Search Results (7,019)

Small-scale and urban indoor farms typically rely on manual observation, which delays stress detection and yields inconsistent crop quality. While large language models (LLMs) and retrieval-augmented generation (RAG) have been explored for agricultural advisory systems, their integration into a single cloud-free indoor-farming platform that couples multimodal symptom interpretation with autonomous environmental control remains largely unexamined. This study presents an integrated platform built around an agentic AI advisory pipeline that runs entirely on-device on commodity hardware. The pipeline couples a RAG-grounded Mistral 7B language model with a LLaVA 7B vision-language model through condition-based routing, intent classification, multi-step reasoning, and an LLM validation gate, delivering context-aware text and image-based symptom-level guidance from a conversational interface. The advisory layer operates alongside vision-based plant monitoring and a deliberately isolated threshold-based control layer, in which an ESP32 microcontroller autonomously actuates irrigation and lighting against predefined thresholds while a Raspberry Pi 5 performs continuous plant detection and browning monitoring. On Cos lettuce, the advisory pipeline achieved 82.00% weighted accuracy across 50 queries spanning health, symptom, watering, pest, root-health, and growth-stage categories, scored against established plant pathology and postharvest references, with no incorrect responses recorded. The study contributes the design of an agentic advisory pipeline and its integration into a working, cloud-free indoor-farming platform, providing an on-device foundation for intelligent small-scale farming. Full article

Virtual reality (VR) applications are often designed as episodic experiences, with limited support for persistence, longitudinal revisitation, and structured integration of interaction data across sessions. This paper presents a persistent multi-user VR garden architecture that combines snapshot-based state restoration, structured event, movement, and transcript records, an asymmetric owner–visitor workflow, cloud-mediated speech transcription, and deferred AI-supported synthesis. The architecture separates the current spatial configuration of the environment from the interaction traces through which it evolves, supporting repeated access, state restoration, historical consultation, and post-hoc processing. A controlled technical validation using synthetic or researcher-generated inputs was conducted through Unity Editor/backend tests and on Meta Quest 3 hardware. Persistence was evaluated at 20, 100, and 250 objects in the Editor and at 1, 50, and 100 objects on Quest, with two Quest replicas per load. Application-level visitor restrictions were verified across nine prohibited write operations. A frozen production speech-to-text corpus completed 40/40 requests with a micro-averaged word error rate of 9.50% and a median end-to-end latency of 1.619 s. The deferred AI pipeline was additionally verified as a functioning technical integration, while limitations in semantic-detail preservation were observed. The results support the implementation-level feasibility of the proposed persistent and traceable VR architecture. The present evaluation does not establish backend-level authorization guarantees, general AI or NPC-grounding performance, user outcomes, or clinical effectiveness. Full article

In rock reinforcement systems, bond stress at the joint between rock bolts and grout exerts a crucial influence on determining the bonding capacity of rock bolts. However, much less research has been conducted to study the joint bond stress (JBS) distribution with analytical deduction. Therefore, this paper adopted an analytical model to evaluate JBS distribution. The novelty of this paper is that the JBS distribution under different grouting conditions can be quantitatively studied. Based on variable controlling techniques, the influence of different parameters on JBS distribution state was studied. These studied parameters included joint strength, joint remaining strength and relative slide at joint remaining strength. Results showed that after rock bolts were loaded, the loading process can be basically divided into five different stages. Joint strength exerts a crucial influence on determining the JBS distribution state before the joint fully disconnected. After the joint started disconnecting, larger joint strength led to a slower propagation speed of the maximum JBS. When the remaining strength of the joint increased, after the whole joint deformed plastically, the disconnected length became longer, and frictional resistance in the disconnected section became larger. This led to an increase in the bonding capacity of rock bolts. Compared with joint strength and joint remaining strength, relative slide at joint remaining strength had a slightly smaller influence on JBS distribution. Before the joint disconnected, increasing relative slide at joint remaining strength was likely to improve the propagation speed of the maximum JBS. This study is valuable for further understanding the stress-transferring mechanism in rock reinforcement systems. Full article

Promoting morality among youth is an important educational priority in Thailand’s Southern Border Provinces. However, few culturally grounded learning interventions based on the Sufficiency Economy Philosophy have been systematically developed and evaluated. This study aimed to develop and preliminarily evaluate the Digital Learning Innovation for Strengthening Society (DLISS), a culturally grounded learning intervention designed to promote self-reported morality among youth. A four-phase research and development design was employed, including needs assessment, intervention and instrument development, preliminary field implementation, and post-implementation expert appraisal. Self-reported morality was assessed using the Sufficiency Economy Moral Scale (SEMS), which underwent psychometric evaluation with 1640 youth. Preliminary implementation involved 22 youth leaders assessed at pretest, immediate posttest, and post-reinforcement assessment. The SEMS demonstrated satisfactory psychometric properties. Participants showed statistically significant within-participant changes in the five measured dimensions of self-reported morality across the three assessment occasions. Experts rated the DLISS model favorably for its appropriateness, feasibility, utility, and contextual relevance. The DLISS model represents a promising culturally grounded learning intervention for promoting self-reported morality among youth. The findings provide preliminary evidence supporting further evaluation using larger samples and controlled research designs. Full article

High-frequency induction hardening enhances the surface wear resistance and contact fatigue life of high-speed linear guides, but simultaneously produces an inhomogeneous, layered cross-sectional structure comprising a high-strength, low-ductility outer hardened layer and a low-strength, high-ductility inner core. This structural heterogeneity renders conventional straightening stroke prediction models—predicated on homogeneous material assumptions—fundamentally inadequate. Moreover, the iterative trial-bending operations ubiquitous in industrial practice progressively accumulate plastic strain, causing guide rails to exhibit erratic positive-to-negative deflection reversal during sequential straightening passes. To address these critical challenges, this study proposes a novel two-stage hybrid straightening strategy based on inverse finite element analysis (FEA) and closed-loop experimental feedback. An equivalent hardened layer depth (HD₀) is introduced as a parametric descriptor to construct a layered elastoplastic finite element model, and an inverse simulation strategy is developed to generate a comprehensive three-dimensional stroke–residual deflection prediction dataset encompassing both vertical and lateral straightening conditions across multiple support spans. Displacement-controlled three-point bending experiments validate the layered model and elucidate the mechanism by which cumulative plasticity progressively amplifies cross-sectional plastic sensitivity under repeated loading. Grounded in this physical insight, a hybrid straightening algorithm is formulated, combining dataset-driven initial stroke prediction for rapid large-deformation elimination with an upper-bound constraint and a measurement-feedback-driven sequential reduction compensation scheme for fine-tuning. Comparative experiments demonstrate that the proposed strategy effectively suppresses the oscillatory over-straightening characteristic of conventional empirical trial-and-error approaches, consistently reducing residual deflection below 0.05 mm within two to three loading cycles. This work bridges the gap between theoretical simulation and the complex physical state of actual machining, substantially improving both the efficiency and precision of straightening for guide rails with induction-hardened layers. Full article

Polypropylene (PP) fiber-reinforced soil slopes undergo progressive resistance degradation under wet–dry cycling (WDC), while stochastic seismic loading introduces additional uncertainty, challenging deterministic seismic assessment. This study develops a time-dependent seismic reliability framework integrating the probability density evolution method and the equivalent extreme value event method. The cohesion and internal friction angle of unreinforced soil measured at different WDC states are represented as cross-correlated lognormal random fields and combined with random fiber configurations and weighted nonstationary stochastic ground motions in a nonlinear dynamic model. The main contribution is a unified uncertainty-propagation scheme that incorporates experimentally characterized WDC degradation and multiple uncertainty sources into the evolution of response probability and multilevel first-passage reliability. With increasing WDC number and PGA, the extreme displacement distributions shift toward larger values, accompanied by increased response dispersion, tail risk, and reliability loss. The reliability evolution exhibits three stages, namely initial stability, rapid degradation, and residual convergence, during the 70 s excitation. PP fiber reinforcement improves reliability, although the marginal gain becomes limited when the fiber content exceeds 0.15% under the present numerical conditions. The proposed framework provides a probabilistic basis for the seismic assessment and deformation control of PP fiber-reinforced soil slopes at different WDC degradation states. Full article

Canopy 3D architecture is a critical determinant of light interception, photosynthetic efficiency, and final yield in cotton, yet its rapid and accurate characterisation remains challenging in field conditions. To achieve efficient, non-destructive, and quantitative monitoring of the canopy structure of field-grown cotton, this study proposes a 3D structure-based technology stack for high-efficiency, low-cost, and high-precision phenotyping extraction. This stack directly addresses the technical bottlenecks of traditional 3D data acquisition, namely high cost, long processing time, and low operational efficiency, which have hindered large-scale application. We developed a pipeline that integrates a fast reconstruction algorithm with a scale-recovery mechanism using ground control points (GCPs), enabling the generation of true-scale 3D point clouds from UAV aerial images in a cost- and time-effective manner. Using only 141 UAV images and with a reconstruction time of approximately 20 min, we efficiently reconstructed high-quality, scale-accurate point clouds of two 5.5 m × 5.5 m cotton plots, significantly outperforming SfM-MVS and Instant-NGP in terms of both reconstruction efficiency and point cloud completeness. This method, whose current validation is confined to a single season, one growth stage, and two experimental plots, not only achieves a breakthrough by using fewer input images with high efficiency, but also ensures point cloud accuracy and completeness, showing strong potential for rapid field monitoring and real-time management. Based on the high-quality reconstructed point clouds, we further quantitatively evaluated canopy characteristics at harvest, analyzing the coefficient of variation of canopy height, porosity distribution, and canopy volume fraction. The core shortcomings and optimization strategies for the existing canopy structure were identified, providing scientific data support and practical technical references for precision cultivation management and mechanization-compatible planting in cotton. Full article

Photovoltaic–thermal (PVT) collectors coupled to thermochemical energy storage (TCES) can turn intermittent low-grade solar heat into a dispatchable service, but solar intermittency poses a closed-loop control problem. A companion study established the feedback-only lower bound: a 937 kJ accumulated exergy-delivery-deficit benchmark under optimally tuned proportional–integral–derivative (PID) flow control. The corresponding upper bound is quantified here by means of a deliberately idealised search-based predictive controller that, at each 10 s step, enumerates 51 candidate pump rates, predicts the reactor-inlet temperature by a single forward-Euler step, and is granted perfect future irradiance. On the experimentally validated shared plant (matched to the companion baseline), against an optimally tuned PID, the perfect-foresight advantage is marginal: +0.96% daily exergy on synthetic days and +0.07–0.24% on two measured Walailak University monsoon days, all controllers tracking within 6–13 K on the measured days. Under tropical-monsoon irradiance, the 95 °C desorption setpoint is rarely sustained, so the delivered exergy is nearly controller-independent: the perfect-foresight upper bound lies just above the feedback-only lower bound, and together the two results bracket the exergy envelope available to any flow-rate controller of this system. A horizon sweep localises the bottleneck to internal-model fidelity, not anticipation depth. The eight-node plant is validated against measured module temperature (root-mean-square error 3.5 °C, coefficient of determination R² = 0.89) and a copper-tube PVT prototype (1.5 °C; peak hot water up to 79 °C). The central contribution is therefore a rigorously defined, experimentally grounded upper bound showing that, at this scale and latitude, deployability rather than anticipation is the effective design lever. Full article

Parkinson’s disease (PD) is associated with alterations in gait symmetry and plantar loading that can be examined using vertical ground reaction force (VGRF) recordings. This study presents a confidence-stratified, physiology-anchored multiple-instance learning framework with concept-bottleneck-inspired pathways (implementation identifier: DRO-PAS-MIL-CBM; hereafter, PAS-MIL) for retrospective session-level PD-versus-control classification. Each gait session is represented as a bag of temporal windows. Eight predefined bilateral signal descriptors are combined with eight learned latent temporal dimensions, aggregated through attention-based pooling, and processed by concept-guided, prototype, anchor-only, and static-feature expert pathways. The evaluation used five-fold person-grouped cross-validation on 306 sessions from 165 participants in the PhysioNet Gait in Parkinson’s Disease database.Inner person-grouped out-of-fold ExtraTrees probabilities were used to construct the confidence strata and distillation targets. PAS-MIL achieved a pooled session-level area under the receiver operating characteristic curve of 0.771, average precision of 0.890, and a mean fold AUC of $0.826\pm 0.041$. Relevance analysis identified C05 (asymmetry variability) and C08 (bilateral change mismatch) as the highest-weighted predefined physiological anchor descriptors. Protocol-stratified sensitivity analysis showed variation across the three source sub-studies, with AUCs ranging from 0.740 to 0.790. Probability calibration remained suboptimal after temperature scaling (mean per-fold ECE, $0.291\pm 0.042$). The results demonstrate the feasibility of integrating physiology-informed descriptors, temporal representation learning, and session-level aggregation. The study is a retrospective proof of concept and does not establish external robustness or clinical deployment readiness. Full article

Smart-environment cybersecurity increasingly depends on experimental platforms that can reproduce attacks against buildings, homes, and cities under realistic conditions. However, the literature remains fragmented across testbed design, attack demonstration, and defensive validation. This makes it particularly difficult to judge what kind of security evidence each study actually provides. This review systematically analyses 28 experimentally grounded studies published from 2020 onwards, focusing on how testbed realism, cyber–physical coupling, and evaluation mode shape the strength of the resulting claims. The corpus spans physical, hybrid, emulated, and dataset-driven environments across smart buildings, smart homes, and smart cities. Through our investigation, we discern a clear asymmetry in the field. Detection-oriented studies dominate, especially those based on emulation or public datasets, while live evidence for prevention, response, containment, and recovery is comparatively scarce. Availability and integrity/control attacks are the most frequently exercised, whereas authentication compromise and software exploitation remain rare because they are harder to stage on real hardware. Moreover, an important observation we arrive at is that physical and hardware-in-the-loop platforms support the strongest cyber–physical evidence, but emulated and replayed environments remain valuable for scale and reproducibility. At the same time, public datasets and offline classification results do not by themselves establish operational resilience in a live smart environment. To make these distinctions explicit, we introduce a cross-domain taxonomy of testbed architectures, attack families, and defensive control coverage, and map the evidence strength of reported mitigations using NIST cybersecurity framework-derived operational functions. Last, we identify open challenges, including weak recovery evaluation, limited reuse of reference testbeds, and the need for live, context-aware datasets, outlining promising future directions. Full article

Retrieval-augmented fine-tuning (RAFT) can improve small language models (sLMs) on retrieval-grounded question answering, but the synthetic training data produced by commercial large language models (LLMs) vary in quality. This paper contributes a quality-aware selection protocol—rather than a new RAFT or QLoRA method—that scores LLM-generated alternatives along four embedding-based dimensions (question relevance, answer faithfulness, QA coherence, and semantic similarity) and selects one alternative per task before parameter-efficient fine-tuning. Under pre-specified paired-bootstrap contrasts with Holm correction, the parameter-free faithfulness-based selector only-AF significantly exceeds random selection on Gemma-2-9B-IT (ΔF1 = +0.106, 95% CI [+0.043, +0.174], Holm-corrected p = 0.019), and its pre-specified weighted companion af-70 (w_AF = 0.70) shows the same confirmed pattern (Holm-corrected p = 0.002). Both effects persist under a Korean character-level F1 that removes particles and punctuation (Holm-corrected p = 0.004 and p = 0.042), indicating robustness to the choice of lexical metric. Relative to training on the full 150-row augmented pool, the quality-selected 50-row sets are statistically indistinguishable while using one third of the training data, which we interpret as data efficiency rather than superiority. Across six instruction-tuned models (2B–27B), a significant selector-by-model interaction indicates that the optimal quality axis is model-dependent, and the two smallest models show no benefit from selection. The study’s confirmatory contrasts use a small controlled Korean corpus under a transductive design; two pre-registered validation experiments probe external validity. On an independent five-fold larger corpus with a passage-level train/test split, fine-tuning transfers strongly and the selected one-third subsets show no significant difference from the full pool, while the advantage over random selection is directionally positive but small and not significant; under controlled corruption of 35% of the pool, the metrics detect the damaged rows, and for the score-sum selector the selection-versus-random benefit is significantly larger than on the clean pool (difference-in-differences p = 0.0014; directionally consistent but not significant for the faithfulness selectors). Within this scope, quality-aware selection is a promising, data-efficient safeguard for synthetic RAFT data—performing comparably to full-pool training at one third of the cost, with growing value as pool quality degrades—and larger-scale external validation remains future work. Full article

The core predicament of energy transition lies not in the availability of clean technologies, but in whether an economy possesses the institutional capacity and social foundation to systematically regulate its carbon-energy metabolic processes. Drawing upon co-evolutionary theory from evolutionary economics, this paper constructs a composite indicator of Low-Carbon Green Energy Eco-Co-evolution (LCEE) encompassing three functional dimensions: efficiency advancement, kinetic energy replacement, and boundary adherence. Concurrently, by integrating innovation ecosystem theory with inclusive development theory, we propose the concept of “Inclusive Potential of Regional Innovation Ecosystems” (IEP), characterizing the systemic potential for transforming innovation outcomes into social welfare across four dimensions: Knowledge Matrix Abundance (KMF), Cultural Capillary Permeation (CCP), Technological Community Succession (TCS), and Social Root Nourishment (SRN). Taking China’s 2016 intellectual property (IP) powerhouse construction pilot as the institutional prototype of Industrial Intellectual Property Upgrading Reform (IPR), we incorporate IPR, IEP, and LCEE into a unified causal analytical framework, proposing a testable transmission logic of ‘institutional supply → ecological development → co-evolutionary synergy. Using panel data from 30 Chinese provincial-level administrative regions over 2010–2022, we employ a Spatial Durbin Difference-in-Differences (SDM-DID) model to identify the direct and spatial spillover effects of IPR on LCEE, and embed a Double Machine Learning (DML) framework to test the mediating mechanism of IEP while controlling for high-dimensional nonlinear interference. The findings reveal that IPR exerts a significant and robust direct promoting effect on LCEE, generating positive spatial spillovers to neighboring regions through the public disclosure of patent information. IEP significantly promotes local LCEE, yet its spatial spillover lacks statistical support due to structural conflicts in inter-dimensional transmission attributes. IEP plays a significant partial mediating role between IPR and LCEE, with the indirect effect accounting for over one-third of the total effect, a finding robust to alternative machine learning algorithms, sample split adjustments, and exclusion of contemporaneous competing policies. Sub-path tests reveal that KMF bears the strongest mediating efficacy, serving as the primary transmission channel, while CCP exhibits full mediation—the institutional effect on LCEE in the cultural dimension depends almost entirely on the mediating transformation through the public cultural service system. Heterogeneity analysis further demonstrates full mediation in the Low-Carbon Green Energy Eco-Kinetic Replacement (KER) dimension, indicating that the institutional catalytic effect on clean energy substitution must be realized through IEP transformation. This paper provides empirical evidence for the proposed causal pathway through which institutional public goods indirectly enhance the synergistic quality of carbon-energy transition via the inclusive potential of innovation ecosystems, providing theoretical foundations and policy implications that, while grounded in China’s institutional context, may offer valuable reference points for emerging market economies facing similar dual pressures of technological constraints and green transition. Full article

Targeting the rock burst hazard induced by the hard roof key stratum during deep mining at the Mengcun Coal Mine in the Binchang mining area, this study takes the No. 403109 working face as the engineering background and systematically investigates the rockburst prevention mechanism and effectiveness of ground hydraulic fracturing through theoretical analysis, UDEC numerical simulation, and surface microseismic monitoring. The results indicate that fracturing pre-weakens the overlying key stratum, transforming its load-bearing mode from a long-beam rigid support to a segmented flexible support. This significantly reduces the cantilever length, lowers the accumulation of elastic strain energy, and enables flexible load transfer and stress redistribution in the overburden. Numerical simulations reveal that after fracturing, the breakage timing of the key stratum advances, the fragmentation size decreases, and the over-burden movement shifts from stepwise fracturing to sequential caving, with the stress concentration zone substantially narrowed. In the field, a total of 44 fracturing stages were implemented in wells MC-05L and MC-06L, creating a fracture network with an average fracture length of 317 m and an average fracture height of 55 m, achieving an effective stimulated volume ratio of 86.7%. During the mining period, microseismic events exhibited a median energy of only 868.14 J, characterized by high frequency and low energy. The average weighting interval was 13.69 m, the peak coal stress was controlled within 5.0–6.7 MPa, and the loads on roadway bolts and cables remained within safe limits. This study validates the source-control effect of ground hydraulic fracturing on working faces with strong rock burst risks in deep mining, providing a theoretical basis and engineering reference for mines with analogous conditions. Full article

External resources, digital competence, and psychological characteristics have each been linked to Artificial Intelligence (AI) literacy, but almost always in isolation, so the unique contribution of any one domain remains unknown. Grounded in social cognitive theory, this study models all three domains jointly. Data were collected from 303 undergraduates at a liberal arts university in Hong Kong and analysed using hierarchical multiple regression. Perceived resources and perceived teacher support explained 29% of the variance in AI literacy; prior digital competence added a further 3%; and personal innovativeness, technology anxiety, and growth mindset in technology added a further 6%, with all three being significant in the final model (total R² = 0.39). The external predictors remained significant throughout but attenuated substantially, indicating that institutional provision is necessary but not sufficient. Technology anxiety, which correlated negatively with AI literacy at the zero-order level, emerged as a positive predictor once other determinants were controlled. Analyses of the five AI literacy subdimensions localised this effect to critical evaluation and ethical competence and found no association with the three performance-oriented dimensions, suggesting that anxiety operates through heightened vigilance rather than enhanced operational skill. These findings, which no single-domain design could have produced, indicate that fostering AI literacy requires attention to students’ psychological readiness and foundational digital skills alongside the provision of infrastructure. Full article

Underground coal mining can induce deformation and failure of overlying strata and ground surface, which seriously endangers the safety of human life and property. During mining, the immediate roof rock successively experiences initial caving (fixed support on four sides) and periodic caving (fixed support on three sides and free on one side). Different boundary conditions alter the unloading and deformation processes such as cracking and fracturing of immediate roof rock, thereby affecting its subsequent mechanical behavior of compaction and deformation, and resulting in differences in the movement law of overlying strata. In this paper, the numerical simulation method is adopted to investigate the variation laws of unloading and fracturing characteristics of immediate roof rock under initial caving and periodic caving with thickness-width ratio (t/w), length-width ratio (l/w), unloading stress (σ_u) and specimen strength (σ_c), and the corresponding action mechanism is revealed. The fractal evolution law of fractured immediate roof rock obtained from this study can quantitatively evaluate the compaction characteristics of caved rock, provide refined parameter support for surface subsidence prediction and possess guiding significance for stope surrounding rock control engineering. The results show that the fragments formed after the failure of immediate roof rock are mainly block-strip shaped under both first caving and periodic caving conditions. With the increase in the thickness-width ratio, the flexural rigidity of immediate roof rock increases and crack propagation is restrained, so that the particle-size–mass fractal dimension of fragments increases first and then decreases for the two caving modes. The increase in length-width ratio weakens the propagation of secondary fractures and raises the particle size of fragments, while the overall variation in particle-size–mass fractal dimension is small under the two working conditions. As the unloading stress continuously rises, the coupled tension-shear effect inside the rock gradually intensifies, and the failure mode changes from tension-shear failure to global shear failure. Accordingly, both the particle-size–mass fractal dimension and fractal dimension of crack distribution increase first and then decrease under first caving and periodic caving conditions. The increase in the strength of immediate roof rock raises the energy consumption during rock failure, and large-size fragments are more likely to be generated, which reduces the particle-size–mass fractal dimension and increases the particle size of fragments under both caving modes. Meanwhile, internal micro-fractures continuously initiate and propagate with the growth of rock strength. For specimens with relatively high strength, crack propagation is inhibited and the development of secondary fractures is weakened, leading to an evolution trend that the fractal dimension of crack distribution increases first and then decreases. Under identical parameter conditions, the particle-size distribution and crack complexity for first caving are mainly affected by geometric parameters; the particle size of fragments is primarily controlled by specimen strength; and the unloading stress threshold governs the transition of failure mode. For periodic caving, the crack-initiation location is first determined by asymmetric boundary constraints. The thickness-width ratio dominates the particle-size distribution of fragments, and unloading stress as well as specimen strength further regulate the complexity of cracks. Full article