Search Results (31)

Polar-orbiting microwave radiometers provide indispensable all-weather measurements of the atmospheric state, yet revisit intervals of many hours leave critical gaps during rapidly evolving weather events. To address this limitation, we developed MIDAS (Microwave Inference via Diffusion Across Satellites), a probabilistic framework that estimates microwave brightness temperature (BT) fields across the geostationary full-disk domain from infrared observations at 10 min intervals. This study focuses on the five Microwave Humidity Sounder-2 (MWHS-2) humidity-sounding channels near 183 GHz, which provide vertically resolved water vapor information. MIDAS achieves relative errors below 0.5% for the majority of cases, with a channel-averaged mean absolute error of 1.15 K, outperforming a deterministic U-Net baseline (1.43 K). Beyond per-sample evaluation, MIDAS reproduces large-scale climatological patterns across the full-disk domain over a three-month summer period, consistent with Radiative Transfer for TOVS–Scattering (RTTOV-SCATT) simulations. In deep convective scenes where reconstruction is most difficult, the ensemble spread naturally tracks reconstruction difficulty, providing a built-in indicator of prediction confidence. Notably, MIDAS incorporates real-time polar-orbiting observations as physical constraints via a merge-sampling mechanism, reducing ensemble RMSE by over 20% and improving probabilistic calibration by more than 30%. Proof-of-concept assimilation experiments for two high-impact weather cases show that MIDAS-generated fields yield forecast improvements comparable to those from real satellite observations, reducing tropical cyclone track errors from approximately 110 km to 40 km and improving heavy precipitation forecasts at extreme rainfall thresholds where direct infrared assimilation shows no benefit. Overall, our framework demonstrates the potential of generative models to supplement sparse observational coverage and provide physically plausible microwave humidity fields for downstream applications. Full article

The final stage of urban logistics does not end at the building entrance but continues within complex, vertically structured indoor environments, where conventional ground-based delivery systems face limitations in efficiency, flexibility, and scalability. This study introduces the concept of last-meter delivery, defined as unmanned aerial vehicle (UAV)-enabled transport from the building envelope to the recipient within global navigation satellite system (GNSS)-denied, building-regulated indoor space, and systematically reviews the literature from two traditionally separate domains: indoor-UAV operation in GNSS-denied spaces, and outdoor-UAV-based logistics. Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, 297 studies are synthesized through a two-stream thematic synthesis. The review makes three contributions. First, a unified analytical framework is developed across four dimensions (spatial mobility, logistical capability, social acceptance, and operational coordination) through which the two bodies of literature are shown to be largely complementary, with the gaps in one stream coinciding with the strengths of the other. Second, indoor aerial delivery is found to be subject to a distinct set of operational constraints, including micro-scale navigation accuracy, strict geometric safety envelopes, close human–UAV interaction, and privacy sensitivity, implying that indoor transport-UAVs cannot be realized through simple miniaturization of outdoor platforms but require precision-oriented, human-centric, and building-aware design. Third, the four dimensions are translated into a building-management-oriented indicator framework covering spatial compliance, handover standardization, building information modeling (BIM) integration, occupant consent, and liability allocation, reframing last-meter requirements in terms that are actionable for building planners and facility managers. By framing these challenges within the last-meter perspective, this review identifies the gap between current last-mile theories and emerging in-building aerial logistics and provides a structured foundation for future research. Full article

Addressing the problems of complex forms, low structurization and insufficient reliability of automatic acquisition of professional knowledge sources in the manufacturing industry, this paper proposes an improved multi-person collaborative knowledge acquisition method for professional fields. Drawing on the quality control concept of “three reviews and three proofs” in the publishing industry and combining the characteristics of professional knowledge acquisition tasks, this method constructs a knowledge acquisition process with the collaborative participation of editors and professionals. This paper designs a quality assurance mechanism from three dimensions, namely personnel quality, process quality and result quality; introduces triangular fuzzy numbers to evaluate personnel quality; and establishes a process quality control model under multi-level inspection. Taking the knowledge acquisition project of CNC Machining Manual as an example, 39 professionals completed large-scale professional knowledge processing tasks within 60 working days. Compared with the traditional manual knowledge acquisition method, under similar workload conditions, the proposed method reduces the task completion time by approximately 40% and improves knowledge quality by approximately 10%. The research results show that this method can enhance the organization, inspectability and result stability of the complex professional knowledge acquisition process, and is suitable for constructing vertical domain knowledge bases with high quality requirements. Full article

Manual case review in legal and courtroom workflows faces efficiency bottlenecks. While LLMs offer potential for vertical domains, they often struggle with domain-specific accuracy and hallucinations. This paper introduces ILL, a lightweight model for legal and courtroom assistance trained via QLoRA. By employing 4-bit quantization on an RTX 4060 GPU, ILL achieves precise knowledge transfer with low computational costs. The model attained a BertScore F1 of 0.8037, and a perplexity of 1.89, while largely preserving TruthfulQA performance after fine-tuning and demonstrating competitive results on MMLU tasks. Experiments demonstrate that this method performs excellently on small-scale datasets and shows approximate convergence at scales of about 1200 sentences. This work validates the feasibility and efficiency of constructing high-quality vertical auxiliary models using limited computational resources. Full article

Traditional high-precision canopy segmentation heavily relies on tedious pixel-level manual annotation, while general-purpose zero-shot visual detection algorithms are prone to boundary adhesion and excessive computational load in dense forest areas. To address this, this study proposes a human–machine collaborative, efficient canopy segmentation and canopy cover inversion paradigm, combining the zero-shot pre-annotation capabilities of text-driven object detection with the high-precision segmentation advantages of the lightweight proprietary network LGBU-Net. In the offline annotation stage, this method automatically locates candidate canopy regions using Grounding DINO combined with text prompts and generates initial pixel-level masks using SAM. A high-quality training set is then constructed through minimal manual correction, significantly reducing the cost of traditional fully manual annotation. Subsequently, an improved LGBU-Net designed for complex forest conditions is used for supervised learning. In the feature extraction stage, a lightweight phantom-coordinate attention module (LG-CAM) is introduced to enhance the network’s focus on the geometric center of the tree canopy and suppress semantic interference caused by the forest background, light spots, and shadows. In the decoding stage, a boundary difference fusion module (BDF-Block) is deployed to alleviate the problem of adjacent tree canopy boundaries adhering by utilizing high-frequency gradient information from the underlying layers of UAV imagery. Combined with a boundary-aware hybrid loss function, the clarity of individual tree boundaries is further improved in the gradient domain. Experiments based on UAV imagery of high-density mixed and coniferous forests in Baishan, Jilin Province, show that, with low manual annotation costs, LGBU-Net achieves a canopy segmentation IoU of 90.45% and an individual tree separation F1 score of 89.35%, significantly outperforming general visual algorithms with zero-shot direct inference, and with only 4.85 M model parameters. Furthermore, the segmentation results are used for plot-level canopy vertical cover (CC) inversion, and the estimated values are highly consistent with ground-based measurements. This research provides a high-precision, low-annotation-cost technical solution with good edge deployment potential for large-scale forest resource surveys and forest understory light environment assessment. Full article

The assessment of managerial competencies in information technology (IT) organizations requires robust and validated instruments capable of predicting performance in volatile, uncertain, complex, and ambiguous environments. This study presents a preliminary validation of the MLPD (Machine Learning Predictive Development) model, which integrates 360° multidimensional evaluation, situational awareness, and exploratory analytics. Conceived as a pilot application and proof-of-concept, the research was conducted within the IT organization of a Chilean Defense Institution responsible for the management and administration of information and communication technologies. This study aims to determine how the three most commonly cited managerial competency domains (Transformational Leadership, Situational Awareness, and Collaborative Management) are weighted in additive models of 360° performance evaluation in a defense IT context, and also seeks to determine whether these weightings differ between civilian and military evaluators. Although the study focuses on a specialized case study with a limited sample of 9 IT leaders, the robustness of the preliminary findings is supported by the analysis of 165 rating records from 360° evaluations clustered within 9 leaders. Through this granular data set, multiple linear regression models were developed to examine the predictive relationships among three core competency domains—Transformational Leadership, Situational Awareness, and Collaborative Management—and their impact on overall managerial performance. The results identify Collaborative Management as the strongest predictor of performance, and highlight significant differences between civilian and military evaluators. This finding challenges conventional assumptions about leadership effectiveness in IT contexts and suggests that horizontal coordination capabilities are more critical than vertical authority. These preliminary results validate the model’s internal structure within a highly hierarchical environment, establishing a foundational benchmark for future large-scale applications of the MLPD model in diverse organizational contexts. Full article

Accurate braking-distance prediction for heavy-duty multi-axle trucks remains challenging due to the large gross vehicle weight, tandem-axle interactions, and strong transient load transfer during emergency braking. Recent studies on tire–road friction estimation, commercial-vehicle braking control (EBS/AEBS), and weigh-in-motion (WIM) sensing have highlighted that unmeasured vertical-load dynamics and time-varying friction are key sources of prediction uncertainty. To address these limitations, this study proposes an integrated sensing–simulation–AI framework that combines real-time axle-load estimation, full-scale robotic braking tests, fused road-friction sensing, and physics-consistent machine-learning modeling. A micro-electro-mechanical systems (MEMS)-based load-angle sensor was installed on the leaf-spring panel linking tandem axles, enabling the continuous estimation of dynamic vertical loads via a polynomial calibration model. Full-scale on-road braking tests were conducted at 40–60 km/h under systematically varied payloads (0–15.5 t) using an actuator-based braking robot to eliminate driver variability. A forward-looking optical friction module was synchronized with dynamic axle-load estimates and deceleration signals, and additional scenarios generated in a commercial ASM environment expanded the operational domain across a broader range of friction, grade, and loading conditions. A gradient-boosting regression model trained on the hybrid dataset reproduced measured stopping distances with a mean absolute error (MAE) of 1.58 m and a mean absolute percentage error (MAPE) of 2.46%, with most predictions falling within ±5 m across all test conditions. The results indicate that incorporating real-time dynamic axle-load sensing together with fused friction estimation improves braking-distance prediction compared with static-load assumptions and purely kinematic formulations. The proposed load-aware framework provides a scalable basis for advanced driver-assistance functions, autonomous emergency braking for heavy trucks, and infrastructure-integrated freight safety management. All full-scale braking tests were carried out at approximately 60% of the nominal service-brake pressure, representing non-panic but moderately severe braking conditions, and the proposed model is designed to accurately predict the resulting stopping distance under this prescribed braking regime rather than to minimize the absolute stopping distance itself. Full article

Large-eddy simulation (LES) within a weather research and forecasting (WRF) model coupled with an active scalar transport equation was used to simulate Atmospheric Boundary Layer conditions during the Mosquito fire, the largest wildland fire in California during September 2022. The simulations were conducted with realistic boundary conditions derived from the National Oceanic and Atmospheric Administration (NOAA) High Resolution Rapid Refresh (HRRR) model, with the aim of better understanding the two-way coupling between the ABL and plume dynamics. The terrain was extremely inhomogeneous, and the topography varied significantly within the numerical domain. Initially, LES of the smoke-free ABL was conducted on nested domains, and detailed ABL data were gathered from 8 to 9 September 2022. LES simulations were validated using four Automated Surface Observing System (ASOS) stations and NOAA meteorological (MET) observations, as well as NOAA met Twin Otter measurements, and the desired accuracy was established. The smoke plume was then released into the ABL at noon on 9 September 2022, and the plume simulations were conducted for a period of one hour following the release. During this period, the ABL transitioned from convective to buoyancy-shear-driven regimes. Late-night and early-morning conditions are influenced by the complex topography and low-level jet, whereas buoyancy and shear control the ABL dynamics during the morning and afternoon hours. The plume vertical transport is influenced by the ABL depth and the size of the vertical turbulence structures during that time, whereas the wind conditions and turbulent kinetic energy within the ABL dictate the horizontal transport scales of the plume. In addition, the results demonstrate that the plume modifies the microclimate along its path. Full article

Nowadays, driven by the exponential growth of parameters and training data of AI applications and Large Language Models, a single GPU is no longer sufficient in terms of computing power and storage capacity. Building high-performance multi-GPU systems or a GPU cluster via vertical scaling (scale-up) has thus become an effective approach to break the bottleneck and has further emerged as a key research focus. Given that traditional inter-GPU communication technologies fail to meet the requirement of GPU interconnection in vertical scaling, a variety of high-performance inter-GPU communication protocols tailored for the scale-up domain have been proposed recently. Notably, due to the emerging nature of these demands and technologies, academic research in this field remains scarce, with limited deep participation from the academic community. Inspired by this trend, this article identifies the challenges and requirements of a scale-up network, analyzes the bottlenecks of traditional technologies like PCIe in a scale-up network, and surveys the emerging scale-up targeted technologies, including NVLink, OISA, UALink, SUE, and other X-Links. Then, an in-depth comparison and discussion is conducted, and we express our insights in protocol design and related technologies. We also highlight that existing emerging protocols and technologies still face limitations, with certain technical mechanisms requiring further exploration. Finally, this article presents future research directions and opportunities. As the first review article fully focusing on intra-node GPU interconnection in a scale-up network, this article aims to provide valuable insights and guidance for future research in this emerging field, and we hope to establish a foundation that will inspire and direct subsequent studies. Full article

The construction of large-scale, dynamic datasets for specialized domain models often suffers with problems of low efficiency and poor consistency. This paper proposes a method that integrates multi-role collaboration with automated annotation to address these issues. The framework introduces two new roles, data augmentation specialists and automatic annotation operators, to establish a closed-loop process that includes dynamic classification adjustment, data augmentation, and intelligent annotation. Two supporting tools were developed: an image classification modification tool that automatically adapts to changes in categories and an automatic annotation tool with rotation-angle perception based on the rotation matrix algorithm. Experimental results show that this method increases annotation efficiency by 40% compared to traditional approaches, while achieving 100% annotation consistency after classification modifications. The method’s effectiveness was validated using the WATER-DET dataset, a collection of 1500 annotated images from the water conservancy engineering field. A model trained on this dataset achieved an F1-score of 0.9 for identifying water environment problems in rivers and lakes. This research offers an efficient framework for dynamic dataset construction, and the developed methods and tools are expected to promote the application of artificial intelligence in specialized domains. Full article

To enhance the accuracy and reliability of large language models (LLMs) in regulatory question-answering tasks, this study addresses the complexity and domain-specificity of regulatory texts by designing a retrieval-augmented generation (RAG) testing framework. It proposes a dimensionality reduction-based semantic similarity measurement method and a retrieval optimization approach leveraging information reasoning. Through the construction of the technical route of the intelligent knowledge management system, the semantic understanding capabilities of multiple mainstream embedding models in the text matching of financial regulations are systematically evaluated. The workflow encompasses data processing, knowledge base construction, embedding model selection, vectorization, recall parameter analysis, and retrieval performance benchmarking. Furthermore, the study innovatively introduces a multidimensional scaling (MDS) based semantic similarity measurement method and a question-reasoning processing technique. Compared to traditional cosine similarity (CS) metrics, these methods significantly improved recall accuracy. Experimental results demonstrate that, under the RAG testing framework, the mxbai-embed-large embedding model combined with MDS similarity calculation, Top-k recall, and information reasoning effectively addresses core challenges such as the structuring of regulatory texts and the generalization of domain-specific terminology. This approach provides a reusable technical solution for optimizing semantic matching in vertical-domain RAG systems, particularly for MDSs such as law and finance. Full article

In the field of construction engineering, there exists a dynamic evolution of extensive technical standards and specifications (e.g., GB/T and ISO series) that permeate the entire lifecycle of design, construction, and operation–maintenance. These standards require continuous version iteration to adapt to technological innovations. Engineers require specialized knowledge bases to assist in understanding and updating these standards. The advancement of large language models (LLMs) and Retrieval-Augmented Generation (RAG) technologies provides robust technical support for constructing domain-specific knowledge bases. This study developed and tested a vertical domain knowledge base construction scheme based on RAG architecture and LLMs, comprising three critical components: entropy-optimized dynamic text segmentation (EDTS), vector correlation-based chunk ranking, and iterative optimization of prompt engineering. This study employs an EDTS method to ensure information clarity and predictability within limited chunk lengths, followed by selecting 10 relevant chunks to form prompts for input into LLMs, thereby enabling efficient retrieval of vertical domain knowledge. Experimental validation using Qwen-series LLMs with a test set of 101 expert-verified questions from Chinese construction industry standard demonstrates that the overall test accuracy reaches 76%. The comparative experiments across model scales (1.5B, 3B, 7B, 14B, 32B, and 72B) quantitatively reveal the relationship between model size, answer accuracy, and execution time, providing decision-making guidance for computational resource-accuracy tradeoffs in engineering practice. Full article

Floating breakwaters (FBs) are frequently used to protect marinas, fisheries, or other bodies of water subject to wave attacks of moderate intensity. New forms of FBs are frequently introduced and investigated in the literature as a consequence of technological advancements. In particular, a new possibility is offered by High-Density Polyethylene (HDPE) by extruding pipes of large diameters (e.g., 2.5 m in diameter) and with virtually no limit in length (hundreds of meters). By connecting two or three such pipes in a vertical layout, a novel low-cost floating breakwater with deep draft is devised. This note investigates numerically and experimentally the efficiency of this type of multi-cylindrical FBs in evaluating different geometries and aims at finding design guidelines. Due to the extraordinary length of the breakwater, the investigation is carried out in two dimensions. The 2D numerical model is based on the solution of the rigid body motion in the frequency domain, where the hydrodynamic forces are evaluated (thanks to a linear potential flow model), and the mooring forces do not include dynamic effects nor drag on the lines. The numerical predictions are compared to the results of a 1:10 scale experimental investigation. An atypical shape of the wave transmission ($k_t$) curve is found, with a very low minimum in correspondence with the heave resonance frequency. The results essentially point out the influence of the position of the gravity center, the stiffness, and the mutual distance among cylinders on $k_t$. Full article

Spar is one of the promising floating platforms to support offshore wind turbines. Wind heeling moment is large in the case of floating offshore wind turbines and, therefore, it is important to reduce the pitch motion of the floating platform. To address this issue, a spar platform with damping plates is proposed and investigated in this study. (i) Type-A, (ii) Type-B, and (iii) Type-C models of 1/120 scale were fabricated with similar stability parameters. Type-A is a classic spar, Type-B and Type-C are spar with damping plates by replacing the ballast water part with horizontal plates and vertical plates, respectively. The rotor model consists of (i) no disk and (ii) with disk conditions. A series of model scale experiments were carried out in the water tank in regular waves, and motion response was measured. A calculation method based on classic frequency-domain was developed to incorporate damping plates and validated with the experiment results in no disk and with disk conditions. When pitch response of Type-B and Type-C were compared with respect to Type-A, it was found that the spar platform with damping plates reduced the pitch response in most wave frequencies. Full article

High-impact weather (HIW) events, such as typhoons, usually have sensitive regions where additional observations can be deployed and sensitive observations assimilated, which can improve forecasting accuracy. The ensemble transform sensitivity (ETS) method was employed to estimate the sensitive regions in the “Chaba” case in order to explore the impact of observation data in sensitive areas on typhoon forecasting during the rapid intensification phase. A set of observation system simulation experiments were conducted, with assimilations of sensitive observations (SEN), randomly selected observations (RAN), whole domain observations (ALL), and no assimilation (CTRL). The results show that (1) the sensitive areas of Typhoon “Chaba” are primarily located in the southwest of the typhoon center and are associated with the distribution of the wind field structure; (2) the typhoon intensity and tracks simulated by the SEN and RAN experiments are closer to the truth than the CTRL; (3) the SEN experiment, with only 3.6% of assimilated data observations, is comparable with the ALL experiment during the rapid intensification phase of the typhoon; (4) the uncertainty of the mesoscale model can be improved by capturing large-scale vertical wind shear and vorticity features from the GEFS data and then using the data assimilation method, which makes the vertical shear and vorticity field more reasonable. Full article