Search Results (602)

This work examines the nonlinear dynamics of a generalized Korteweg–de Vries–Zakharov–Kuznetsov equation, a model that appears in plasma physics, shallow water flows, and nonlinear wave propagation. By applying a solitary-wave transformation, the governing partial differential equation is reduced to an autonomous dynamical system, enabling a direct study of its phase portraits and equilibrium behavior. Stability of the fixed points is assessed through Jacobian matrices and eigenvalue classification, revealing parameter regimes that admit saddle states, centers, and oscillatory structures. The system’s richer behavior is explored by varying key parameters, with phase-space trajectories exhibiting periodic, quasiperiodic, and irregular wave patterns. To probe the onset of complexity, we employ several diagnostic tools, including time-series evolution, Lyapunov exponents, bifurcation analysis, sensitivity tests, and Poincaré sections, which together indicate transitions to chaotic motion. The resulting dynamics are further captured using a nonlinear autoregressive neural network, which accurately reproduces the observed trajectories. The combination of analytical and computational perspectives provides a clear framework for understanding this generalized equation and offers a practical approach for investigating other nonlinear systems with a similar structure. Full article

Hepatitis B virus (HBV) infection remains a major public-health challenge in Peru, particularly in historically hyperendemic Amazonian and Andean regions; however, the structure, evolution, and equity of national HBV research have not been systematically evaluated. We conducted a PRISMA-informed bibliometric analysis of all peer-reviewed and theses on HBV in Peru published between 1988 and 2023 using Scopus, Google Scholar, and the Peruvian National Repository (RENATI). Bibliometric indicators, collaboration networks, thematic structure, and temporal thematic evolution were analyzed in R using bibliometrix- and network-based approaches. The final corpus comprised 232 documents, with a marked increase in production after 2005 and a publication peak in 2018. Scientific output was strongly concentrated in Lima-based institutions, while several departments historically associated with HBV endemicity exhibited minimal or absent research production. Nearly half of the corpus corresponded to undergraduate and postgraduate theses. Thematic analyses revealed persistent predominance of epidemiology, seroprevalence, and vaccination-related research, whereas molecular virology, therapeutics, and translational research remained peripheral or poorly represented. International collaboration was markedly limited. Overall, Peruvian HBV research has expanded quantitatively but remains geographically centralized and shows only limited correspondence with the contemporary geographic distribution of HBV incidence, while also remaining only partially aligned with the contemporary global HBV research frontier. These findings provide an evidence-based framework to guide research-priority setting, territorial equity policies, and strategic investment in infectious disease research capacity in Peru. Moreover, the weak association observed between scientific production and departmental HBV incidence suggests that factors beyond contemporary epidemiological burden contribute to the current distribution of research activity in Peru, highlighting a critical but often overlooked dimension of health inequity in low- and middle-income countries (LMIC) research systems. Full article

Postharvest quality deterioration of Korla fragrant pear (Pyrus sinkiangensis Yu) severely constrains its market value, yet the regulatory role of preharvest soil management in shaping postharvest performance remains poorly understood. Although green manure is widely adopted to ameliorate orchard soil degradation, species-specific modulation of postharvest storage trajectories and the quantitative fidelity of soil-to-fruit nutrient transmission have rarely been resolved for climacteric pear species. This study investigated how green manure species modulate fruit quality at harvest and during postharvest storage life and their underlying soil–fruit linkages. Three preharvest treatments were imposed, as follows: control (CK), sweet clover (CM), and alfalfa (MX). Fruits were harvested and stored at 4 °C, with samplings at 1, 5, 10, 15, and 20 d. A critical quality transition was identified at 15 d, characterized by the concurrent peaking of soluble sugars, organic acids, vitamin C, and anthocyanins alongside an optimal sugar–acid ratio. Beyond this inflection point, CM and MX diverged markedly: CM enhanced soluble sugar accumulation, anthocyanin retention, and ester volatile production—most notably hexyl acetate, which increased over 14.4-fold—thereby generating a pronounced fruity aroma bouquet. Conversely, MX sustained higher amino acid and vitamin C levels and conferred superior late-storage stability, evidenced by a three-fold lower coefficient of variation in the sugar–acid ratio relative to CK. Partial-least-squares structural equation modeling (PLS–SEM) revealed soil fertility as the principal exploratory associative factor of fruit quality, but the fidelity of soil-to-fruit transmission was species-dependent. MX exhibited the highest observed associative strength (R² = 0.971), whereas CM exhibited attenuated transmission fidelity (R² = 0.777), with network analysis further indicating that CM exhibited divergent associative patterns of key soil–fruit correlations. These findings suggest that green manure identity is linked to postharvest quality through divergent soil–fruit coupling pathways: alfalfa shows nutrient transmission efficiency and stabilizes nutritional quality, whereas sweet clover promotes sugar-aroma accumulation at the cost of reduced soil–fruit conversion fidelity. Species-specific green manure selection thus offers a viable strategy for targeted modulation of postharvest traits in Korla fragrant pear. Full article

Roman amphorae are among the most frequently excavated archaeological ceramics and provide valuable information on ancient manufacturing techniques and trade networks. This work presents a comparative Mössbauer spectroscopy study of amphora sherds recovered from kiln sites in the Roman provinces of Hispania Lusitania and Baetica in order to reconstruct firing procedures and assess the influence of raw material composition on ceramic appearance and iron speciation. Three representative samples—São Lourenço (Lusitania), Lebrija (Baetica), and Arva (Baetica)—were investigated by ⁵⁷Fe Mössbauer spectroscopy at room temperature and 4.2 K, complemented by X-ray diffraction (XRD), X-ray fluorescence (XRF), and controlled laboratory re-firing experiments under reducing and oxidising atmospheres. The Mössbauer results reveal that all studied amphorae were originally fired under variable redox conditions, involving a reducing stage followed by partial oxidation during the final stages of firing and cooling. Re-firing experiments reproduced the original spectra and demonstrated that the observed iron phase assemblages reflect primary technological processes rather than post-depositional alteration. Low-calcium ceramics developed abundant hematite and characteristic red colours after oxidation, whereas calcium-rich ceramics favoured incorporation of iron into gehlenite and vitreous phases, producing buff or grey colouration. The Arva samples further illustrate how incomplete oxygen penetration generated red oxidised surfaces surrounding partially reduced cores. These results demonstrate that Roman amphora production at the kiln sites studied relied on empirically controlled firing regimes in open updraught kilns and that calcium content was a major factor governing both iron mineralogy and final ceramic appearance. Full article

Background: Diabetic cardiomyopathy (DCM) is characterized by metabolic dysfunction, inflammation, extracellular matrix (ECM) remodeling, and myocardial fibrosis. Increasing evidence suggests that ferroptosis-associated oxidative injury may contribute to cardiac remodeling; however, the interaction between ferroptosis-related pathways and fibrosis-associated molecular networks remains incompletely understood. This study explored the ferroptosis–fibrosis axis using an integrative transcriptomic and systems pharmacology framework. Methods: Differentially expressed genes were identified from the GSE5406 myocardial transcriptomic dataset comparing nonfailing donor hearts with ischemic and idiopathic cardiomyopathy samples and analyzed using functional enrichment, protein–protein interaction, and disease-association approaches. Cross-dataset comparison and exploratory sample-level external evaluation were performed using the independent GSE263297 DCM-related dataset. Candidate genes were further evaluated by receiver operating characteristic (ROC) analysis and machine learning-based feature selection using least absolute shrinkage and selection operator (LASSO), random forest, and support vector machine-recursive feature elimination (SVM-RFE). Representative compounds associated with fibrosis-, oxidative stress-, inflammation-, and ferroptosis-related pathways were subsequently assessed by molecular docking against TGFBR1, STAT3, GPX4, AKT1, SMAD3, and ACSL4. Results: Transcriptomic analyses highlighted ECM organization, collagen-containing ECM, and fibrosis-related pathways as dominant biological themes. Cross-dataset comparison showed partial preservation of transcriptional patterns between independent myocardial cohorts, with 20 of 51 evaluated genes demonstrating concordant expression direction across datasets. ROC analysis identified LUM and ASPN as having the highest area under the curve (AUC) values among candidate genes, whereas COL1A1, COL1A2, and COL3A1 also showed elevated AUC values. Machine learning analyses identified FCN3, HOPX, CNN1, and GLUL as the core signature consistently prioritized across all three algorithms, whereas LUM was additionally identified by two of three algorithms. Internal validation yielded a cross-validated AUC of 0.934 (95% CI: 0.820–1.000), and exploratory sample-level external evaluation of the four-gene signature in GSE263297 yielded an AUC of 0.673 (95% CI: 0.380–0.967). Exploratory docking analyses suggested potential structural compatibility between several candidate compounds and fibrosis-, inflammation-, and ferroptosis-associated targets, with comparatively lower predicted binding-energy values observed for selected ligand–target combinations. Conclusions: The findings are consistent with a fibrosis-dominant remodeling signature and suggest potential network-level links between ferroptosis-associated processes and cardiac fibrosis. These observations should be regarded as exploratory and hypothesis-generating and require validation in independent cohorts and experimental studies. Full article

Background: Late-onset preeclampsia (LO-PE) is the most prevalent clinical phenotype of preeclampsia and, although traditionally considered less strongly associated with placental dysfunction than early-onset disease, increasing evidence supports the presence of relevant placental alterations. The extracellular matrix (ECM) is a key regulator of villous architecture, tissue mechanics, trophoblast behavior, vascular remodeling, and angiogenesis. This study aimed to characterize ECM remodeling in placentas from women with LO-PE. Patients and Methods: A prospective observational study was conducted in 111 pregnant women, including 68 with LO-PE and 43 healthy controls. Placental samples were collected immediately after delivery. Gene expression of elastogenesis-related markers, cross-linking enzymes, fibrillar collagens, matrix-remodeling regulators, and endothelial–matrix signaling molecules was assessed by RT-qPCR. Protein expression was evaluated by immunohistochemistry. Differences between groups were analyzed using non-parametric tests with Benjamini–Hochberg correction, and correlations among ECM markers were explored using Spearman analysis. Results: LO-PE placentas showed significantly increased expression of tropoelastin (TE), fibulin-4 (FBLN-4), fibulin-5 (FBLN-5), fibrillin-1 (FBN-1), lysyl oxidase (LOX), lysyl oxidase-like 1 (LOXL-1), collagen type I (COL-I), collagen type III (COL-III), and matrix metalloproteinase-2 (MMP-2) at both gene and protein levels. Conversely, gene and protein expression of tissue inhibitor of metalloproteinase-2 (TIMP-2) and epidermal growth factor-like domain 7 (EGFL7) showed a marked decrease in the placentas of pregnant women with LO-PE. These findings indicate enhanced elastogenesis, increased matrix cross-linking, greater fibrillar collagen deposition, and an imbalance in matrix turnover. Correlation analysis further suggested that, although the FBLN-4/FBLN-5 axis remained preserved, LO-PE placentas displayed partial disruption of the broader ECM transcriptional network. Conclusions: LO-PE placentas exhibit a coordinated but dysregulated ECM remodeling phenotype involving elastic, collagenous, proteolytic, and endothelial–matrix regulatory pathways. These alterations support ECM remodeling as a relevant biological feature of LO-PE placental pathophysiology. Full article

The rising prevalence of chronic kidney disease represents a major global health burden. Limitations of current renal replacement therapies, including donor organ shortages, rejection, and dialysis-related complications, underscore the need for innovative treatment options. This narrative review assesses the feasibility of bioengineered kidneys as an alternative to current treatments by discussing advances in decellularization, recellularization, and the transplantation of cell-on-scaffold kidneys. We propose that the development of functional bioengineered kidneys follows a hierarchical, staged process, in which vascular patency is the primary prerequisite for graft survival, followed by partial restoration of glomerular filtration, with complete tubular function remaining the final and most challenging milestone. Perfusion-based whole-organ decellularization has made significant progress in preserving the extracellular matrix, enabling the production of acellular human kidney scaffolds. However, complete recellularization of whole kidneys has not yet been achieved. Nevertheless, partially repopulated kidney scaffolds have been shown to withstand physiological blood pressure, produce urine, and exhibit filtration in large-animal models. Complete endothelial coverage of the vascular network proved essential for preventing thrombosis after transplantation. Current work on bioengineered kidneys shows promising results regarding feasibility for clinical application. It is important to note that most of the included studies are proof-of-concept, characterized by small sample sizes and short observation periods. Although these findings are crucial for further research, they cannot be generalized, and larger trials are recommended. In addition to cell-on-scaffold kidneys, 3D bioprinting is a promising technique that could eliminate the need for donor scaffolds. Full article

Large vision–language models have advanced embodied navigation by integrating visual perception with natural-language reasoning. However, vision-and-language navigation (VLN) for unmanned aerial vehicles in low-altitude urban airspaces remains challenging due to occluded views, dynamic layouts, limited communication bandwidth, and partial observability. Existing methods mainly focus on single-agent egocentric navigation and lack explicit modeling of uncertainty and inter-agent dependencies in collaborative multi-UAV settings. We propose Collaborative Low-Altitude Space Navigation (Co-LASN), a dynamic Bayesian network-based framework for collaborative VLN in embodied multi-agent systems. Co-LASN jointly models environmental dynamics, linguistic constraints, and inter-agent dependencies in a unified probabilistic representation, allowing each UAV to update its belief state and incorporate information from neighboring agents when making navigation decisions. Experiments on a low-altitude subset of the HaL-13k benchmark show that, under the evaluated simulation protocol, Co-LASN achieves higher navigation metrics than single-agent and partially collaborative baselines. In the 3-agent setting, Co-LASN increases the any-success rate (ASR) from $12.37\%$ to $15.23\%$ and reduces the min navigation error (MNE) from $99.86$ to $89.46$. These results demonstrate the relative effectiveness of belief-aware collaboration within the evaluated simulation setting. Full article

Aspartame is a widely used artificial sweetener, but its possible relationship with rheumatoid arthritis (RA) remains insufficiently understood. This study aimed to explore, rather than prove, potential molecular links between aspartame-related targets and RA-associated gene networks. Three public RA transcriptomic datasets (GSE55235, GSE55457, and GSE77298) from the Gene Expression Omnibus (GEO) database were integrated as discovery/training data. Because these datasets included different tissue origins, batch correction was used to reduce dataset-level technical variation, whereas tissue-origin-related biological variation was not assumed to be fully removable. After differential expression analysis, RA-associated differentially expressed genes (DEGs) were identified. The single-cell dataset GSE200815 was used for cell annotation and cellular expression visualization; because its comparator group consists of psoriatic arthritis (PsA) samples rather than healthy controls, single-cell results were interpreted as RA-vs-PsA observations and were not treated as disease-versus-healthy-control evidence. Potential targets of aspartame were retrieved from ChEMBL, SwissTargetPrediction, and the Similarity Ensemble Approach (SEA), and were intersected with RA-related DEGs to construct an aspartame-gene-RA regulatory network. Diagnostic models were developed using 113 machine-learning algorithm combinations to determine an optimal multigene model and its core genes. HLA-DRB1 was selected for exploratory scTenifoldKnk-based virtual knockout mainly because it was included in the optimal model and has a well-established role in RA immunogenetics; the single-cell analysis was used only to describe cellular distribution in the RA/PsA dataset. Molecular docking was then used to evaluate the possible interaction between aspartame and HLA-DRB1. Forty-four intersected genes linked the predicted aspartame targets with RA DEGs. The random forest plus partial least-squares generalized linear model (RF + plsRglm) identified 16 core genes. Network-level interpretation indicated that these genes were distributed across immune/antigen-processing, inflammatory-signaling, protease/extracellular-matrix-remodeling, adhesion, metabolic, and proliferation-related modules; therefore, HLA-DRB1 was treated as a prioritized immune-module candidate rather than as the sole driver of the network. Following virtual knockout of HLA-DRB1, affected genes were enriched in extracellular matrix organization, extracellular structure organization, extracellular matrix, collagen trimer, extracellular matrix structural constituent, and collagen binding. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways included integrin signaling, focal adhesion, proteoglycans in cancer, cytoskeleton in muscle, and phosphoinositide 3-kinase/protein kinase B (PI3K/AKT) signaling. Molecular docking showed a minimum binding energy of −6.7 kcal/mol, which was more negative than the preset stability criterion of −5.0 kcal/mol, and the docking pose suggested contacts around ARG-146. This integrative analysis suggests a hypothesis-generating association between aspartame-related predicted targets and RA-relevant molecular networks involving HLA-DRB1 and other core genes. The findings do not establish causality and require experimental, epidemiological, biophysical, and tissue-stratified validation before any causal or clinical inference can be made. Full article

Underwater acoustic communication (UAC) is of great strategic importance for marine resource exploration and security collaboration. However, its open physical nature exposes communication links to severe eavesdropping and localization threats, while limited bandwidth and severe attenuation further exacerbate the difficulty of secure transmission. To address this, this study introduces the underwater acoustic reconfigurable intelligent surface (UARIS) to reconfigure acoustic propagation paths, leveraging its programmable reflection capability to enhance link quality and provide additional spatial degrees of freedom for location privacy protection. Accounting for the partial observability caused by the coarse observations of a mobile eavesdropping user (EU), noisy channel state information (CSI), and the practical constraint of UARIS discrete phase quantization, a utility maximization problem is formulated to jointly optimize the secrecy rate and location privacy. To tackle the strong non-convexity and coupled constraints in dynamic environments, a Gated Recurrent and Conformal-calibrated Soft Actor–Critic (GC-SAC) algorithm is proposed. Specifically, GC-SAC employs a gated recurrent unit (GRU) to capture the temporal statistical features of channel evolution. By integrating a risk prediction network with a conformal calibration mechanism, conservative estimation and robust regulation of multidimensional constraint risks are enhanced. Simulation results demonstrate that the GC-SAC algorithm achieves faster convergence and superior stability in dynamic underwater environments. Compared with representative baselines, the proposed algorithm exhibits significant advantages in secrecy rate and location privacy protection, validating its effectiveness for UARIS-assisted secure resource optimization in underwater scenarios. Full article

Understanding the drivers and barriers of wine consumption is of substantial importance for both market development and sensory science research, and this is particularly salient in rapidly changing non-Western markets. Young, highly educated Chinese consumers represent one of the fastest-growing segments in the global wine market, yet large-scale studies of their consumption preferences and rejection patterns remain limited. This study aimed to characterize the conditional dependence structure of wine-consumption behavior in this population and to examine the associations between common consumption barriers and sociodemographic variables. A nationwide cross-sectional online survey collected 4823 valid responses. Non-parametric tests were used to compare sociodemographic groups, and a regularized Gaussian graphical model (GGM) was estimated to characterize the conditional associations among 15 consumption-behavior variables. The sample was dominated by young respondents (18–24 years) and individuals with higher education. The three most frequently endorsed barriers were taste aversion (51.1%), price sensitivity (38.7%), and lack of knowledge (19.6%). Age and education were the most central sociodemographic variables in the network. The knowledge barrier showed a moderate negative conditional association with education (partial r ≈ −0.171), whereas taste aversion—although the most frequently endorsed barrier—did not show clear conditional associations with sociodemographic variables in the network. Gender was not conditionally associated with any other variable in the network. These observations suggest that the three consumption barriers may operate through different network pathways and may therefore have different implications for intervention design, a possibility that warrants further confirmatory and longitudinal research. Full article

The aim of this study was to evaluate the possibility of using rapeseed protein–fiber concentrate (RPFC) as a functional ingredient for wheat pasta fortification, with emphasis on dough rheology, gel-like network formation, microstructure, and cooking quality. For this purpose, five formulations of rigatoni pasta were produced by partially substituting wheat flour with 0, 5, 10, 15, and 20% RPFC. Dough rheological behavior was assessed by frequency sweep and creep–recovery tests, while mixing and pasting behavior was evaluated using the Mixolab device. Microstructure was analyzed by scanning electron microscopy (SEM), and pasta technological and chemical parameters were determined using standard methods. All dough systems exhibited viscoelastic, gel-like behavior characterized by the dominance of the storage modulus (G’) over the loss modulus (G”), confirming the formation of a structured gluten-based network. Moderate RPFC incorporation (5–15%) enhanced G′, indicating reinforcement of the continuous protein–starch gel matrix and improved structural integrity and deformation resistance. Mixolab results showed a significant increase in water absorption and dough stability with RPFC addition, reflecting improved hydration and strengthening of the gel-forming protein network. SEM observations confirmed the development of a more compact and continuous starch–protein gel system, associated with reduced stickiness and improved structural cohesion. However, higher RPFC levels (15–20%) disrupted the continuity of the gel network, leading to increased cooking losses (8.8–10.4%), higher fracturability, and reduced firmness of cooked pasta. According to the data obtained, RPFC represents a promising functional protein ingredient for gel-like food systems such as cereal-based products, particularly pasta. These findings offer feasible formulation strategies and support its use as a sustainable, high-quality plant protein ingredient in pasta production. Full article

Skeleton-based human action recognition (HAR) requires models that preserve the local kinematic structure of the human body while capturing long-range spatiotemporal dependencies under noisy or incomplete joint observations. Traditional Graph Convolutional Networks (GCNs) provide topology-aligned inductive bias but are often limited by local information aggregation from neighboring joints. In contrast, attention-based mechanisms capture global interactions, yet they may attend to spurious correlations when skeletal constraints are weakly enforced. This paper proposes Differential Hyperedge Attention-enhanced GCN (DHA-eGCN), a hybrid architecture that couples structure-aware Differential Hyperedge Attention with multi-scale temporal convolution for spatiotemporal skeleton sequence processing. DHA injects skeletal structure into attention via hop-distance relative positional encoding and hyperedge context tokens generated via joint-to-part pooling. It further employs differential attention to suppress shared noisy correlations and enhance interaction selectivity. To strengthen spatial grounding, an explicit GCN branch is added under partial- or full-depth configurations, where the first four or all ten layers are applied with graph convolutions. The model further employs an ensemble strategy that combines predictions from multiple complementary model instances. Our experiments on NTU RGB+D 60 under the X-Sub and X-View protocols, NTU RGB+D 120 under the X-Sub and X-Set protocols, and Northwestern-UCLA demonstrate that DHA-eGCN consistently outperforms or remains competitive with strong graph-based, transformer-based, and hybrid state-of-the-art methods based on the same four-stream architecture. The best configuration achieves 93.7% and 97.0% on NTU RGB+D 60 X-Sub and X-View, respectively; 90.9% and 91.9% on NTU RGB+D 120 X-Sub and X-Set, respectively; and 97.6% on Northwestern-UCLA. Full article

Safety-critical Industrial Internet of Things (IIoT) sensor networks deployed in disaster scenarios require intelligent routing mechanisms that prioritize mission-critical packets without relying on centralized coordination. Federated learning on resource-constrained edge nodes presents three primary challenges: the absence of an interpretable supervisory signal, the inability to act conservatively based on per-inference confidence, and vulnerability to partial node availability. The proposed FedCARE framework addresses these issues by employing a Mamdani Fuzzy Inference System to generate traceable criticality labels from multi-modal sensor telemetry, a dropout-aware aggregation protocol that normalizes over only reachable nodes, and a confidence-gated resolver that defers to symbolic fuzzy classification when model confidence is insufficient, otherwise applying an auditable maximization rule to prevent under-prioritization of safety-critical data. Evaluation on 50-, 100-, and 200-node Watts–Strogatz topologies under fault rates up to 50%, using the Edge-IIoTset and WUSTL-IIoT-2021 benchmarks, demonstrates 99.00% critical recall and up to $1.8\times$ higher overall-packet delivery compared to RPL-RP under severe fault conditions. Routing improvements are primarily attributed to fuzzy criticality labeling and multi-path replication. These findings indicate that fuzzy-supervised federated inference offers a practical and interpretable solution for safety-critical IIoT routing, with an observed energy overhead of 7.8% per delivered packet. Full article

In partially observable adversarial environments, heterogeneous unmanned aerial vehicle (UAV) swarms must generate tasks online from noisy observations while respecting platform capabilities, consumable resources, and structural dependencies among tasks. This paper proposes a task graph generation method that converts local observations, target beliefs, and UAV resource states into executable task graphs with explicit resource semantics and inter-task relations. The method first constructs a sufficiently expressive candidate task graph in the belief and resource spaces. An offline search teacher then evaluates future trajectory particles, resource feasibility, and structural interaction values to produce supervision for node selection, marginal task value, and relation prediction. A relation-biased graph attention network learns to generate task graphs online, and a task manager further performs task filtering, dependency repair, conflict completion, and resource checking. Simulation results under complex observation pressure and unseen adversarial strategies show that the proposed method consistently improves structural generation quality and execution feasibility. Compared with Graphormer, it improves the task-graph utility, task-edge F1-score, and executable-graph ratio by 5.83%, 5.41%, and 2.68%, respectively, while reducing the infeasible-task ratio by 35.14%. These results indicate that combining an offline search teacher with resource-constrained graph modeling provides an effective front-end task organization mechanism for heterogeneous UAV swarm planning. Full article