodal and deep learning-enabled semantic scene understanding. Intelligent decision-making has evolved from experience-driven approaches toward physics-informed, data-driven, and knowledge-enhanced frameworks for adaptive operational optimization. Precision control has progressed from conventional PID control toward adaptive, learning-based, and digital twin-enabled control strategies, achieving robust high-precision closed-loop regulation. However, several critical challenges persist: limited cross-domain generalization and robustness of perception models under environmental distribution shift, constrained interpretability and trustworthiness of data-driven decision systems, and insufficient adaptability of control architectures under multi-disturbance coupled field conditions. To address these gaps, future research should prioritize multi-source heterogeneous data fusion and standardization, collaborative control frameworks integrating mechanistic knowledge with data-driven learning, and explainable artificial intelligence combined with agricultural domain expertise—advancing toward genuinely autonomous, trustworthy, and resilient agricultural systems. Full article

With the advancement of Earth observation technology and the improvement of multisource data acquisition capabilities, heterogeneous remote sensing image change detection technology has become increasingly important. However, existing heterogeneous remote sensing image change detection methods rely largely on fixed network architectures, which makes adapting to complex modal differences and severe noise interference difficult. To address these issues, in this paper, a decoupled differential search-based graph change detection network (DDS-Net) is proposed. First, the model designs a decoupled differential search-based dual-stream graph encoder (DDSGE). By decoupling the search space from the optimization strategy, it automatically optimizes feature extraction operators and graph topologies for different modalities, thereby significantly increasing feature adaptability while reducing computational complexity. Second, to address nonlinear geometric distortions between heterogeneous images, in this paper, a heterogeneous spatiotemporal alignment module that is based on differential localization search (HSTAM) is proposed. This module uses a local soft attention mechanism to dynamically correct registration errors in the feature space. Furthermore, to suppress erroneous graph connections caused by noise, structural consistency and smooth denoising (SCSD) loss is introduced, and deep semantic feedback and graph smoothing regularization constraints are collaboratively used to dynamically generate graphs, thereby effectively increasing the internal consistency of the transformed graph and suppressing misconnection noise. Extensive experimental results demonstrate that this method significantly improves the robustness and accuracy of the model in complex registration error scenarios while maintaining computational efficiency. Full article

Accurate monocular pose estimation of spacecraft is essential for on-orbit servicing and proximity operations, yet this remains challenging because of large variations in the target scale and limited onboard computational resources. To address these challenges, this paper proposes SAPose, a lightweight geometry-guided pose-estimation framework that integrates multi-task keypoint prediction with geometric pose recovery. The proposed network jointly performs spacecraft detection and semantic keypoint localization using a stage-aware lightweight backbone and a scale-adaptive feature fusion structure, thereby enhancing geometric feature representation across different target scales. For distant spacecraft occupying only a small portion of the image, an adaptive coarse-to-fine inference strategy selectively activates ROI-based secondary refinement to recover weakened structural details while avoiding unnecessary computation for medium- and large-scale targets. In addition, confidence-ranked keypoint selection and object-space nonlinear refinement are employed to improve the stability and accuracy of pose recovery. Experiments on the Spacecraft Pose Estimation Dataset (SPEED) and the Spacecraft Keypoint Dataset (SKD) demonstrate that SAPose achieves competitive pose estimation accuracy with a compact model size and efficient inference, providing a practical balance between accuracy and computational cost for monocular spacecraft pose estimation. Full article