To address the difficulty of accurately identifying driving behavior preferences in highway scenarios, vehicle trajectory data from NGSIM US-101are used in this paper.The data features including speed, acceleration, jerk, time headway, spacing, and lane-changing behavior are extracted. A three-class labeling method for driving behavior preferences, namely cautious, normal, and aggressive, is constructed by combining the density-based spatial clustering of applications with noise (DBSCAN) algorithm with an aggressiveness score. On this basis, a bidirectional long short-term memory (BiLSTM) identification model integrated with an attention mechanism is established to classify driving behavior sequences. The results show that, among 2 490 valid vehicles, the proportions of the three classes are 26.1%, 44.7%, and29.1%, respectively. Significant differences are observed among the classes in terms of speed level, longitudinal control intensity, and car-following safety margin. The proposed model achieves an accuracy of 90.25% in driving behavior preference identification. This method provides a reference for behavior prediction, decision-making and planning, and safety control of intelligent connected vehicles.

To address the problems of low detection accuracy and insufficient robustness caused by variable target scales and complex environments in autonomous driving, an improved depth and geometry aware object detection algorithm based on YOLO26, DGA-YOLO (depth and geometry aware YOLO), is proposed. Firstly, an R-G-Depth early fusion and iterative depth completion strategy is designed to enhance 3D spatial perception. Secondly, a Semantic-Geometric Aggregator (SGA) is introduced into the backbone to extract local semantics and macroscopic geometric context in parallel through two streams. Thirdly, an Anisotropic Geometric Convolution (AGC) is constructed, which parallels depthwise separable and pointwise convolutions to extract anisotropic geometric features, combined with a capacity-bounded scale-aware attention. Finally, a Large Selective Reconstructor (LSR) is deployed in the fusion layer, smoothing noise with large-kernel dilated convolutions and purifying geometric features with cross-strip convolutions. Experiments show that the proposed algorithm improves the mean Average Precision (mAP) on the KITTI dataset, with mAP@50 and mAP@50∶95 exceeding the baseline by 2.5 and 4.4 percentage points, respectively; it also significantly outperforms the baseline under physically simulated rain, fog, and rain-fog extreme conditions and in zero-shot generalization tests, demonstrating excellent anti-interference robustness and generalization capability.

Aiming at the challenges of actuator dynamic limitations, time-varying lung mechanics parameters, and conflicting control requirements between inspiratory pressure build-up and steady-state maintenance in constant-pressure ventilation, a parameter-aware pressure-speed dual closed-loop fuzzy adaptive control method is proposed. An inner speed loopis constructed, in which a Proportional-Integral-Derivative (PID) controller is employed to exploit the fast fan dynamics, while an outer pressure loop is designed with stage-dependent PID gains that are tuned by fuzzy rules according to the respiratory phase. Meanwhile, the feedforward pressure build-up model is corrected using online estimates of airway resistance and lung compliance, thereby reducing reliance on feedback regulation. Experimental results show that, compared with a conventional PID controller, the proposed method reduces rise time by 62.2% and settling time by 67.3%; compared with a state-of-the-art method reported in the literature, the settling time is further reduced by 28.6%, and a steady-state error of ±0.03 cmH₂O is achieved. These results indicate that the proposed method provides improved dynamic response and pressure regulation accuracy for constant-pressure ventilation control.

Landslide susceptibility assessment is an important tool for landslide risk early warning. To address the issue of insufficient capability in extracting and utilizing the physical effects and intrinsic correlations of environmental factors during the assessment process, which makes it difficult to effectively integrate multiple factors for accurate prediction, this study proposes a landslide susceptibility assessment method that combines physics-informed information with deep learning. First, environmental factors are divided into natural conditions and driving factors based on their physical roles in landslides. Then, based on deep learning methods, a natural condition assessment module and a driving factor assessment module are designed to evaluate potential landslide risks and external driving intensity, respectively. Finally, the assessment results are integrated by considering the sequence and causal relationships of the factors to accurately predict landslide probability. The results show that the new method significantly outperforms seven commonly used methods in terms of AUC (0.898 4), overall accuracy (0.829 7), precision (0.795 8), and recall(0.863 6). The self-attention mechanism within the method can accurately identify the key driving factors of landslides, providing precise guidance for disaster prevention and mitigation efforts.

Cancer research collaboration recommendation can be naturally formulated as a link prediction problem on continuous-time dynamic graphs. However, existing temporal graph models are primarily designed for general scenarios and struggle to incorporate domain-specific semantic information. To address this limitation, an adaptive temporal graph network model, termed AdaTGN-CR, is proposed for research collaboration recommendation. Specifically, an expert-driven adaptive neighbor sampling mechanism is introduced to dynamically adjust the neighborhood size and neighbor importance weights according to expert status. In addition, an expert feature modeling method is developed to quantify academic activity, collaboration intensity, and expertise level. Furthermore, a layer-normalized gated recurrent unit (GRU) is employed to enhance the stability of temporal memory updates and capture long-term collaboration patterns. Experimental results on the cancer research dataset CATCD demonstrate that the proposed model achieves average precision (AP) of 93.99% and93.81% under transductive and inductive settings, respectively, consistently outperforming state-of-the-art baseline models. These results indicate that integrating domain semantic information with adaptive temporal graph modeling significantly enhances the performance of research collaboration recommendation.

The message queuing telemetry transport (MQTT) protocol is widely used for data transmission in earthquake early warning, rapid reporting, and disaster assessment due to its high concurrency and low latency. However, the protocol suffers from security vulnerabilities such as simplistic authentication, plaintext transmission, and susceptibility to message tampering. Moreover, existing security solutions struggle to simultaneously meet the requirements of lightweight design, compatibility with domestic innovation ecosystems, and multi-terminal security protection, which restricts its application in high-security earthquake-related operations. To address these issues, this study proposes a domestic cryptographic-enhanced MQTT secure transmission scheme tailored for domestic innovation ecosystems. Based on the enhanced authentication mechanism of MQTT 5.0, the scheme designs a domestic cryptographic-enhanced authentication algorithm. Utilizing SM2 encryption and digital signature technology, it enables bidirectional trusted authentication between the client and the broker in just four rounds of interaction, achieving a balance between security and lightweight performance. Additionally, a key re-encryption mechanism is introduced to establish a "one-key-per-subscriber" protection model, effectively mitigating the risk of key leakage in multi-subscriber scenarios caused by the "single-topic-single-key" approach. Security analysis confirms that the scheme can resist typical attacks such as identity forgery, man-in-the-middle interception, and tampering. Simulation results show that in 25 repeated tests, the average time for identity authentication is 13.36 ms, the average time for topic key acquisitionis 0.977 ms, and the average data transmission time in scenarios with 300 subscribers is 551.78 ms. Constructed with domestic cryptographic algorithms, this scheme is compatible with domestic innovation ecosystems and low-computing-power devices such as earthquake early warning terminals, fully meeting the core requirements of high security and low latency for earthquake-related data transmission.

To address the defects of the Chinese pangolin optimization algorithm(CPO), such as uneven initial population distribution, a tendency to fall into local optima in high dimensions, and convergence oscillation in the later stage, an improved Chinese pangolin optimization algorithm(ICPO) based on multi-strategy enhancement is proposed. The Logistic chaotic map is integrated to optimize the initial population distribution and improve the coverage of the solution space. A dynamically triggered Cauchy mutation mechanism is introduced to enhance the ability of the algorithm to escape from local optima. A full-process elite strategy is designed, including elite pool construction, guided update, and protection mechanisms, to strengthen the guidance of high-quality solutions and improve convergence stability. In the simulation experiments, 12 test functions are adopted to compare the performance of ICPO with that of CPO and other seven swarm intelligence optimization algorithms. Friedman and Wilcoxon rank-sum tests confirm that its performance advantages are statistically significant. The proposed algorithm is further applied to three types of engineering optimization problems. The results show that the ICPO algorithm has better search accuracy and stability, which verifies its adaptability and favorable performance.

The dynamical behavior of a predator system with Smith growth and cross-diffusion is investigated. Firstly, the direction of Hopf bifurcation and the stability of periodic solutions for the ODE are analyzed using normal form theory and center manifold theorem. Secondly, an analytical criterion for cross-diffusion inducing the Turing instability of the periodic solutions is derived through Floquet theory and eigenvalue perturbation method. The results show that when cross-diffusion coefficients exceed a critical value, the originally stable spatially homogeneous periodic solutions lose stability under spatial perturbations, leading to spatial pattern formation. Finally, numerical simulations are performed using Matlab to verify the theoretical results.

An improved model named FF-ResPINN, combining Fourier feature mapping with residual connections, is proposed to address the challenges of spectral bias and gradient pathology encountered by Physics-Informed Neural Networks (PINN) when solving high-order nonlinear partial differential equations. Firstly, input coordinates are projected into a high-dimensional feature space by Fourier feature mapping. The spectral bandwidth of the Neural Tangent Kernel is regulated, and the capability of the network to capture high-frequency oscillatory solutionsis enhanced. Secondly, residual connections are introduced into the network to establish direct propagation paths for gradients. Consequently, the vanishing gradient problem triggered by high-order derivative calculations is effectively mitigated. The fourth-order Boussinesq equation and the fifth-order Sawada-Kotera equationare selected for numerical verification. Experimental results show that, compared with the standard PINN model, the solution accuracy and convergence stability are obviously improved by FF-ResPINN. The relative L₂ error reaches O(10^-5). The effectiveness of the proposed method in solving high-order nonlinear partial differential equations is verified.

This study investigated the secondary metabolites of the endophyticfungus Talaromyces sp. HSHL-F40 isolated from Ardisia mamillata Hance and their antibacterial and antifungal activities. Seventeen compounds were isolated from the rice fermentation of the fungus by using column chromatography and semi-preparative HPLC. Their structures were identified as(2E,5E)-4-oxo-2-propylidene-5,7-octadienoic acid(1), peniciisocoumarin H (2), penicimarin C (3), penicilloxalone B (4), (R)-3-((3S, 4S)-3, 4-dihydroxypentyl)-8-hydroxyisochroman-1-one (5), (R)-3-((R)-4, 5-dihydroxypentyl)-8-hydroxyisochroman-1-one (6), talaroisocoumarin A (7), peniciisocoumarin F (8), aspergillumarins B (9), aspergillumarin A (10), (3S, 4S)-sclerotinin A (11), 3-(hydroxymethyl)-6, 8-dimethoxy-2H-1-benzopyran-2-one (12), talarohydrazone B(13), sequoiamonascin C (14), vermistatin (15), 4-hydroxy-2-methoxyacetanilide (16), and (E)-3-(2, 5-dioxo-3-(propan-2-ylidene)pyrrolidin-1-yl)acrylic acid (17), by using NMR and HR-ESI-MS. Among these, compound 1 is a new compound, and the other compounds were isolated from the endophytic fungi of Ardisia mamillata Hance for the first time. Compound13 exhibited antibacterial activity against Staphylococcus aureus, Bacillus subtilis and Bacillus megaterium, with MIC values of 8 mg/L. Antifungal assays revealed that compound 1 showed moderate inhibitory activity against Colletotrichum fructicola, with an MIC value of 32 mg/L.

A phytochemical investigation of the ethanolic extract of Pteris semipinnata L. resulted in the isolation of eight terpenoids. The isolation was performed using a combination of chromatographic techniques, including silica gel, Sephadex LH-20, and semi-preparative HPLC. Structural elucidation was accomplished through comprehensive spectroscopic analysis, encompassing MS, 1D and 2D NMR, polarimetry, and electronic circular dichroism (ECD). The compounds were identified as (2R,3R)-14-norpterosin L(1), norpterosin C(2), norpterosin B(3), pterosin B(4), spelosin 3-O-β-D-glucopyranoside (5), 9-hydroxy-15-oxo-ent-kaur-16-en-18-oic acid (6), ent-(4α,11α)-11-hydroxy-15-oxo-kaur-16-en-18-oic acid (7), and ent-(7α,9α)-dihydroxy-kaur-16-en-19-oic acid (8). Compound1 was determined to be a new pterosin sesquiterpenoid, and compounds 5 and8 represent the first report of these constituents from this species. The isolated constituents were evaluated forin vitro antiproliferative activity with the MTS colorimetric assay against a five-member human tumour cell-line panel, and compounds 6 and7 possessed varying degrees of activity.

Sanye Qingteng is a characteristic Zhuang medicine in Guangxi. In folk medicine, Sanye Qingteng is often used to make wine to treat joint swelling and pain. To explore the effect and mechanism of the ethanol extract of Sanye Qingteng on joint damage in adjuvant-induced arthritis (AIA) rats, this study employed network pharmacology, molecular docking and other methods in combination with animal experiments. The results showed that the main components of the ethanol extract of Sanye Qingteng were enriched in pathways such as PI3K/Akt, and the main components, including glycocholic acid and emodin, had a strong binding ability with PI3K/Akt. After intervention with the ethanol extract of Sanye Qingteng, the joint swelling and bone damage in AIA rats were alleviated, the number of osteoclasts decreased, the expression of PI3K/Akt pathway proteins was down-regulated, the expression of bone damage-related proteins and genes such as MMP-9 was reduced, and the expression of type Ⅱ collagen was increased. The study indicated that the ethanol extract of Sanye Qingteng could alleviate joint damage in AIA rats, and the mechanism might be related to the inhibition of the PI3K/Akt pathway, the reduction of osteoclast numbers, and the down-regulation of bone damage-related proteins and genes.

This studyinvestigated the analgesic mechanism of the ethanolic extract of Dalbergia hancei stem by using ultra-high performanceliquid chromatography-quadrupole time-of-flight tandemmass spectrometry (UPLC-Q-TOF-MS/MS) combined withnetwork pharmacology. The chemical constituents of the ethanolic extract, as well as those in the medicated serum, feces, and urine, were analyzed using UPLC-Q-TOF-MS/MS. Active components and analgesic targets were screened through relevant databases. Subsequently, a “component-target” network and a protein-protein interaction (PPI) network were constructed, followed bygene ontology (GO) and Kyoto encyclopedia of genes and genomes (KEGG) pathway enrichment analyses. A total of seventy-seven compounds were identified from the ethanolic extract, with thirty-six of them screened as active components. Among these active components, fifteen were confirmed as prototype constituents in vivo. Network pharmacology analysis yielded 139 intersecting targets, from which PTGS2, TIMP1, SERPINE1, and PPARG were identified as the core targets. Pathway enrichment analysis revealed that the analgesic effect was associated with five key pathways, primarily involving the calcium signaling pathway and the serotonergic synapse. The results suggest that Dalbergia hancei stem likely exerts its analgesic effect through a multi-component, multi-target, and multi-pathway mechanism. Active components such as geranylacetone and butein act on multiple targets including PTGS2, and regulate signaling networks involving pathways such as the calcium signaling pathway.

This study aimed to investigate the active components and molecular mechanism of Tiaoqi Rougan Sanjie Pill (TRSP) in inhibiting angiogenesis, migration and invasion of hepatocellular carcinoma in vitro. A total of 42 chemical components were identified by UPLC-QTRAP-MS/MS; furthermore, based on network pharmacology, 128 potential targets related to HCC metastasis were screened, and four core targets (HIF-1α, TNF, CASP3, and SRC) were identified by topological analysis, with HIF-1α occupying a central position in the network and being identified as the core target. Molecular docking showed that multiple components in TRSP exhibited good binding activity with HIF-1α, with binding energies all below -20.9 kJ/mol. In vitro experiments confirmed that TRSP dose-dependently inhibited the proliferation, migration and invasion of hepatoma cells (HepG2 and MHCC-97H), and significantly suppressed the angiogenic ability of human umbilical vein endothelial cells (HUVECs). Mechanistic studies revealed that TRSP downregulated the protein expression of HIF-1α and its downstream factor VEGFA in hepatoma cells. In conclusion, TRSP may exert anti-angiogenic and anti-metastatic effects against HCC by inhibiting the HIF-1α/VEGF signaling pathway, providing experimental evidence for its clinical application.

To explore the visual preference characteristics of landscapes integrating natural and artificial elements in urban parks, and to verify the applicability of eye tracking technology and the segment anything model (SAM)in landscape quantitative analysis, this study selected Guilin Shanshui Park as the research object. A multi-dimensional evaluation system combining landscape features, visual behavior and subjective assessment was constructed byusing eye-tracking technology, the scenic beauty estimation (SBE) method and the SAM. Forty participants were recruited, and their eye movement data and aesthetic scores were collected. Automatic segmentation and quantitative analysis of landscape elements were carried out with the SAM. Differences in landscape visual behavior among different functional zones, landscape elements and groups in Shanshui Park, as well as their correlation with aesthetic evaluations were systematically compared. The results showed that: 1) Two main high-value cores in the park were revealed by Kernel density analysis, which were located in the south-central part of the Cultural and Sports Entertainment Zone and the northeastern part of the Sightseeing Zone. A highly significant positive correlation was observed between average fixation duration and SBE scores. The highest fixation count (25.34 times) was found in the Sightseeing Zone, while the lowest (22.72 times) was recorded in the Leisure Strolling Zone. 2) Significant differences in visual preferences for landscape elements were detected among different participant groups. The long-term resident group exhibited significantly longer fixation durations in artificial areas such as the Urban Theme Square, while the short-term stay group showed longer fixations on characteristic elements such as landscapesketches. 3) Plants were identified as the dominant landscape element in the park, accounting for 39.0%. A significant correlation between the landscape element segmentation results of the SAM and eye tracking attention indicators was confirmed, which indicates that the quantitative indexes of the model can effectively reflect human visual attention distribution and provide a new approach for objective and efficient landscape feature analysis. This study provided empirical evidence for the landscape optimization of Guilin′s landscape city parks and expanded the application paths of eye-tracking technology and the SAM in the research on landscapes of specific region.

This study establishes a scientific assessment framework, diagnostic methodology, and management pathways for the ecological conservation and sustainable development of karst tourism destinations, offering valuable insights for coordinating tourism development and ecological preservation in environmentally fragile regions. Taking the Yulong River Karst Wetland Scenic Area in Guilin, Guangxi Zhuang Autonomous Region, China as a case study, this paper systematically evaluates the spatio-temporal evolution of three key ecosystem services-water conservation, habitat quality, and carbon storage-from 2003 to 2023 using the InVEST model. The findings indicate that, driven by tourism-induced land-use changes, the scenic area’s ecosystem services have generally degraded. This is reflected in the continuous decline of water conservation capacity, polarized habitat quality, and a stepwise reduction in carbon storage, collectively forming a spatial differentiation pattern characterized by "high-value preservation in the upper reaches, median-value expansion in the middle reaches, and low-value aggregation in the lower reaches". The study further reveals that tourism activities drive land use change through three primary pathways: facility expansion, agricultural transformation, and operational emissions, which subsequently trigger coordinated degradation of ecosystem services. Based on the evaluation results and spatial coupling patterns, this paper proposes a three-tier spatial management system-comprising core conservation areas, ecological restoration zones, and intensive utilization areas-along with ecologically tailored optimization measures such as terraced wetlands, permeable pavements, and native vegetation restoration. These recommendations provide a scientific basis and practical reference for advancing ecological protection and sustainable development in karst tourist destinations.

This study developed a microsatellite marker system suitable for the Black-necked Crane (Grus nigricollis) and the Common Crane (Grus grus), aiming to provide a reliable tool for their conservation genetics research. Based on 1 408 feather samples and two muscle DNA samples collected from the Caohai National Nature Reserve in Guizhou, microsatellite markers were employed for individual identification and genetic relationship analysis of the two crane species. Through locus screening and optimization, individual identification panels comprising 13 loci (for Black-necked Crane) and 16 loci (for Common Crane) were constructed, respectively. The systems demonstrated extremely high discriminatory power: the probability of identity for unrelated individuals (PID) and the probability of identity among siblings (PIDsib) for the Black-necked Crane were 5.797×10^-11 and 4.886×10^-5, respectively, and 1.308×10^-18 and 2.700×10^-7 for the Common Crane. A total of 66 individual Black-necked Cranes and 68 individual Common Cranes were successfully identified. Furthermore, this study evaluated the impact of locus number on the resolution of genetic relationship analysis. The results showed that the 16-locus panel for the Common Crane could consistently distinguish different genetic relationship categories, while increasing the loci to 16 enabled more accurate discrimination of various genetic relationship grades for the Black-necked Crane. The microsatellite system established in this study provides crucial technical support for subsequent analyses of the social structure, population dynamics, and scientific conservation management of these two crane species.

Tetramethylpyrazine is a pharmaceutically active substance and a flavor substance in Baijiu. A strain (Bacillus sp. TTMP20) with a high tetramethylpyrazine yield (3 896.36 mg/L) was screened from Daqu and fermented grains of Danquan Baijiu in this study. To better understand the mechanism of tetramethylpyrazine production by TTMP20 and perform its genetic modification, genome sequencing of TTMP20 was conducted. This study described the genomic characteristics of TTMP20. It analyzed genes and pathways associated with tetramethylpyrazine synthesis in detail. Gene function was analyzed by KEGG, COG, GO, NR and CAZy annotation. These pathways included fructose metabolism pathway, pathways of starch and sucrose metabolism, glycolysis pathway, pentose phosphate pathway, partial butanoate metabolism pathway and nitrogen metabolism pathway. Results showed that Bacillus sp. TTMP20 had complete metabolic pathways for acetoin and tetramethylpyrazine, and TTMP20 could utilize multiple complex carbon sources. Genome information thoroughly revealed the connection between genes and tetramethylpyrazine production.