Efficient task allocation and coordination are critical for heterogeneous multi-agent systems operating in dynamic field environments. This paper presents a closed-loop framework that integrates Large Language Models (LLMs) with graph-based optimisation to enable end-to-end task decomposition, allocation, and adaptive execution. High-level task scripts are initially parsed by an LLM into structured execution flows, which are transformed into Directed Acyclic Graphs (DAGs) capturing action-level dependencies. A Genetic Algorithm (GA) then optimises agent-to-task assignments by minimising makespan under capability and battery constraints. To ensure robustness, the framework incorporates an LLM-driven recovery module that enables localised replanning under execution failures without interrupting unaffected agents. System-level experiments in a high-fidelity agroforestry simulation demonstrate a 37% increase (p<0.001) in harvesting productivity and a 19% reduction in human idle time compared to manual baselines. Under mid-execution failures, the system maintains significantly higher performance, with replanning latencies averaging 24 s. The framework scales to large fleets (up to 1000 agents) and effectively enhances human–robot collaboration through structured, dependency-aware coordination.

Fibers in vortex turbulence fields involve complex gas–solid coupling effects, significantly influencing the spinning process within vortex nozzles. This paper utilizes the Discrete Element Method (DEM) to refine the existing rigid bead–elastic rod model describing fiber constitutive behavior. This improved model is used to numerically simulate the dynamic behavior of a single flexible fiber within the vortex field of the nozzle. Based on elastic mechanics, this study establishes mapping functions converting relative displacement and angular displacement between beads into internal forces and torques within the beads. A contact model is also developed to handle fiber–wall interactions. The effects of different nozzle structures on fiber motion are investigated. The improved model successfully simulates the entire motion process of a single fiber during spinning. Its reliability is validated by comparing with experimentally collected fiber motion data. The study reveals that a twist chamber diameter of 6 mm, a conical cavity angle of 55 degrees, and a distance of 1.05 mm between the jet orifice and the hollow spindle yield optimal fiber twist count and wrapping density. This research provides effective insights for developing textile equipment that relies on airflow to drive fiber spinning and contributes to establishing a comprehensive twisting mechanism.

Control of dynamic technical objects in locked-loop systems with negative feedback is widely used in robotics and all complex technical devices. The theory of ensuring accuracy, speed, and control quality is well developed for the case of stationary objects. If the parameters of the controlled object change over time, the development of adaptive systems is recommended. Many articles and monographs have been written on the topic of adaptive control, but most often they provide only general indications that an additional external loop is necessary to change the controller parameters so that the modified controller more successfully ensures control accuracy and quality than its previous version. At the same time, methods for solving the two most complex sub-problems almost never were disclosed: determining how exactly the mathematical model of the object has changed and deciding how exactly the controller parameters should be changed to provide the required properties of the system as a whole. This paper proposes some general principles for designing adaptive systems and demonstrates their effectiveness using several examples.

Underwater radiated noise (URN) has attracted increasing attention due to its environmental impact, with cavitation recognized as the dominant source. This study investigates cavitation-generation mechanisms and associated noise radiation for open-type and gap-type wings using high-fidelity numerical simulations. Cavitation noise was predicted using the Ffowcs Williams–Hawkings (FW–H) equation. The Fitzpatrick–Strasberg bubble noise model was independently employed for analysis to relate cavitation dynamics and cavity-volume variation to the resulting acoustic emissions. The results show that the gap-type configuration produces significantly stronger low-frequency noise, with the Tip Leakage Vortex Cavitation (TLVC) contributing up to 15 dB/Hz higher noise levels than the Tip Separation Vortex Cavitation (TSVC). This enhancement is attributed to the strong interaction between TLVC and TSVC, which amplifies cavitation dynamics and acoustic emissions. Analysis of three gap sizes reveals that, for small gaps, this interaction induces periodic cavitation behavior, generating a distinct harmonic component at St ≈ 2. As the gap size increases, the TLVC-TSVC interaction weakens, and the cavitation behavior transitions toward that of the open-type configuration, leading to the disappearance of the tonal component. These findings highlight the critical role of gap-induced vortex interaction in determining URN characteristics.

Differentiable Neural Architecture Search has emerged as a powerful paradigm for automated network design, yet it suffers from a fundamental optimization inconsistency problem: Architectures optimized under continuous relaxation often fail to maintain their performance after discretization. To address this challenge, we propose Fisher-DARTS—a Fisher information-driven differentiable NAS framework. The proposed method introduces three key innovations: (1) a Fisher information-based momentum update mechanism that guides architectural parameters toward statistically significant operations, aligning the search objective with discrete deployment; (2) a progressive three-region pruning strategy that adaptively eliminates redundant operations with low Fisher information, ensuring architectural compactness; and (3) a cell-weighted fusion module that preserves multi-scale features across stacked cells. Additionally, the search space is expanded by incorporating attention mechanisms to enhance feature representation capability. The proposed framework is generic and applicable to a wide range of vision tasks. To validate its effectiveness, we apply it to gaze estimation—a core technology in multimodal human–computer interaction. Experimental results on three public datasets, MPIIFaceGaze, RT-GENE, and ETH-XGaze, demonstrate that Fisher-DARTS achieves mean angular errors of 3.22°, 5.45°, and 4.12°, respectively, outperforming hand-designed networks and existing NAS-based gaze estimation models. These results validate the effectiveness of the proposed Fisher-driven NAS framework and its generalization capability across diverse scenarios.

This study proposes an improved tree-augmented Bayesian network (TAN-BN) method for analyzing the severity of ship collision accidents by introducing the information contribution rate (ICR) for edge orientation and flexible filtering constraints for structure optimization. Based on 634 ship collision accident reports, a Bayesian network covering accident attributes and causal factors was constructed. The results show that the improved model achieved an overall AUC of 0.864, higher than that of the traditional TAN model (0.827). Mutual information analysis identified ship length as the factor most strongly associated with accident severity, with a mutual information value of 0.0868. Sensitivity analysis based on true risk impact (TRI) further showed that ship length, time, and ship type were the most influential factors, with average TRI values of 19.4%, 8.8%, and 7.2%, respectively. The proposed model effectively captures the dependency relationships between accident severity and multiple influencing factors and can provide quantitative support for risk warning and accident prevention in maritime traffic safety.

Background: Malocclusion can influence not only oral function, but also adolescents’ perception of dental appearance, psychosocial status, and oral health-related quality of life. The present cross-sectional study aimed to examine the association between the clinical severity of malocclusion and adolescents’ self-perception, as well as to assess the level of agreement between clinician and patient ratings of aesthetic impairment. Methods: The study included 160 consecutive adolescents aged 12 to 17 years who were referred for orthodontic evaluation. Malocclusion severity was determined using the Index of Orthodontic Treatment Need (IOTN), considering both the Dental Health Component (DHC) and the Aesthetic Component (AC). Patients’ perceptions of malocclusion severity, aesthetic concerns, and functional limitations were collected through numeric rating scales, while psychosocial impact was measured using the Psychosocial Impact of Dental Aesthetics Questionnaire (PIDAQ). Correlations were examined with Spearman’s coefficients, agreement between clinician- and self-rated AC scores was evaluated with weighted kappa statistics, and multivariable regression analyses were used to identify independent predictors of patient-reported outcomes. Results: Moderate positive correlations emerged between IOTN-DHC and self-perceived malocclusion severity (rho = 0.42, p < 0.001), and between clinician-rated IOTN-AC and self-perceived severity (rho = 0.47, p < 0.001). Associations were stronger for aesthetic and psychosocial measures than for functional complaints, with the highest correlation observed between clinician-rated IOTN-AC and perceived aesthetic impact (rho = 0.58, p < 0.001). Greater clinical severity was also associated with increased psychosocial burden, as reflected by higher PIDAQ total scores. Agreement between clinician and patient evaluations of aesthetic impairment was fair to moderate (weighted kappa = 0.34) but improved when AC scores were grouped into broader categories (weighted kappa = 0.46). Overall, adolescents tended to perceive their dental aesthetic impairment as more severe than did clinicians. In the multivariable models, clinician-rated IOTN-AC remained an independent predictor of perceived severity, aesthetic impact, and psychosocial burden. Conclusions: Clinical measures of malocclusion severity and adolescents’ self-perception were significantly related, although they reflected partly different aspects of orthodontic treatment need. Combining normative clinical indices with patient-reported outcomes may allow a more complete and patient-centred evaluation of malocclusion.

Providencia stuartii (Ps) is a clinically significant opportunistic pathogen often associated with “difficult-to-treat resistance” (DTR) infections due to pan-resistance to first-line antimicrobials. We report the clinical diagnosis and rapid genomic characterization of strain Ps-CMC-4104, recovered from a human splenic abscess in a patient with infected necrotizing pancreatitis. To resolve the complex genetic architecture of this strain, we utilized hybrid sequencing combining Oxford Nanopore (long-read) and Illumina (short-read) technologies. Analysis revealed a 4,504,925 bp circular chromosome featuring a unique genomic resistance island (GRI) closely related to Salmonella SGI1. Notably, the PsGRI contains multiple copies of NDM-1 and PER-1 carbapenem-resistance and -inhibitor genes, a repetitive structure typically unresolvable by standard short-read methods. Additionally, a large 278,489 bp low-copy circular plasmid harbored single copies of these carbapenemase and extended-spectrum β-lactamase genes alongside other antimicrobial resistance determinants and ISCR1 insertion sequences. Nanopore technology allowed us to precisely identify the duplications, providing critical insights into the strain’s pan-resistant phenotype. This study serves as proof-of-concept for the importance of integrating long-read sequencing into clinical workflows to identify complex resistance mechanisms in DTR pathogens, facilitating targeted antimicrobial stewardship and infection control.

This editorial collection brings together a series of recent studies offering focused insights into clinical challenges in modern dentistry [...]

To solve key challenges in machine vision-driven online defect recognition of soft packaging, such as inadequate ability to capture defect deformation, difficulty in extracting defect features, and limited generalization performance of models, an online detection method for soft packaging defects is proposed by integrating edge computing and domain adaptation. By replacing the backbone network with GhostNet and optimizing feature fusion through an adaptive feature pyramid network (AFPN), the number of model parameters was significantly reduced by approximately 30%. A multi-scale domain adversarial neural network (DANN) was introduced to enable rapid adaptation to target domains by leveraging historical multi-category data. A three-tier edge computing architecture of “terminal–edge–cloud” was built, and the lightweight YOLOv8 model was deployed on edge nodes, significantly reducing detection latency. Experimental results demonstrated that the proposed method achieved an average detection accuracy of 97.5% across five types of soft packaging products, with an inference time of only 10.9 ms and an average system response time of 148 ms. This approach significantly enhances detection speed and accuracy for soft packaging defect recognition, effectively meeting the real-time requirements of industrial inspection.