Processing Capabilities Research Articles

Electrode-Dependent and Tunable Sub-to-Super-Linear Responsivity in Mott Material-Enabled Near-Infrared Photodetectors for Advanced Near-Sensor Image Processing.

Brain-like intelligence is ushering humanity into an era of the Internet of Perceptions (IoP), where the vast amounts of data generated by numerous sensing nodes pose significant challenges to transmission bandwidth and computing hardware. A recently proposed near-sensor computing architecture offers an effective solution to reduce data processing delays and energy consumption. However, a pressing need remains for innovative hardware with multifunctional near-sensor image processing capabilities. In this work, Mott material (vanadium dioxide)-based photothermoelectric near-infrared photodetectors are developed that exhibit electrode-dependent and tunable super-linear photoresponse (exponent α > 33) with ultralow modulation bias. These devices demonstrate an opto-thermo-electro-coupled phase transition, resulting in a large photocurrent on/off ratio (>105), high responsivity (≈500 A W-1), and well detectivity (≈3.9× 1012 Jones), all while maintaining rapid response speeds (τr = 2 µs and τd = 5 µs) under the bias of 1V. This electrode-dependent super-linear response is found to arise from the electron doping effect determined by the polarity of the Seebeck coefficient. Furthermore, the work showcases intensity-selective near-sensor processing and night vision pattern reorganization, even with noisy inputs. This work paves the way for developing near-sensor devices with potential applications in medical image preprocessing, flexible electronics, and intelligent edge sensing.

Molecular Reconfiguration of Disordered Tellurium Oxide Transistors with Biomimetic Spectral Selectivity.

Reconfigurable devices with field-effect transistor features and neuromorphic behaviors are promising for enhancing data processing capability and reducing power consumption in next-generation semiconductor platforms. However, commonly used 2D materials for reconfigurable devices require additional modulation terminals and suffer from complex and stringent operating rules to obtain specific functionalities. Here, a p-type disordered tellurium oxide is introduced that realizes dual-mode reconfigurability as a logic transistor and a neuromorphic device. Due to the disordered film surface, the enhanced adsorption of oxygen molecules and laser-induced desorption concurrently regulate the carrier concentration in the channel. The device exhibits high-performance p-type characteristics with a field-effect hole mobility of 10.02 cm2 V-1 s-1 and an Ion/Ioff ratio exceeding 106 in the transistor mode. As a neuromorphic device, the vision system exhibits biomimetic bee vision, explicitly responding to the blue-to-ultraviolet light. Finally, in-sensor denoising and invisible image recognition in static and dynamic scenarios are achieved.

Comparative Analysis of the Response Accuracies of Large Language Models in the Korean National Dental Hygienist Examination Across Korean and English Questions.

Large language models such as Gemini, GPT-3.5, and GPT-4 have demonstrated significant potential in the medical field. Their performance in medical licensing examinations globally has highlighted their capabilities in understanding and processing specialized medical knowledge. This study aimed to evaluate and compare the performance of Gemini, GPT-3.5, and GPT-4 in the Korean National Dental Hygienist Examination. The accuracy of answering the examination questions in both Korean and English was assessed. This study used a dataset comprising questions from the Korean National Dental Hygienist Examination over 5 years (2019-2023). A two-way analysis of variance (ANOVA) test was employed to investigate the impacts of model type and language on the accuracy of the responses. Questions were input into each model under standardized conditions, and responses were classified as correct or incorrect based on predefined criteria. GPT-4 consistently outperformed the other models, achieving the highest accuracy rates across both language versions annually. In particular, it showed superior performance in English, suggesting advancements in its training algorithms for language processing. However, all models demonstrated variable accuracies in subjects with localized characteristics, such as health and medical law. These findings indicate that GPT-4 holds significant promise for application in medical education and standardized testing, especially in English. However, the variability in performance across different subjects and languages underscores the need for ongoing improvements and the inclusion of more diverse and localized training datasets to enhance the models' effectiveness in multilingual and multicultural contexts.

Scale matters: Large language models with billions (rather than millions) of parameters better match neural representations of natural language.

Recent research has used large language models (LLMs) to study the neural basis of naturalistic language processing in the human brain. LLMs have rapidly grown in complexity, leading to improved language processing capabilities. However, neuroscience researchers haven't kept up with the quick progress in LLM development. Here, we utilized several families of transformer-based LLMs to investigate the relationship between model size and their ability to capture linguistic information in the human brain. Crucially, a subset of LLMs were trained on a fixed training set, enabling us to dissociate model size from architecture and training set size. We used electrocorticography (ECoG) to measure neural activity in epilepsy patients while they listened to a 30-minute naturalistic audio story. We fit electrode-wise encoding models using contextual embeddings extracted from each hidden layer of the LLMs to predict word-level neural signals. In line with prior work, we found that larger LLMs better capture the structure of natural language and better predict neural activity. We also found a log-linear relationship where the encoding performance peaks in relatively earlier layers as model size increases. We also observed variations in the best-performing layer across different brain regions, corresponding to an organized language processing hierarchy.

The Role of Information Processing and Digital Supply Chain Technology in Supply Chain Resilience through Supply Chain Risk (Management in Manufacturing Companies in DKI Jakarta)

The 2019 COVID-19 pandemic highlighted the importance of supply chain resilience in the face of global disruptions. Supply chain disruptions caused by isolation measures during the pandemic triggered market turmoil, logistical uncertainty, and reduced production capacity, which sometimes led to shutdowns. In this context, research on supply chain resilience is crucial to prepare companies for future disruptions. This research aims to examine the impact of supply chain resilience on supply chain continuity in manufacturing companies in DKI Jakarta, focusing on the role of digital supply chains, information processing capabilities, and supply chain risk management. This research method uses a quantitative approach. Data were collected from 300 respondents in the basic and chemical industry, miscellaneous industries, and consumer goods sectors through purposive sampling techniques. The analysis was conducted using Structural Equation Modeling (SEM) to identify the relationship between variables. The results showed that supply chain resilience has a significant positive influence on supply chain continuity. In addition, digital supply chains, information processing capabilities, and effective supply chain risk management were shown to support supply chain resilience in the studied sectors. The implication of this research is that it is important for companies to invest in digital technology, develop better information processing capabilities, and implement effective risk management to strengthen their supply chain resilience and continuity in the face of future global uncertainty.

How Can Circular Economy Strengthen the Recycling of Plastic Packaging Wastes? Exploratory and Qualitative Case Study of the Danish Plastic Industry

This paper investigates how companies in the plastic value chain can comply with the requirements put forward in the European Union’s Extended Producer Responsibility (EPR) scheme for plastic packaging wastes, which must be fully adopted in a Danish context ultimo 2024. Plastic packaging waste for recycling is expected to increase when the EPR scheme enters into force, just as national regulation seeks to avoid the incineration of plastics. Multiple case studies, applying an exploratory and qualitative approach, were conducted with three best-case companies within the plastic value chain to illuminate technical capabilities in sorting and further processing of plastics, current recycles strategies and market positioning. To access the possibility to prevent, reuse, recycle, and recover plastic wastes we make use of the waste hierarchy in combination with a circular economy approach, where suggestions to activities along the cascading chain are proposed to qualify the analysis. This paper reveals that only few recycling plastic value chains have been established that originate from plastic packaging waste from household, and there are none for food packaging reusage; the current unfavorable management practices mostly lead to downcycling of the plastic value chain to recovery only. It is further concluded that strengthening the EPR scheme is important, by applying, for example, eco-design, hereunder to utilize mono plastics of polyethylene (PE), polypropylene (PP) and polyethylene terephthalate (PET), and by fee-modulation using taxes/penalties to spur on the plastic industry to produce environmentally friendly products. Finally, alternative sources of handling and producing plastics are outlined as watermark technology, bio-plastics and PtX-nafta.

Job Scheduling in Hybrid Clouds With Privacy Constraints: A Deep Reinforcement Learning Approach

ABSTRACTWith the proliferation of cloud computing and the escalating demand for extensive data processing capabilities, an increasing number of enterprises are embracing hybrid cloud solutions. However, as more businesses move toward hybrid clouds, the need for effective solutions to privacy and security concerns becomes increasingly important. Although current scheduling approaches for cloud computing have addressed privacy protection to some extent, few have adequately considered the unique challenges posed by hybrid clouds. To address this gap, we propose a novel approach for scheduling jobs in hybrid clouds that prioritizes privacy protection. Our approach, called PH‐DRL, leverages Deep Reinforcement Learning (DRL) to intelligently allocate jobs to virtual machines, optimizing both privacy and Quality of Service (QoS), while minimizing response time. We present the detailed implementation of our approach and our experimental results demonstrate the superior performance of PH‐DRL in terms of privacy protection compared to existing methods.

Leveraging Artificial Intelligence to Automate and Enhance Security Operations: Balancing Efficiency and Human Oversight

It is now clear that as new cyberthreats emerge, old security measures are less successful in thwarting increasingly sophisticated cyberattacks. One of the most significant and influential technologies of the last few years, artificial intelligence (AI) is a game-changer for security operations because it can automate critical processes, maintain real-time threat recognition, and respond with equal speed and efficiency. Based on the automation of threats, vulnerabilities, and events, this abstract analyzes the use of AI in security. It looks at specific AI use cases, such as automated patch management, UEBA, and machine learning-based anomalous activity detection, and explains how the technologies greatly increase the speed and accuracy of thwarting cyberthreats. As a tool, AI offers massive advantages: it helps minimize human effort, act faster, and scale easily but is not without risks inherent to operating on the internet. Namely, some responsibilities involve inputting relevant context, making ethical decisions, and interpreting what AI systems have to offer to human operators will always be needed.This abstract also explores the further discussion of the integration of AI security processing capabilities and experienced security personnel so that both facets are achieved efficiently and rightfully ethically. Finally, the abstract concludes that the key to the future advancement of cybersecurity is the ability to find the best conditions for the collaboration of artificial intelligence and human thought, to strengthen the security of organizations and reduce the consequences of plan implementation.

Fault tolerant & priority basis task offloading and scheduling model for IoT logistics

The Internet of Things (IoT) has connected millions of devices to the internet for communication and computation purposes. Due to constraints such as limited battery power, processing capabilities, and storage capacity, IoT devices often rely on remote fog nodes for task execution, a practice known as task offloading. The NP-hard nature of offloading and scheduling IoT tasks onto fog nodes poses significant challenges. Current IoT task offloading often relies on single parameters like device MIPS, RAM, or battery power. Similarly, task allocation on fog nodes frequently prioritizes VM MIPS. To address energy consumption and task failures in IoT systems, robust fault tolerance, efficient task offloading, and effective scheduling are crucial. This paper introduces a novel Fault-Tolerant, Priority-Based Task Offloading and Scheduling Model (FP-TOSM) for IoT logistics. Utilizing the Analytic Hierarchy Process (AHP) for multi-criteria decision-making, IoT tasks are prioritized. Tasks below a specified threshold are executed locally, while those within a defined range are offloaded to a fog node. Tasks exceeding this range are delegated to the cloud. The Euclidean formula is used to determine the proximity between the IoT logistics vehicle and fog nodes for offloading decisions. Dynamic re-clustering fault tolerance mechanisms are utilized to manage task failures after offloading. Successful offloaded tasks are allocated to suitable VMs in the cloud and fog nodes. The proposed strategy selects the most suitable VM for offloaded tasks using a multi-criteria decision-making process to reduce energy consumption, SLA violations, and execution costs. The model's performance is evaluated through simulations in iFogSim2. Results demonstrate reductions in energy consumption by up to 5.8 % and 10.4 %, decreases in SLA violations by up to 25.2 %, and a 16.28 % improvement in response time compared to an ACO-based model with a response time of 17.8 %. Additionally, the task failure ratio shows a significant reduction of up to 22.1 %. These findings highlight the effectiveness of FP-TOSM in enhancing efficiency and reliability within fog computing environments.

Inspiration from Visual Ecology for Advancing Multifunctional Robotic Vision Systems: Bio-inspired Electronic Eyes and Neuromorphic Image Sensors.

In robotics, particularly for autonomous navigation and human-robot collaboration, the significance of unconventional imaging techniques and efficient data processing capabilities is paramount. The unstructured environments encountered by robots, coupled with complex missions assigned to them, present numerous challenges necessitating diverse visual functionalities, and consequently, the development of multifunctional robotic vision systems has become indispensable. Meanwhile, rich diversity inherent in animal vision systems, honed over evolutionary epochs to meet their survival demands across varied habitats, serves as a profound source of inspirations. Here, recent advancements in multifunctional robotic vision systems drawing inspiration from natural ocular structures and their visual perception mechanisms are delineated. First, unique imaging functionalities of natural eyes across terrestrial, aerial, and aquatic habitats and visual signal processing mechanism of humans are explored. Then, designs and functionalities of bio-inspired electronic eyes are explored, engineered to mimic key components and underlying optical principles of natural eyes. Furthermore, neuromorphic image sensors are discussed, emulating functional properties of synapses, neurons, and retinas and thereby enhancing accuracy and efficiency of robotic vision tasks. Next, integration examples of electronic eyes with mobile robotic/biological systems are introduced. Finally, a forward-looking outlook on the development of bio-inspired electronic eyes and neuromorphic image sensors is provided.

MONITORING THE SAUSAGE PRODUCT USING LANEY DEMERIT CHART BASED ANALYTICAL HIERARCHY PROCESS

Ready-to-eat sausage is a food product that has a limited shelf life. Therefore, regularly monitoring the quality of packaged ready-to-eat sausage products is important to ensure that the products meet the established quality standards. Twelve types of product defects need to be observed in the final checking process to meet the quality standards, namely Wrinkle, Dots, Leaking, Product Stain, Non-standard Form, Poor Print Quality, Vacuum Leaks, Weak Ties, Body Defects, Uneven Length, Broken Node, and Small Stain. This study aims to apply the Laney Demerit Control Chart (LDCC) and Analytical Hierarchy Process-Integrated Statistical Process Control (AHP-ISPC) methods to monitor the quality of packaged ready-to-eat sausage production at XYZ Inc. The data is from the quality testing of ready-to-eat sausage products taken from XYZ Inc. for six months from April 1, 2023, until September 30, 2023. The findings reveal that conventional control charts (u control chart, demerit control chart, and AHP-based demerit control chart) exhibited oversensitivity because it is attributed to the large number of samples produced by the company, prompting the need for a more balanced approach. Implementing the Laney u control chart, Laney demerit control chart, and the AHP-based Laney demerit control charts successfully achieved statistical control in phase I. In contrast, phase II still demonstrated challenges, particularly with the AHP-based Laney Demerit Control Chart detecting the highest number of out-of-control points. This suggests that phase II remains statistically out of control, necessitating further analysis or corrective measures to enhance process stability. Additionally, the process capability analysis indicated that the production process during the specified period lacked capability, as evidenced by a capability index value below one (0.883).

NAN-DETR: noising multi-anchor makes DETR better for object detection.

Object detection plays a crucial role in robotic vision, focusing on accurately identifying and localizing objects within images. However, many existing methods encounter limitations, particularly when it comes to effectively implementing a one-to-many matching strategy. To address these challenges, we propose NAN-DETR (Noising Multi-Anchor Detection Transformer), an innovative framework based on DETR (Detection Transformer). NAN-DETR introduces three key improvements to transformer-based object detection: a decoder-based multi-anchor strategy, a centralization noising mechanism, and the integration of Complete Intersection over Union (CIoU) loss. The multi-anchor strategy leverages multiple anchors per object, significantly enhancing detection accuracy by improving the one-to-many matching process. The centralization noising mechanism mitigates conflicts among anchors by injecting controlled noise into the detection boxes, thereby increasing the robustness of the model. Additionally, CIoU loss, which incorporates both aspect ratio and spatial distance in its calculations, results in more precise bounding box predictions compared to the conventional IoU loss. Although NAN-DETR may not drastically improve real-time processing capabilities, its exceptional performance positions it as a highly reliable solution for diverse object detection scenarios.

Development of an intelligent-based telemetry hexapod robotic system for surveillance of power system components

A robot is an artificial system that performs specific programmed tasks to aid day-to-day human activities. Robot finds applications in industries for large-scale production, healthcare for computer-aided surgeries, security for surveillance of electrical power system equipment, and the military for carrying out dangerous reconnaissance. Most robotic systems employed for different operations use a wheeled mode of locomotion which has limitations over rough terrains and the issue of communication setup of the system. This work developed a hexapod, a six-legged robotic system, to overcome these identified challenges. The robot was developed in tandem with a telemetry system for surveillance. Although, automating the surveillance process using robotic systems has been in the works for some time. However, this work developed a telemetry hexapod robotic system for enhanced surveillance for the security of power systems equipment and other critical infrastructures using a Raspberry Pi for fast, secure data transmission, and precise system synchronization. An inverse kinematics approach was used to determine joint configurations for better positions of each endpoint. A digital camera was integrated into the robot to relay real-time images of adversaries of power system components intelligently. In addition, the system incorporates data processing capability. In the result obtained, the developed telemetry hexapod robotic system receives instructions over a distance slightly above 300 m. It establishes effective control of the telemetry system without any hindrance of a limited range as observed in Bluetooth-controlled systems. The system's performance includes an average efficiency of 98.2 %, latency of 0.48 s at the peak distance, and an endurance of 4 h. Further analysis of the system shows that the corresponding increase in the latency with the distance is negligible. The system performed better than other related systems considered in the literature. The correct implementation of the developed telemetry hexapod robotic system further enhances mobility, stability in rough terrain, reliable communication system. Also, it brings about a notable reduction in component theft in electrical power systems.

Control charts for monitoring the capability of unstable process

Stability is an important criterion to be investigated at the early stage of any industrial process and stable processes are usually preferred over unstable ones. But a process can be unstable and nevertheless capable and, therefore, there is a clear need to monitor such situations to ensure an acceptable capability. In this study we introduce, in a first step, a new Shewhart-type control chart for normal processes based on the yield index S pk to monitor the capability of potentially unstable processes and then, in a second step, we extend it using an Exponentially Weighted Moving Average (EWMA) scheme. Moreover, the optimal values of the control parameters are obtained using a nonlinear optimization problem for which the out-of-control average run length is the objective function. The introduced charts’ sensitivity is examined for deterministic shift sizes. Numerical and graphical analyses are given to display the features of the proposed charts. A simulation study shows, how in such a situation, the introduced tool is successful to recognize changes in the process capability.

CJE-PCHF: Chinese Joint Entity and Relation Extraction Model Based on Progressive Contrastive Learning and Heterogeneous Feature Fusion

The joint extraction of entities and relations is a critical task in information extraction, and its performance directly affects the performance of downstream tasks. However, existing joint extraction models based on deep learning exhibit weak processing capabilities for the phenomenon of multiple pronunciations of one character and multiple characters of one pronunciation when processing Chinese texts, resulting in a performance loss. To address these issues, this paper introduces part-of-speech (POS) and pinyin features to aid the model in learning semantic features that are more contextually appropriate. We propose a Chinese Joint Entity and Relation Extraction Model based on progressive contrastive learning and heterogeneous feature fusion (CJE-PCHF). During model training, an interactive fusion network based on progressive contrastive learning is employed to learn the dependencies between pinyin, POS, and semantic features. This guides the model in heterogeneous feature fusion, capturing higher-order semantic associations between heterogeneous features. On the commonly used DuIE evaluation dataset for joint extraction, our model achieved a significant improvement, with the F1 score increasing by 5.4% compared to the benchmark model CasRel.

Multi-Layered Unsupervised Learning Driven by Signal-to-Noise Ratio-Based Relaying for Vehicular Ad Hoc Network-Supported Intelligent Transport System in eHealth Monitoring.

Every year, about 1.19 million people are killed in traffic accidents; hence, the United Nations has a goal of halving the number of road traffic deaths and injuries by 2030. In line with this objective, technological innovations in telecommunication, particularly brought about by the rise of 5G networks, have contributed to the development of modern Vehicle-to-Everything (V2X) systems for communication. A New Radio V2X (NR-V2X) was introduced in the latest Third Generation Partnership Project (3GPP) releases which allows user devices to exchange information without relying on roadside infrastructures. This, together with Massive Machine Type Communication (mMTC) and Ultra-Reliable Low Latency Communication (URLLC), has led to the significantly increased reliability, coverage, and efficiency of vehicular communication networks. The use of artificial intelligence (AI), especially K-means clustering, has been very promising in terms of supporting efficient data exchange in vehicular ad hoc networks (VANETs). K-means is an unsupervised machine learning (ML) technique that groups vehicles located near each other geographically so that they can communicate with one another directly within these clusters while also allowing for inter-cluster communication via cluster heads. This paper proposes a multi-layered VANET-enabled Intelligent Transportation System (ITS) framework powered by unsupervised learning to optimize communication efficiency, scalability, and reliability. By leveraging AI in VANET solutions, the proposed framework aims to address road safety challenges and contribute to global efforts to meet the United Nations' 2030 target. Additionally, this framework's robust communication and data processing capabilities can be extended to eHealth monitoring systems, enabling real-time health data transmission and processing for continuous patient monitoring and timely medical interventions. This paper's contributions include exploring AI-driven approaches for enhanced data interaction, improved safety in VANET-based ITS environments, and potential applications in eHealth monitoring.

Classical and Bayesian inference of Cpc$\mathcal {C}_{pc}$ for Wilson–Hilferty distribution under progressively first‐failure type‐II censoring samples

AbstractThis study uses two frequentist approaches and the Bayesian method of estimation using progressively first‐failure type‐II censored data to estimate process capability index (PCI), , for the Wilson–Hilferty (WH) distribution. A competitive maximum product of spacing (MPS) method for estimation of is proposed in the frequentist method as an alternative to conventional likelihood (LK)‐based estimation. We have also taken into account the PS function in the Bayesian setup as an alternative to the conventional LK function, and both LK and PS‐based Bayesian estimates are given for . Further, the Bayes estimates are evaluated by applying Tierney–Kadane (TK) approximation method and Markov chain Monte Carlo (MCMC) method with the help of both LK and PS functions and Jeffrey's prior. Additionally, the approximate confidence intervals based on the suggested frequentist techniques as well as the Bayes highest posterior density (HPD) credible intervals are built for the index . In addition, two bootstrap confidence intervals are obtained. In the simulation exercise, the performance of the Bayes and traditional estimates of is assessed in terms of their mean squared errors, and the average width and coverage probabilities of the CIs and HPD intervals are compared. An actual data set from the electronic industries is reanalyzed in order to show the efficacy of the proposed index and estimation methodology.

A lightweight defect detection algorithm for escalator steps

In this paper, we propose an efficient target detection algorithm, ASF-Sim-YOLO, to address issues encountered in escalator step defect detection, such as an excessive number of parameters in the detection network model, poor adaptability, and difficulties in real-time processing of video streams. Firstly, to address the characteristics of escalator step defects, we designed the ASF-Sim-P2 structure to improve the detection accuracy of small targets, such as step defects. Additionally, we incorporated the SimAM (Similarity-based Attention Mechanism) by combining SimAM with SPPF (Spatial Pyramid Pooling-Fast) to enhance the model’s ability to capture key information by assigning importance weights to each pixel. Furthermore, to address the challenge posed by the small size of step defects, we replaced the traditional CIoU (Complete-Intersection-over-Union) loss function with NWD (Normalized Wasserstein Distance), which alleviated the problem of defect missing. Finally, to meet the deployment requirements of mobile devices, we performed channel pruning on the model. The experimental results showed that the improved ASF-Sim-YOLO model achieved an average accuracy (mAP50) of 96.8% on the test data set, which was a 22.1% improvement in accuracy compared to the baseline model. Meanwhile, the computational complexity (in GFLOPS) of the model was reduced to a quarter of that of the baseline model, while the frame rate (FPS) was improved to 575.1. Compared with YOLOv3-tiny, YOLOv5s, YOLOv8s, Faster-RCNN, TOOD, RTMDET and other deep learning-based target recognition algorithms, ASF-Sim-YOLO has better detection accuracy and real-time processing capability. These results demonstrate that ASF-Sim-YOLO effectively balances lightweight design and performance improvement, making it highly suitable for real-time detection of step defects, which can meet the demands of escalator inspection operations.

BEW-YOLOv8: A deep learning model for multi-scene and multi-scale flood depth estimation

As global climate change intensifies, the frequency of extreme weather events, including urban flooding, has risen, posing a significant threat to the safety of urban areas. Efficient and precise flood depth monitoring is crucial for a dynamic understanding of disaster conditions, enabling informed decision-making and reducing the potential for significant loss of life and property. Although object detection models have shown promise in estimating flood depths and have achieved notable results, enhancements are still needed in applications across multi-scene and multi-scale. This research enhances the YOLOv8 model by integrating Bidirectional Feature Pyramid Network (BiFPN), Effective Squeeze and Excitation (EffectiveSE), and Wise-IoU (WIoU), targeting submerged cars to develop the BEW-YOLOv8 model. Testing shows that the BEW-YOLOv8 model improves Precision, Recall, F1 Score, mAP@0.5, and mAP@0.5–0.95 by 14.1%, 2.2%, 7.9%, 8.7%, and 6.3%, respectively, compared to the unmodified YOLOv8 model. Compared with other existing models, BEW-YOLOv8 achieves the best performance with less than one-tenth of the parameters. Furthermore, this study verifies that the BEW-YOLOv8 model holds considerable promise for further enhancements with dataset expansions. Its precision and dependability make the BEW-YOLOv8 model well-suited for real-time flood depth assessment across multiple scenes and scales. The model’s capability for real-time processing supports urgent flood response needs, offering solid technical assistance for managing urban floods and mitigating disaster risks.

Large scale multi-GPU based parallel traffic simulation for accelerated traffic assignment and propagation

Traffic simulation is a critical tool for congestion analysis, travel time estimation, and route optimization in urban planning, benefiting navigation apps, transportation network companies, and state agencies. Traditionally, traffic micro-simulation frameworks are based on road segments and can only support a limited number of main roads. Efficient traffic simulation on a regional scale remains a significant challenge due to the complexity of urban mobility and the large scale of spatiotemporal data. This paper introduces a Large Scale Multi-GPU Parallel Computing based Regional Scale Traffic Simulation Framework (LPSim), which leverages graphical processing unit (GPU) parallel computing to address these challenges. LPSim utilizes a multi-GPU architecture to simulate extensive and dynamic traffic networks with high fidelity and reduced computation time. Using the parallel processing capabilities of GPUs, LPSim can perform tens of millions of individual vehicle dynamics simulations simultaneously, significantly outperforming traditional CPU-based approaches. The framework is designed to be scalable and can easily accommodate the increasing complexity of traffic simulations. We present the theory behind GPU-based traffic simulation, the architecture of single- and multi-GPU based simulations, and the graph partition strategies that enhance computation resource load balance. Our experimental results demonstrate the effectiveness of LPSim in simulating large-scale traffic scenarios. LPSim is capable of completing simulations of 2.82 million trips in just 6.28 min on a single GPU machine equipped with 5120 CUDA cores (Tesla V100-SXM2). Furthermore, utilizing a Google Cloud instance with two NVIDIA V100 GPUs, which collectively offer 10240 CUDA cores, LPSim successfully simulates 9.01 million trips within 21.16 min. We further tested our simulator with the same demand on dual NVIDIA A100-PCIE-40GB GPUs, which finished the simulation in 0.0398 h, approximately 113 times faster than the same simulation scenario running on an Intel(R) Xeon(R) Gold 6326 CPU @ 2.90 GHz, which takes 4.49 h to complete. This performance not only demonstrates its speed and scalability advantages over traditional simulation techniques but also highlights LPSim’s unique position as the first traffic simulation framework that is scalable for both single- and multiple-GPU configurations. Consequently, LPSim provides an invaluable tool for individuals and extensive research teams alike, enabling the acquisition of large-scale traffic simulation results in a time-efficient manner. LPSim code is available at: https://github.com/Xuan-1998/LPSim