[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Sign in to use this feature.

Years

Between: -

Subjects

remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline

Journals

remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline

Article Types

Countries / Regions

remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline

Search Results (4,575)

Search Parameters:
Keywords = edge-computing

Order results
Result details
Results per page
Select all
Export citation of selected articles as:
35 pages, 3825 KiB  
Article
Research on Unmanned Aerial Vehicle Emergency Support System and Optimization Method Based on Gaussian Global Seagull Algorithm
by Songyue Han, Mingyu Wang, Junhong Duan, Jialong Zhang and Dongdong Li
Drones 2024, 8(12), 763; https://doi.org/10.3390/drones8120763 (registering DOI) - 17 Dec 2024
Abstract
In emergency rescue scenarios, drones can be equipped with different payloads as needed to aid in tasks such as disaster reconnaissance, situational awareness, communication support, and material assistance. However, rescue missions typically face challenges such as limited reconnaissance boundaries, heterogeneous communication networks, complex [...] Read more.
In emergency rescue scenarios, drones can be equipped with different payloads as needed to aid in tasks such as disaster reconnaissance, situational awareness, communication support, and material assistance. However, rescue missions typically face challenges such as limited reconnaissance boundaries, heterogeneous communication networks, complex data fusion, high task latency, and limited equipment endurance. To address these issues, an unmanned emergency support system tailored for emergency rescue scenarios is designed. This system leverages 5G edge computing technology to provide high-speed and flexible network access along with elastic computing power support, reducing the complexity of data fusion across heterogeneous networks. It supports the control and data transmission of drones through the separation of the control plane and the data plane. Furthermore, by applying the Tammer decomposition method to break down the system optimization problem, the Global Learning Seagull Algorithm for Gaussian Mapping (GLSOAG) is proposed to jointly optimize the system’s energy consumption and latency. Through simulation experiments, the GLSOAG demonstrates significant advantages over the Seagull Optimization Algorithm (SOA), Particle Swarm Optimization (PSO), and Beetle Antennae Search Algorithm (BAS) in terms of convergence speed, optimization accuracy, and stability. The system optimization approach effectively reduces the system’s energy consumption and latency costs. Overall, our work alleviates the pain points faced in rescue scenarios to some extent. Full article
18 pages, 31070 KiB  
Article
Flow-Induced Stress Analysis of a Large Francis Turbine Under Different Loads in a Wide Operation Range
by Xingxing Huang, Hua Ou, Hao Huang, Zhengwei Wang and Gang Wang
Appl. Sci. 2024, 14(24), 11782; https://doi.org/10.3390/app142411782 - 17 Dec 2024
Abstract
Francis turbines, being widely used in hydropower plants, operate under different loads which significantly affect their hydraulic characteristics and structural dynamics. It is essential to carry out the flow-induced dynamics analysis of the large prototype Francis turbines under different loads in a wide [...] Read more.
Francis turbines, being widely used in hydropower plants, operate under different loads which significantly affect their hydraulic characteristics and structural dynamics. It is essential to carry out the flow-induced dynamics analysis of the large prototype Francis turbines under different loads in a wide load operation range to optimize the hydraulic performance, ensure structural reliability, and prevent mechanical failure. This work analyzes the flow-induced dynamics of a large Francis turbine prototype with a rated power of 46 MW. Computer-aided design (CAD) models of the Francis turbine unit are first constructed, including the fluid and structural domains. After generating the computational meshes of the flow passages in the Francis turbine unit, Computational fluid dynamics (CFD) calculations are carried out under four typical operating conditions from 25% load to 100% load, and the pressure files obtained from CFD calculations are applied to the finite element model to analyze the flow-induced stresses of the runner. The results show that the pressure inside the Francis turbine runner decreases gradually from the spiral case to the draft tube under 25%, 50%, 75%, and 100% loads, but the local pressure distribution in the crown chamber of the Francis turbine unit varies under different loads. The locations of the maximum stress of the runner under the four different operating conditions vary with the power output. The flow-induced maximum stress of the runner at 25% load is located on the chamfer of the connection between the blade trailing edge and the crown. But from 50% load to 100% load, the maximum stress of the runner appears on the chamfer of the connection between the blade leading edge and the band. From 25% load to full load, the maximum stress of the unit is one-fifth of the yield stress of the runner material, and the runner will not be damaged during normal use. The calculation method with a fully three-dimensional fluid–structure interaction (FSI) method and the conclusions proposed in this study can provide important references for the design and evaluation of other hydraulic turbine units. Full article
Show Figures

Figure 1

Figure 1
<p>Damage of Francis turbine runners [<a href="#B6-applsci-14-11782" class="html-bibr">6</a>].</p>
Full article ">Figure 2
<p>Flow domain of the CFD analysis of the Francis turbine unit.</p>
Full article ">Figure 3
<p>CFD calculation model of the Francis turbine unit.</p>
Full article ">Figure 4
<p>Normalized pressure distribution in the flow domain of the Francis turbine unit under different loads: (<b>a</b>) 25% load, (<b>b</b>) 50% load, (<b>c</b>) 75% load, (<b>d</b>) 100% load.</p>
Full article ">Figure 5
<p>Local pressure distribution of crown chamber of the Francis turbine under different loads: (<b>a</b>) 25% load, (<b>b</b>) 50% load, (<b>c</b>) 75% load, (<b>d</b>) 100% load.</p>
Full article ">Figure 5 Cont.
<p>Local pressure distribution of crown chamber of the Francis turbine under different loads: (<b>a</b>) 25% load, (<b>b</b>) 50% load, (<b>c</b>) 75% load, (<b>d</b>) 100% load.</p>
Full article ">Figure 6
<p>Local pressure distribution around guide vanes of the Francis turbine under different loads: (<b>a</b>) 25% load, (<b>b</b>) 50% load, (<b>c</b>) 75% load, (<b>d</b>) 100% load.</p>
Full article ">Figure 7
<p>Local pressure distribution of Francis turbine runner under different loads: (<b>a</b>) 25% load, (<b>b</b>) 50% load, (<b>c</b>) 75% load, (<b>d</b>) 100% load.</p>
Full article ">Figure 8
<p>Flow streamlines of the Francis turbine under different loads: (<b>a</b>) 25% load, (<b>b</b>) 50% load, (<b>c</b>) 75% load, (<b>d</b>) 100% load.</p>
Full article ">Figure 8 Cont.
<p>Flow streamlines of the Francis turbine under different loads: (<b>a</b>) 25% load, (<b>b</b>) 50% load, (<b>c</b>) 75% load, (<b>d</b>) 100% load.</p>
Full article ">Figure 9
<p>Finite element model of the Francis turbine for the flow-induced stress analysis.</p>
Full article ">Figure 10
<p>Transfer the pressure data to the finite element model of the Francis turbine.</p>
Full article ">Figure 11
<p>Normalized stress distribution of the Francis runner at 25% load.</p>
Full article ">Figure 12
<p>Normalized stress distribution of the Francis runner from 50% to 100% load.</p>
Full article ">Figure 13
<p>Normalized stress of the Francis runner from 25% to 100% load.</p>
Full article ">Figure 14
<p>Cracks on Francis turbine runners from other power plants.</p>
Full article ">
16 pages, 4714 KiB  
Article
Computer Vision System for Multi-Robot Construction Waste Management: Integrating Cloud and Edge Computing
by Zeli Wang, Xincong Yang, Xianghan Zheng, Daoyin Huang and Binfei Jiang
Buildings 2024, 14(12), 3999; https://doi.org/10.3390/buildings14123999 - 17 Dec 2024
Abstract
Sorting is an important construction waste management tool to increase recycling rates and reduce pollution. Previous studies have used robots to improve the efficiency of construction waste recycling. However, in large construction sites, it is difficult for a single robot to accomplish the [...] Read more.
Sorting is an important construction waste management tool to increase recycling rates and reduce pollution. Previous studies have used robots to improve the efficiency of construction waste recycling. However, in large construction sites, it is difficult for a single robot to accomplish the task quickly, and multiple robots working together are a better option. Most construction waste recycling robotic systems are developed based on a client-server framework, which means that all robots need to be continuously connected to their respective cloud servers. Such systems are low in robustness in complex environments and waste a lot of computational resources. Therefore, in this paper, we propose a pixel-level automatic construction waste recognition platform with high robustness and low computational resource requirements by combining multiple computer vision technologies with edge computing and cloud computing platforms. Experiments show that the computing platform proposed in this study can achieve a recognition speed of 23.3 fps and a recognition accuracy of 90.81% at the edge computing platform without the help of network and cloud servers. This is 23 times faster than the algorithm used in previous research. Meanwhile, the computing platform proposed in this study achieves 93.2% instance segmentation accuracy on the cloud server side. Notably, this system allows multiple robots to operate simultaneously at the same construction site using only a single server without compromising efficiency, which significantly reduces costs and promotes the adoption of automated construction waste recycling robots. Full article
(This article belongs to the Section Construction Management, and Computers & Digitization)
Show Figures

Figure 1

Figure 1
<p>System architecture of construction waste collection platform.</p>
Full article ">Figure 2
<p>Construction waste collection platform workflow.</p>
Full article ">Figure 3
<p>YOLO algorithm.</p>
Full article ">Figure 4
<p>A part of the data set.</p>
Full article ">Figure 5
<p>Pseudocode of data format conversion.</p>
Full article ">Figure 6
<p>Data augmentation result.</p>
Full article ">Figure 7
<p>Pseudocode of data augmentation.</p>
Full article ">Figure 8
<p>Workflow of computer vision system in construction waste collection platform.</p>
Full article ">Figure 9
<p>Construction waste collecting robot.</p>
Full article ">Figure 10
<p>Comparison of accuracy between YOLO-Tiny and YOLO.</p>
Full article ">Figure 11
<p>Missed detection problem of YOLO-Tiny.</p>
Full article ">Figure 12
<p>Instance segmentation results of YOLACT.</p>
Full article ">
34 pages, 3160 KiB  
Article
Energy-Efficient Collision-Free Machine/AGV Scheduling Using Vehicle Edge Intelligence
by Zhengying Cai, Jingshu Du, Tianhao Huang, Zhuimeng Lu, Zeya Liu and Guoqiang Gong
Sensors 2024, 24(24), 8044; https://doi.org/10.3390/s24248044 - 17 Dec 2024
Abstract
With the widespread use of autonomous guided vehicles (AGVs), avoiding collisions has become a challenging problem. Addressing the issue is not straightforward since production efficiency, collision avoidance, and energy consumption are conflicting factors. This paper proposes a novel edge computing method based on [...] Read more.
With the widespread use of autonomous guided vehicles (AGVs), avoiding collisions has become a challenging problem. Addressing the issue is not straightforward since production efficiency, collision avoidance, and energy consumption are conflicting factors. This paper proposes a novel edge computing method based on vehicle edge intelligence to solve the energy-efficient collision-free machine/AGV scheduling problem. First, a vehicle edge intelligence architecture was built, and the corresponding state transition diagrams for collision-free scheduling were developed. Second, the energy-efficient collision-free machine/AGV scheduling problem was modeled as a multi-objective function with electric capacity constraints, where production efficiency, collision prevention, and energy conservation were comprehensively considered. Third, an artificial plant community algorithm was explored based on the edge intelligence of AGVs. The proposed method utilizes a heuristic search and the swarm intelligence of multiple AGVs to realize energy-efficient collision-free scheduling and is suitable for deploying on embedded platforms for edge computing. Finally, a benchmark dataset was developed, and some benchmark experiments were conducted, where the results revealed that the proposed heuristic method could effectively instruct multiple automatic guided vehicles to avoid collisions with high energy efficiency. Full article
(This article belongs to the Section Vehicular Sensing)
Show Figures

Figure 1

Figure 1
<p>The VEI-based architecture for a machine/AGV system.</p>
Full article ">Figure 2
<p>The task state transition diagram.</p>
Full article ">Figure 3
<p>The machine/AGV state transition diagram.</p>
Full article ">Figure 4
<p>A schematic diagram of the APC-based edge computing framework.</p>
Full article ">Figure 5
<p>The APC algorithm flow for the energy-efficient collision-free machine/AGV scheduling problem.</p>
Full article ">Figure 6
<p>Benchmark roadmap in a production workshop.</p>
Full article ">Figure 7
<p>The processing time for energy-efficient collision-free AGV scheduling. (The same color is used for different processes of the same task, and different colors are randomly assigned to different tasks).</p>
Full article ">Figure 8
<p>The transportation time for energy-efficient collision-free AGV scheduling. (The same color is used for different transportations of the same AGV, and different colors are randomly assigned to different AGVs).</p>
Full article ">Figure 9
<p>The convergence curves of the proposed APC algorithm.</p>
Full article ">Figure 10
<p>Comparison of parameter changes. ((<b>a</b>) The trends of the four main indicators with different machine numbers; (<b>b</b>) the trend of the four main indicators with different AGV numbers).</p>
Full article ">
13 pages, 6856 KiB  
Article
Mind the Step: An Artificial Intelligence-Based Monitoring Platform for Animal Welfare
by Andrea Michielon, Paolo Litta, Francesca Bonelli, Gregorio Don, Stefano Farisè, Diana Giannuzzi, Marco Milanesi, Daniele Pietrucci, Angelica Vezzoli, Alessio Cecchinato, Giovanni Chillemi, Luigi Gallo, Marcello Mele and Cesare Furlanello
Sensors 2024, 24(24), 8042; https://doi.org/10.3390/s24248042 - 17 Dec 2024
Abstract
We present an artificial intelligence (AI)-enhanced monitoring framework designed to assist personnel in evaluating and maintaining animal welfare using a modular architecture. This framework integrates multiple deep learning models to automatically compute metrics relevant to assessing animal well-being. Using deep learning for AI-based [...] Read more.
We present an artificial intelligence (AI)-enhanced monitoring framework designed to assist personnel in evaluating and maintaining animal welfare using a modular architecture. This framework integrates multiple deep learning models to automatically compute metrics relevant to assessing animal well-being. Using deep learning for AI-based vision adapted from industrial applications and human behavioral analysis, the framework includes modules for markerless animal identification and health status assessment (e.g., locomotion score and body condition score). Methods for behavioral analysis are also included to evaluate how nutritional and rearing conditions impact behaviors. These models are initially trained on public datasets and then fine-tuned on original data. We demonstrate the approach through two use cases: a health monitoring system for dairy cattle and a piglet behavior analysis system. The results indicate that scalable deep learning and edge computing solutions can support precision livestock farming by automating welfare assessments and enabling timely, data-driven interventions. Full article
Show Figures

Figure 1

Figure 1
<p>The dairy cattle two-camera setup. The side-view camera is used for the estimation of pose and locomotion features, while the top-view camera supports markerless identification.</p>
Full article ">Figure 2
<p>Architecture of the AI health status assessment system.</p>
Full article ">Figure 3
<p>Ordered list of pose key points and an annotation example over a training set image.</p>
Full article ">Figure 4
<p>Structure of the locomotion score classifier. The process starts with a clip, divided into a sequence of frames (<b>left</b>), which are then processed into a sequence of meaningful features (<b>middle</b>). These features are evaluated by the classifier, resulting in a locomotion score prediction (<b>right</b>).</p>
Full article ">Figure 5
<p>Structure of the body condition score classifier. The process begins with a snapshot, from which the subject is extracted using a detection model. The extracted subject is then evaluated by the body condition score classifier, producing an estimated BCS value.</p>
Full article ">Figure 6
<p>Structure of the markerless identification algorithm. The process starts with a snapshot, where the subject is extracted using a detection model (<b>left</b>). The extracted subject is then encoded into a latent space via an encoder network (<b>middle</b>). Finally, the closest anchor point is identified using the Euclidean distance (<b>right</b>).</p>
Full article ">Figure 7
<p>Outline of the pipeline for behavioral analysis.</p>
Full article ">Figure 8
<p>(<b>A</b>) Example of how perspective transformation is applied to map pixel coordinates to real-world coordinates, aimed at achieving more precise quantification and characterization of traveled distances. (<b>B</b>) Example of body-part segmentation, with the five segmentation classes (mouth, right ear, left ear, body, and tail) highlighted in different colors and labeled accordingly.</p>
Full article ">Figure 9
<p>The pig interactions pipeline. (<b>A</b>) Inference annotations; (<b>B</b>) Daily interaction distribution for Week 1 in Pen 1: total interaction time (minutes) per pig per hour; (<b>C</b>) Classification of interactions based on body-part segmentation; (<b>D</b>) Example of quantitative analysis of an interaction.</p>
Full article ">Figure 10
<p>Ear tag-based and livestock mark-based identification systems.</p>
Full article ">Figure 11
<p>Visual comparison of different instances (with the two on the left from in-field acquisitions, and the one on the right from the NWAFU dataset [<a href="#B9-sensors-24-08042" class="html-bibr">9</a>]), with the annotated bounding boxes shown in green, and the predicted bounding boxes in purple.</p>
Full article ">Figure 12
<p>Visual comparison of different instances (with the two on the left from in-field acquisitions, and the one on the right from the NWAFU dataset [<a href="#B9-sensors-24-08042" class="html-bibr">9</a>]), with the annotated key points shown in green, the predicted key points shown in purple, and the distance between the predicted coordinates and their ground-truth values highlighted in red.</p>
Full article ">
20 pages, 1708 KiB  
Article
Sustainability in Industry 4.0: Edge Computing Microservices as a New Approach
by Leandro Colevati dos Santos, Maria Lucia Pereira da Silva and Sebastião Gomes dos Santos Filho
Sustainability 2024, 16(24), 11052; https://doi.org/10.3390/su162411052 - 17 Dec 2024
Viewed by 128
Abstract
The importance of the electronics sector in the modern world is unquestionable, as it demonstrates clean technology, dry processes, and efficient design, which favor Industry 4.0 and sustainability. Nonetheless, the large number of instruments developed, and their correspondent quick obsolescence, imply an increment [...] Read more.
The importance of the electronics sector in the modern world is unquestionable, as it demonstrates clean technology, dry processes, and efficient design, which favor Industry 4.0 and sustainability. Nonetheless, the large number of instruments developed, and their correspondent quick obsolescence, imply an increment in electronic waste. Therefore, in this work, with the aim of diminishing obsolescence, we developed and customized one application that runs independently of systems and takes advantage of the existing computing structures. The application is a new edge computing structure (the AIFC) that is based on an enterprise service bus (ESB) developed in decentralized microservices. In this study, we conducted action research involving the collaboration of researchers and practitioners, and the tests involved six different scenarios; they used existing low-cost, basic computing environments and ranged from the proof of concept, prototype, minimum viable product, and scalability to the roadmap for the structure implementation. The six scenarios emulated sections of a small and medium-sized enterprise (SME), and all the developed microservices communicate with each other to perform data filtering, processing, storage, query, and sensor data acquisition. The results show that it is possible to carry out these functions with low latency and without any decrement or even increase in performance when compared with more conventional cloud computing structures, and it is also possible to manipulate different products that do not have single, consolidated structures. Moreover, there is no need to update machines or communication structures, which are the main factors of rapid obsolescence. Therefore, following the steps of the AIFC development, the results from the proof of concept to the minimum viable product and scalability tests correspond to a roadmap for a sustainable solution and are an important tool for both Industry 4.0 and SMEs. Full article
Show Figures

Figure 1

Figure 1
<p>Interconnection between Industry 4.0, SMEs, electronics, and sustainability.</p>
Full article ">Figure 2
<p>AIFC block diagram.</p>
Full article ">Figure 3
<p>The proof-of-concept step.</p>
Full article ">Figure 4
<p>Elements added to the ESB for the MVP.</p>
Full article ">Figure 5
<p>Elements added to the ESB for the MVP.</p>
Full article ">Figure 6
<p>Proposed roadmap for developing the MVP AIFC structure.</p>
Full article ">
29 pages, 9712 KiB  
Article
Cloud–Edge–End Collaborative Federated Learning: Enhancing Model Accuracy and Privacy in Non-IID Environments
by Ling Li, Lidong Zhu and Weibang Li
Sensors 2024, 24(24), 8028; https://doi.org/10.3390/s24248028 - 16 Dec 2024
Viewed by 171
Abstract
Cloud–edge–end computing architecture is crucial for large-scale edge data processing and analysis. However, the diversity of terminal nodes and task complexity in this architecture often result in non-independent and identically distributed (non-IID) data, making it challenging to balance data heterogeneity and privacy protection. [...] Read more.
Cloud–edge–end computing architecture is crucial for large-scale edge data processing and analysis. However, the diversity of terminal nodes and task complexity in this architecture often result in non-independent and identically distributed (non-IID) data, making it challenging to balance data heterogeneity and privacy protection. To address this, we propose a privacy-preserving federated learning method based on cloud–edge–end collaboration. Our method fully considers the three-tier architecture of cloud–edge–end systems and the non-IID nature of terminal node data. It enhances model accuracy while protecting the privacy of terminal node data. The proposed method groups terminal nodes based on the similarity of their data distributions and constructs edge subnetworks for training in collaboration with edge nodes, thereby mitigating the negative impact of non-IID data. Furthermore, we enhance WGAN-GP with attention mechanism to generate balanced synthetic data while preserving key patterns from original datasets, reducing the adverse effects of non-IID data on global model accuracy while preserving data privacy. In addition, we introduce data resampling and loss function weighting strategies to mitigate model bias caused by imbalanced data distribution. Experimental results on real-world datasets demonstrate that our proposed method significantly outperforms existing approaches in terms of model accuracy, F1-score, and other metrics. Full article
(This article belongs to the Section Sensor Networks)
Show Figures

Figure 1

Figure 1
<p>Federated learning framework for cloud–edge–end architecture.</p>
Full article ">Figure 2
<p>Illustration of non-IID client data in federated learning.</p>
Full article ">Figure 3
<p>Generator structure of WGAN-GP after adding the self-attention layer.</p>
Full article ">Figure 4
<p>Discriminator structure of WGAN-GP after adding the self-attention layer.</p>
Full article ">Figure 5
<p>Examples of original MNIST dataset and WGAN-GP generated dataset.</p>
Full article ">Figure 6
<p>Examples of AnnualCrop label category from original EuroSAT and WGAN-GP generated dataset of the same label category.</p>
Full article ">Figure 7
<p>Performance of CEECFed, FedGS, FedAvg, and FedSGD based on the original MNIST dataset. (<b>a</b>) Accuracy, (<b>b</b>) Precision, (<b>c</b>) Recall, (<b>d</b>) F1-Score, (<b>e</b>) Average Loss.</p>
Full article ">Figure 8
<p>Performance of CEECFed, FedGS, FedAvg, and FedSGD based on the original EuroSAT dataset. (<b>a</b>) Accuracy, (<b>b</b>) Precision, (<b>c</b>) Recall, (<b>d</b>) F1-Score, (<b>e</b>) Average Loss.</p>
Full article ">Figure 9
<p>Performance of CEECFed based on original MNIST dataset and WGAN-GP generated datasets. (<b>a</b>) Accuracy, (<b>b</b>) Precision, (<b>c</b>) Recall, (<b>d</b>) F1-Score, (<b>e</b>) Average Loss.</p>
Full article ">Figure 10
<p>Performance of FedAvg based on original MNIST dataset and WGAN-GP generated datasets. (<b>a</b>) Accuracy, (<b>b</b>) Precision, (<b>c</b>) Recall, (<b>d</b>) F1-Score, (<b>e</b>) Average Loss.</p>
Full article ">Figure 11
<p>Performance of FedSGD based on original MNIST dataset and WGAN-GP generated datasets. (<b>a</b>) Accuracy, (<b>b</b>) Precision, (<b>c</b>) Recall, (<b>d</b>) F1-Score, (<b>e</b>) Average Loss.</p>
Full article ">
22 pages, 4996 KiB  
Article
Identification of Spatial and Symbolic City Image Elements Through Social Media Data: A Case Study of Hangzhou
by Jiaqi Wang, Yu Shi, Weishun Xu and Yue Wu
Land 2024, 13(12), 2194; https://doi.org/10.3390/land13122194 - 16 Dec 2024
Viewed by 252
Abstract
Despite emerging empirical findings and computational tools that extend city image research to include social dimensions beyond visual perception, methodologies for effectively identifying and analyzing the relationships between the five city image elements remain underdeveloped. This paper addresses the gap by proposing a [...] Read more.
Despite emerging empirical findings and computational tools that extend city image research to include social dimensions beyond visual perception, methodologies for effectively identifying and analyzing the relationships between the five city image elements remain underdeveloped. This paper addresses the gap by proposing a big data-driven method, integrating Weibo check-in data, Baidu Map POI, and ArcGIS algorithms to identify city image elements and further reveal a city’s overall morphological characteristics. Based on different modes of observation, city image elements are categorized as spatial descriptors (“districts”, “nodes”, and “paths”) and symbolic descriptors (“landmarks” and “edges”). Taking Hangzhou as a case study, the findings show a strong alignment between urban development achievements and the distribution patterns of city image elements. “Districts” and “landmarks” stand out as the most prominent, reflecting functional zoning and urban maturity, while “nodes” emphasize the city’s polycentric structure. “Paths” offer clear insight into the city’s development trajectory, while “edges” appear to be legible only in relation to other elements. This method innovates cognitive mapping by merging real-world perceptions with algorithmic precision, offering a valuable tool for understanding urban morphology, monitoring development changes, and fostering participatory urban design. Full article
Show Figures

Figure 1

Figure 1
<p>Study area of Hangzhou, with key natural resources and urban cores marked on the map.</p>
Full article ">Figure 2
<p>Distribution map for Hangzhou Weibo check-in data with POI classification.</p>
Full article ">Figure 3
<p>The cognitive map of “Districts” in Hangzhou.</p>
Full article ">Figure 4
<p>Statistics of dominant functional types within “Districts” in Hangzhou.</p>
Full article ">Figure 5
<p>The cognitive map of “Nodes” in Hangzhou’s main urban area.</p>
Full article ">Figure 6
<p>Statistics of “Landmarks” in six administrative districts of Hangzhou.</p>
Full article ">Figure 7
<p>The cognitive map of “Landmarks” in Hangzhou.</p>
Full article ">Figure 8
<p>The overlay cognitive map of “Districts”, “Nodes”, and “Landmarks” in Hangzhou.</p>
Full article ">Figure 9
<p>The cognitive map of “Paths” in Hangzhou.</p>
Full article ">Figure 10
<p>The cognitive map of “Edges” in Hangzhou: (<b>a</b>) Different types of “edges”; (<b>b</b>) “Edges” overlaid with “nodes” and “districts”; (<b>c</b>) “Edges” overlaid with “paths”.</p>
Full article ">Figure A1
<p>Check-in count distribution based on check-in locations (determining landmark thresholds using the elbow method).</p>
Full article ">
27 pages, 5284 KiB  
Article
A Study on the Feasibility of Natural Frequency-Based Crack Detection
by Xutao Sun, Sinniah Ilanko, Yusuke Mochida and Rachael C. Tighe
Appl. Sci. 2024, 14(24), 11712; https://doi.org/10.3390/app142411712 - 16 Dec 2024
Viewed by 277
Abstract
Owing to the long-standing statement that natural frequency-based crack detection is not sensitive enough to localised damage to identify small cracks, many natural frequency-based crack detection methods are validated by detecting cracks of moderate size. However, a direct comparison between the crack severity [...] Read more.
Owing to the long-standing statement that natural frequency-based crack detection is not sensitive enough to localised damage to identify small cracks, many natural frequency-based crack detection methods are validated by detecting cracks of moderate size. However, a direct comparison between the crack severity causing a measurable natural frequency change and the crack severity reaching the initiation of crack propagation or leading to brittle fracture is constantly ignored. Without this understanding, it is debatable whether the presented crack detection methods are feasible in practical situations. Through natural frequency calculation and linear elastic fracture mechanics, this study is dedicated to filling the above gap in knowledge. To directly utilize the solution of stress intensity factor, common fracture toughness test specimens featuring a single-edge crack are used. These specimens are essentially cracked rectangular plates under uniform uniaxial tension. Considering the stress resultants obtained via the extended finite element method, the natural frequency of the loaded cracked plates is calculated using the Rayleigh–Ritz method incorporating corner functions. In addition, assuming the specimens as structural components under remote uniform tension, the development of critical load versus crack length is derived based on the solution of the stress intensity factor. Thus, critical crack lengths corresponding to a series of safety factors are obtained by equating service load with critical load. After obtaining natural frequencies of the cracked plates with critical crack lengths, the natural frequency drop caused by a critical crack can be computed. Hence, the critical crack length can be compared with the crack length when the frequency drop is measurable. It is found that the brittleness of the employed metals plays a vital role in the feasibility of natural frequency-based crack detection. Full article
(This article belongs to the Special Issue Advances in Structural Health Monitoring in Civil Engineering)
Show Figures

Figure 1

Figure 1
<p>The SENT specimen for fracture toughness test.</p>
Full article ">Figure 2
<p>The methodology of the current study.</p>
Full article ">Figure 3
<p>The divisions of the single-edge cracked plate.</p>
Full article ">Figure 4
<p>The mesh for Gaussian integration and the distribution of integration points.</p>
Full article ">Figure 5
<p>The finite element model discretization in ANSYS.</p>
Full article ">Figure 6
<p>The stress distribution in the cracked square plate when <math display="inline"><semantics> <mrow> <mi>σ</mi> <mo>=</mo> <mn>1.3255</mn> <mo>×</mo> <msup> <mrow> <mn>10</mn> </mrow> <mrow> <mn>6</mn> </mrow> </msup> <mo> </mo> <mi mathvariant="normal">P</mi> <mi mathvariant="normal">a</mi> </mrow> </semantics></math>.</p>
Full article ">Figure 7
<p>Mode shapes when <span class="html-italic">I</span> = <span class="html-italic">J</span> = 8 and <span class="html-italic">N</span> = 3.</p>
Full article ">Figure 8
<p>The development of critical crack length ratio as safety factor increases.</p>
Full article ">Figure A1
<p>Typical crack growth behaviour of ductile materials (for a load-controlled system).</p>
Full article ">Figure A2
<p>Dimensions and coordinates of a side-cracked plate.</p>
Full article ">Figure A3
<p>Mode shapes of a completely free square plate with a slant side crack when <math display="inline"><semantics> <mrow> <mi>I</mi> <mo>=</mo> <mi>J</mi> <mo>=</mo> </mrow> </semantics></math> 8 and <math display="inline"><semantics> <mrow> <mi>N</mi> <mo>=</mo> </mrow> </semantics></math> 3 (<math display="inline"><semantics> <mrow> <mi>α</mi> <mo>=</mo> </mrow> </semantics></math> 30, <math display="inline"><semantics> <mrow> <mi>d</mi> <mo>/</mo> <mi>b</mi> <mo>=</mo> </mrow> </semantics></math> 0.75, and <math display="inline"><semantics> <mrow> <mi>c</mi> <mo>/</mo> <mi>a</mi> <mo>=</mo> </mrow> </semantics></math> 0.5).</p>
Full article ">Figure A4
<p>Mode shapes of a simply supported square plate with a slant side crack when <math display="inline"><semantics> <mrow> <mi>I</mi> <mo>=</mo> <mi>J</mi> <mo>=</mo> </mrow> </semantics></math> 8 and <math display="inline"><semantics> <mrow> <mi>N</mi> <mo>=</mo> </mrow> </semantics></math> 3 (<math display="inline"><semantics> <mrow> <mi>α</mi> <mo>=</mo> </mrow> </semantics></math> 15, <math display="inline"><semantics> <mrow> <mi>d</mi> <mo>/</mo> <mi>b</mi> <mo>=</mo> </mrow> </semantics></math> 0.75, and <math display="inline"><semantics> <mrow> <mi>c</mi> <mo>/</mo> <mi>a</mi> <mo>=</mo> </mrow> </semantics></math> 0.5).</p>
Full article ">
19 pages, 3567 KiB  
Article
Multi-Agent Reinforcement Learning-Based Computation Offloading for Unmanned Aerial Vehicle Post-Disaster Rescue
by Lixing Wang and Huirong Jiao
Sensors 2024, 24(24), 8014; https://doi.org/10.3390/s24248014 - 15 Dec 2024
Viewed by 328
Abstract
Natural disasters cause significant losses. Unmanned aerial vehicles (UAVs) are valuable in rescue missions but need to offload tasks to edge servers due to their limited computing power and battery life. This study proposes a task offloading decision algorithm called the multi-agent deep [...] Read more.
Natural disasters cause significant losses. Unmanned aerial vehicles (UAVs) are valuable in rescue missions but need to offload tasks to edge servers due to their limited computing power and battery life. This study proposes a task offloading decision algorithm called the multi-agent deep deterministic policy gradient with cooperation and experience replay (CER-MADDPG), which is based on multi-agent reinforcement learning for UAV computation offloading. CER-MADDPG emphasizes collaboration between UAVs and uses historical UAV experiences to classify and obtain optimal strategies. It enables collaboration among edge devices through the design of the ’critic’ network. Additionally, by defining good and bad experiences for UAVs, experiences are classified into two separate buffers, allowing UAVs to learn from them, seek benefits, avoid harm, and reduce system overhead. The performance of CER-MADDPG was verified through simulations in two aspects. First, the influence of key hyperparameters on performance was examined, and the optimal values were determined. Second, CER-MADDPG was compared with other baseline algorithms. The results show that compared with MADDPG and stochastic game-based resource allocation with prioritized experience replay, CER-MADDPG achieves the lowest system overhead and superior stability and scalability. Full article
(This article belongs to the Section Intelligent Sensors)
Show Figures

Figure 1

Figure 1
<p>Edge computing architecture.</p>
Full article ">Figure 2
<p>CER-MADDPG algorithm structure.</p>
Full article ">Figure 3
<p>Improved critic network structure.</p>
Full article ">Figure 4
<p>Good and bad behavior guidance model.</p>
Full article ">Figure 5
<p>Selection of learning rates for the critic and actor networks.</p>
Full article ">Figure 6
<p>Selection of replay buffer size.</p>
Full article ">Figure 7
<p>Selection of <math display="inline"><semantics> <msub> <mi>α</mi> <mn>1</mn> </msub> </semantics></math> and <math display="inline"><semantics> <msub> <mi>α</mi> <mn>2</mn> </msub> </semantics></math>.</p>
Full article ">Figure 8
<p>Comparison of system overhead of different algorithms.</p>
Full article ">Figure 9
<p>Comparison of task completion time for different UAV mission sizes.</p>
Full article ">Figure 10
<p>System consumption comparison as the number of UAVs increases.</p>
Full article ">
23 pages, 3917 KiB  
Article
MRP-YOLO: An Improved YOLOv8 Algorithm for Steel Surface Defects
by Shuxian Zhu and Yajie Zhou
Machines 2024, 12(12), 917; https://doi.org/10.3390/machines12120917 - 14 Dec 2024
Viewed by 527
Abstract
The existing detection algorithms are unable to achieve a suitable balance between detection accuracy and inference speed. As the accuracy of the algorithm increases, its complexity also rises, resulting in a decrease in detection speed, which undermines its practicality. This issue is particularly [...] Read more.
The existing detection algorithms are unable to achieve a suitable balance between detection accuracy and inference speed. As the accuracy of the algorithm increases, its complexity also rises, resulting in a decrease in detection speed, which undermines its practicality. This issue is particularly evident in the context of surface defect detection in industrial parts, where low contrast, small target features, difficult feature extraction, and low real-time detection efficiency are prominent challenges. This study proposes a novel method for steel defect detection based on the YOLO v8 algorithm, which improves detection accuracy while maintaining low computational complexity. Firstly, the global background and edge information are adaptively extracted via the MSA-SPPF module in order to obtain a more comprehensive feature representation. Furthermore, the anti-interference ability of the model is enhanced through the deformability of attention and the large convolution kernel characteristics. Secondly, the design of Dynamic Conv and C2f-OREPA enables the model to efficiently reduce the demand for computational resources while maintaining high performance. It is further proposed that the RepHead detection head approximates the multi-branch structure of the original training by a single convolution operation. This approach not only enriches the feature representation but also maintains an efficient inference process. The effectiveness of the improved MRP-YOLO algorithm is verified using the NEU-DET industrial surface defect dataset. The experimental results demonstrate that the mAP of the MRP-YOLO algorithm reaches 75.6%, which is 2.2% higher than that of the YOLOv8n algorithm, while the FLOPs are only 2.3 G higher. It indicates that the detection accuracy is significantly improved with a limited increase in computational complexity. Full article
Show Figures

Figure 1

Figure 1
<p>MRP-YOLO network structure diagram.</p>
Full article ">Figure 2
<p>Dynamic convolution network structure diagram.</p>
Full article ">Figure 3
<p>Comparison of (<b>a</b>) a vanilla convolutional layer, (<b>b</b>) a typical re-param block, and (<b>c</b>) our online re-param block in the training phase. All of these structures are converted to the same (<b>d</b>) inference–time structure.</p>
Full article ">Figure 4
<p>Block Linearization Module structure.</p>
Full article ">Figure 5
<p>Block Squeezing Module structure.</p>
Full article ">Figure 6
<p>Op-Bottleneck module and C2f-OREPA module structure. (<b>a</b>) Op-Bottleneck module architecture; (<b>b</b>) C2f-OREPA module architecture.</p>
Full article ">Figure 7
<p>DLKA Network structure diagram. (<b>a</b>) DLKA Network structure diagram; (<b>b</b>) Deform-DW Con2D structure diagram.</p>
Full article ">Figure 8
<p>MSA-SPPF Network structure diagram.</p>
Full article ">Figure 9
<p>Multiple-branch-assisted training and the inference stage is transformed into a serial structure. (<b>a</b>) Training; (<b>b</b>) Inference.</p>
Full article ">Figure 10
<p>RepHead Structure.</p>
Full article ">Figure 11
<p>Comparison of the heat maps of the improved algorithms. (<b>a</b>) Image after tagging; (<b>b</b>) Heatmap generated with the YOLOv8n model; (<b>c</b>) Heatmap generated with B model; (<b>d</b>) Heatmap generated with D model; (<b>e</b>) Heatmap generated with E model.</p>
Full article ">Figure 12
<p>Comparison of detection performance between the MRP-YOLO algorithm and other algorithms. (<b>a</b>) mAP at IoU = 0.5; (<b>b</b>) mAP for IoU Range 0.5–0.95.</p>
Full article ">Figure 13
<p>Comparison of training loss between the MRP-YOLO algorithm and other algorithms. (<b>a</b>) Training Box Loss; (<b>b</b>)Training Classification Loss; (<b>c</b>)Training DFL Loss.</p>
Full article ">Figure 14
<p>Comparison of detection performance between the MRP YOLO algorithm and the YOLOv8n algorithm, the defect types for each of the three images are Inclusion, Rolled-in scale, and Scratches. (<b>a</b>) Original image; (<b>b</b>) Image after tagging; (<b>c</b>) YOLOv8n detected results; (<b>d</b>) MRP-YOLO detected results.</p>
Full article ">
23 pages, 1529 KiB  
Article
Robust Task Offloading and Trajectory Optimization for UAV-Mounted Mobile Edge Computing
by Runhe Wang, Yang Huang, Yiwei Lu, Pu Xie and Qihui Wu
Drones 2024, 8(12), 757; https://doi.org/10.3390/drones8120757 - 14 Dec 2024
Viewed by 228
Abstract
Mobile edge computing (MEC) deployed in unmanned aerial vehicles (UAVs) has shown special strength by enhancing computational capacity and prolonging the battery lives of terrestrial user equipment (UE). Nevertheless, current research lacks studies of robust offloading scheme scheduling and trajectory planning using terrestrial [...] Read more.
Mobile edge computing (MEC) deployed in unmanned aerial vehicles (UAVs) has shown special strength by enhancing computational capacity and prolonging the battery lives of terrestrial user equipment (UE). Nevertheless, current research lacks studies of robust offloading scheme scheduling and trajectory planning using terrestrial random channels. The state-of-the-art joint task-offloading and trajectory-planning optimization techniques for UAV-mounted MEC are focused on scenarios where only air–ground channels exist rather than time-varying terrestrial channels. By contrast, this paper considers the scenario where both the time-varying/random terrestrial channels and the line-of-sight air–ground channels occur. Aiming at robust resource scheduling for energy-efficient UAV-assisted MEC, we formulate a novel joint optimization of UAV trajectory planning and task offloading, which, however, is highly nonconvex. As a countermeasure, the original optimization is recast as subproblems related to task offloading and trajectory planning and solved by a novel robust iterative optimization algorithm that combines the methods of weighted minimum mean square error, S-procedure, successive convex approximation, etc. Numerical results indicate that, compared to various baselines, the proposed algorithm can effectively reduce energy consumption and optimize the trajectory in the presence of a large number of input tasks. In addition, in terms of stability and effectiveness, the proposed robust iterative optimization algorithm can reduce energy consumption more stably in time-varying/random channels compared to non-robust schemes. Full article
Show Figures

Figure 1

Figure 1
<p>An illustration of the UAV-assisted MEC system, in which the UAV serves as an edge computing server or a relay for further offloading tasks to the base station.</p>
Full article ">Figure 2
<p>The time frame structure of the MEC.</p>
Full article ">Figure 3
<p>The trajectories of the six scenarios after algorithm convergence, and each of scenarios 2 to 6 has only a single parameter change on the baseline Scenario 1. (<b>a</b>) Scenario 1: <math display="inline"><semantics> <mrow> <mrow> <mo>{</mo> <msub> <mi mathvariant="bold">u</mi> <mn>1</mn> </msub> <mo>,</mo> <msub> <mi mathvariant="bold">u</mi> <mn>2</mn> </msub> <mo>,</mo> <msub> <mi mathvariant="bold">u</mi> <mn>3</mn> </msub> <mo>,</mo> <msub> <mi mathvariant="bold">u</mi> <mn>4</mn> </msub> <mo>}</mo> </mrow> <mo>=</mo> <mrow> <mo>{</mo> <mrow> <mo>[</mo> <mo>−</mo> <mn>100</mn> <mo>,</mo> <mo>−</mo> <mn>100</mn> <mo>]</mo> </mrow> <mo>,</mo> <mrow> <mo>[</mo> <mn>100</mn> <mo>,</mo> <mo>−</mo> <mn>100</mn> <mo>]</mo> </mrow> <mo>,</mo> <mrow> <mo>[</mo> <mn>100</mn> <mo>,</mo> <mn>100</mn> <mo>]</mo> </mrow> <mo>,</mo> <mrow> <mo>[</mo> <mo>−</mo> <mn>100</mn> <mo>,</mo> <mn>100</mn> <mo>]</mo> </mrow> <mo>}</mo> </mrow> </mrow> </semantics></math> m, <math display="inline"><semantics> <mrow> <mrow> <mo>[</mo> <msub> <mi>I</mi> <mn>1</mn> </msub> <mo>,</mo> <msub> <mi>I</mi> <mn>2</mn> </msub> <mo>,</mo> <msub> <mi>I</mi> <mn>3</mn> </msub> <mo>,</mo> <msub> <mi>I</mi> <mn>4</mn> </msub> <mo>]</mo> </mrow> <mo>=</mo> <mrow> <mo>[</mo> <mn>4</mn> <mo>,</mo> <mn>4</mn> <mo>,</mo> <mn>4</mn> <mo>,</mo> <mn>4</mn> <mo>]</mo> </mrow> </mrow> </semantics></math> Mbits, <math display="inline"><semantics> <mrow> <mi>H</mi> <mo>=</mo> <mn>30</mn> </mrow> </semantics></math> m, <math display="inline"><semantics> <mrow> <mi>r</mi> <mo>=</mo> <mn>55</mn> </mrow> </semantics></math> m, <math display="inline"><semantics> <mrow> <mi>T</mi> <mo>=</mo> <mn>25</mn> </mrow> </semantics></math> s. (<b>b</b>) Scenario 2: <math display="inline"><semantics> <mrow> <mrow> <mo>{</mo> <msub> <mi mathvariant="bold">u</mi> <mn>1</mn> </msub> <mo>,</mo> <msub> <mi mathvariant="bold">u</mi> <mn>2</mn> </msub> <mo>,</mo> <msub> <mi mathvariant="bold">u</mi> <mn>3</mn> </msub> <mo>,</mo> <msub> <mi mathvariant="bold">u</mi> <mn>4</mn> </msub> <mo>}</mo> </mrow> <mo>=</mo> <mrow> <mo>{</mo> <mrow> <mo>[</mo> <mo>−</mo> <mn>75</mn> <mo>,</mo> <mo>−</mo> <mn>75</mn> <mo>]</mo> </mrow> <mo>,</mo> <mrow> <mo>[</mo> <mn>100</mn> <mo>,</mo> <mo>−</mo> <mn>100</mn> <mo>]</mo> </mrow> <mo>,</mo> <mrow> <mo>[</mo> <mn>100</mn> <mo>,</mo> <mn>100</mn> <mo>]</mo> </mrow> <mo>,</mo> <mrow> <mo>[</mo> <mo>−</mo> <mn>50</mn> <mo>,</mo> <mn>100</mn> <mo>]</mo> </mrow> <mo>}</mo> </mrow> </mrow> </semantics></math> m. (<b>c</b>) Scenario 3: <math display="inline"><semantics> <mrow> <mrow> <mo>[</mo> <msub> <mi>I</mi> <mn>1</mn> </msub> <mo>,</mo> <msub> <mi>I</mi> <mn>2</mn> </msub> <mo>,</mo> <msub> <mi>I</mi> <mn>3</mn> </msub> <mo>,</mo> <msub> <mi>I</mi> <mn>4</mn> </msub> <mo>]</mo> </mrow> <mo>=</mo> <mrow> <mo>[</mo> <mn>2</mn> <mo>,</mo> <mn>4</mn> <mo>,</mo> <mn>3</mn> <mo>,</mo> <mn>4</mn> <mo>]</mo> </mrow> </mrow> </semantics></math> Mbits. (<b>d</b>) Scenario 4: <math display="inline"><semantics> <mrow> <mi>H</mi> <mo>=</mo> <mn>40</mn> </mrow> </semantics></math> m. (<b>e</b>) Scenario 4: <math display="inline"><semantics> <mrow> <mi>r</mi> <mo>=</mo> <mn>65</mn> </mrow> </semantics></math> m. (<b>f</b>) Scenario 6: <math display="inline"><semantics> <mrow> <mi>T</mi> <mo>=</mo> <mn>35</mn> </mrow> </semantics></math>.</p>
Full article ">Figure 4
<p>The weighted sum energy consumption of the UAV and the UE as a function of iteration indexes. Scenarios 1 to 6 are detailed in <a href="#drones-08-00757-f003" class="html-fig">Figure 3</a>.</p>
Full article ">Figure 5
<p>The offloading task <math display="inline"><semantics> <msup> <mi>l</mi> <mi mathvariant="normal">b</mi> </msup> </semantics></math>, <math display="inline"><semantics> <msup> <mi>l</mi> <mi mathvariant="normal">u</mi> </msup> </semantics></math>, <math display="inline"><semantics> <msup> <mi>l</mi> <mi>ub</mi> </msup> </semantics></math> and local computing task <math display="inline"><semantics> <msub> <mi>L</mi> <mi>k</mi> </msub> </semantics></math>.<math display="inline"><semantics> <msubsup> <mi>L</mi> <mrow> <mi>k</mi> </mrow> <mi>U</mi> </msubsup> </semantics></math> versus time.</p>
Full article ">Figure 6
<p>The weighted sum energy consumption as a function of the size of the total task of the MEC system.</p>
Full article ">Figure 7
<p>The weighted sum energy consumption as a function of the period of the MEC system.</p>
Full article ">Figure 8
<p>The weighted sum energy consumption as a function of the height of the UAV.</p>
Full article ">Figure 9
<p>The comparison between the proposed algorithm and the non-robust scheme from different perspectives. (<b>a</b>) Average energy consumption vs. simulation times. (<b>b</b>) Cumulative distribution function vs. energy consumption.</p>
Full article ">Figure 10
<p>The comparison between the proposed algorithm with different <math display="inline"><semantics> <mi>ε</mi> </semantics></math>. (<b>a</b>) Average energy consumption vs. simulation times. (<b>b</b>) Cumulative distribution function vs. energy consumption.</p>
Full article ">
13 pages, 7944 KiB  
Article
Research on Intelligent Identification Method for Pantograph Positioning and Skateboard Structural Anomalies Based on Improved YOLO v8 Algorithm
by Ruihong Zhou, Baokang Xiang, Long Wu, Yanli Hu, Litong Dou and Kaifeng Huang
Algorithms 2024, 17(12), 574; https://doi.org/10.3390/a17120574 - 14 Dec 2024
Viewed by 294
Abstract
The abnormal structural state of the pantograph skateboard is a significant and highly concerning issue that has a significant impact on the safety of high-speed railway operation. In order to obtain real-time information on the abnormal state of the skateboard in advance, an [...] Read more.
The abnormal structural state of the pantograph skateboard is a significant and highly concerning issue that has a significant impact on the safety of high-speed railway operation. In order to obtain real-time information on the abnormal state of the skateboard in advance, an intelligent defect identification model suitable to be used as a monitoring device for the pantograph skateboard was designed using a computer vision-based intelligent detection technology for pantograph skateboard defects, combined with an improved YOLO v8 model and traditional image processing algorithms such as edge extraction. The results show that the anomaly detection algorithm for the pantograph sliding plate structure has good robustness, maintaining recognition accuracy of 90% or above in complex scenes, and the average runtime is 12.32 ms. Railway field experiments have proven that the intelligent recognition model meets the actual detection requirements of railway sites and has strong practical application value. Full article
Show Figures

Figure 1

Figure 1
<p>Network structure diagram of YOLO v8.</p>
Full article ">Figure 2
<p>YOLO algorithm performance comparison.</p>
Full article ">Figure 3
<p>Verification result chart.</p>
Full article ">Figure 4
<p>Model processing statistical chart.</p>
Full article ">Figure 5
<p>Pantograph structural anomaly detection flow chart.</p>
Full article ">Figure 6
<p>Schematic diagram of the pantograph image extraction process.</p>
Full article ">Figure 7
<p>Heat map for anomaly detection.</p>
Full article ">Figure 8
<p>Pantograph structure abnormality judgment.</p>
Full article ">Figure 9
<p>Identification time-consumption curve.</p>
Full article ">
16 pages, 1308 KiB  
Article
Evaluating DL Model Scaling Trade-Offs During Inference via an Empirical Benchmark Analysis
by Demetris Trihinas, Panagiotis Michael and Moysis Symeonides
Future Internet 2024, 16(12), 468; https://doi.org/10.3390/fi16120468 - 13 Dec 2024
Viewed by 493
Abstract
With generative Artificial Intelligence (AI) capturing public attention, the appetite of the technology sector for larger and more complex Deep Learning (DL) models is continuously growing. Traditionally, the focus in DL model development has been on scaling the neural network’s foundational structure to [...] Read more.
With generative Artificial Intelligence (AI) capturing public attention, the appetite of the technology sector for larger and more complex Deep Learning (DL) models is continuously growing. Traditionally, the focus in DL model development has been on scaling the neural network’s foundational structure to increase computational complexity and enhance the representational expressiveness of the model. However, with recent advancements in edge computing and 5G networks, DL models are now aggressively being deployed and utilized across the cloud–edge–IoT continuum for the realization of in situ intelligent IoT services. This paradigm shift introduces a growing need for AI practitioners, as a focus on inference costs, including latency, computational overhead, and energy efficiency, is long overdue. This work presents a benchmarking framework designed to assess DL model scaling across three key performance axes during model inference: classification accuracy, computational overhead, and latency. The framework’s utility is demonstrated through an empirical study involving various model structures and variants, as well as publicly available datasets for three popular DL use cases covering natural language understanding, object detection, and regression analysis. Full article
Show Figures

Figure 1

Figure 1
<p>High-level overview of a deep neural network.</p>
Full article ">Figure 2
<p>Pipeline of performance evaluation trade-offs.</p>
Full article ">Figure 3
<p>Inference quality (classification accuracy and MSE) with respect to model complexity. The presented plots include: (<b>a</b>) BERT model variants, (<b>b</b>) EfficientNet model variants, and (<b>c</b>) MLP-Regression model variants.</p>
Full article ">Figure 4
<p>Computational overhead of inference with respect to model complexity. The presented plots include: (<b>a</b>) BERT model variants, (<b>b</b>) EfficientNet model variants, and (<b>c</b>) MLP-Regression model variants.</p>
Full article ">Figure 5
<p>Inference latency with respect to model complexity. The presented plots include: (<b>a</b>) BERT model variants, (<b>b</b>) EfficientNet model variants, and (<b>c</b>) MLP-Regression model variants.</p>
Full article ">
20 pages, 7528 KiB  
Article
Data Integrity Verification for Edge Computing Environments
by Jun Ye and Yu Jiang
Symmetry 2024, 16(12), 1648; https://doi.org/10.3390/sym16121648 - 13 Dec 2024
Viewed by 275
Abstract
Mobile edge computing (MEC), is a technology that extends the power of cloud computing to the edge of the end device, with the primary goal of providing faster response times and an optimized quality of service. Given that edge devices are often used [...] Read more.
Mobile edge computing (MEC), is a technology that extends the power of cloud computing to the edge of the end device, with the primary goal of providing faster response times and an optimized quality of service. Given that edge devices are often used by smaller organizations with less computing power, data on the edge are more susceptible to data corruption, which is closely related to uneven resource allocation and uneven security protection. It is therefore particularly important to check the integrity of the MEC to ensure that it is intact, which in turn serves the symmetry of data security protection and facilitates the realization and optimization of symmetry in resource allocation. An integrity verification protocol MEC-P is proposed. MEC-P allows a third-party auditor (TPA) to check data integrity on the edge without violating users’ data privacy and query pattern privacy. We rigorously demonstrate the security and privacy guarantees of the protocol. In addition, this protocol carefully considers the scenarios of single edge node and multiple edge nodes, as well as the complexity of dynamic modifications. Both theoretical analysis and experimental results demonstrate that the proposed protocol is effective. Full article
(This article belongs to the Section Computer)
Show Figures

Figure 1

Figure 1
<p>Summary of main notations.</p>
Full article ">Figure 2
<p>Pedersen commitment process based on elliptic curves.</p>
Full article ">Figure 3
<p>System model.</p>
Full article ">Figure 4
<p>Basic audit process.</p>
Full article ">Figure 5
<p>(<b>a</b>) Improved index-linked lists. (<b>b</b>) Insertion. (<b>c</b>) Deletion. (<b>d</b>) Modification.</p>
Full article ">Figure 5 Cont.
<p>(<b>a</b>) Improved index-linked lists. (<b>b</b>) Insertion. (<b>c</b>) Deletion. (<b>d</b>) Modification.</p>
Full article ">Figure 6
<p>(<b>a</b>) TagGen computation. (<b>b</b>) TagGen computation.</p>
Full article ">Figure 7
<p>(<b>a</b>) Time costs of Request for single node. (<b>b</b>) Time costs of Request for multiple nodes.</p>
Full article ">Figure 8
<p>(<b>a</b>) Time costs of Response for single node. (<b>b</b>) Time costs of Response for multiple nodes.</p>
Full article ">Figure 9
<p>(<b>a</b>) Time costs of Verify for single node. (<b>b</b>) Time costs of Verify for multiple nodes.</p>
Full article ">Figure 10
<p>(<b>a</b>) Time costs of process for single node. (<b>b</b>) Time costs of process for multiple nodes.</p>
Full article ">
Back to TopTop