[go: up one dir, main page]
Next Issue
Volume 13, September-1
Previous Issue
Volume 13, August-1
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
5.3
2.6
Journals
Electronics
Volume 13
Issue 16
share announcement

Electronics, Volume 13, Issue 16 (August-2 2024) – 246 articles

Cover Story (view full-size image): Interest in Unmanned Aerial Vehicles (UAVs) has been increasingly growing in recent years, especially for purposes other than those for which they were initially used (civil and military purposes). The present work proposes an intelligent automatic charging system (Intelligent Charging Network) created using PC Engines Alix and an experimental drone prototype using a Raspberry Pi 3 and a Navio 2 module. At the same time, an efficient Intelligent Charging Network–drone communication system and a data transmission system are proposed, which allow images acquired by the drone to be transferred directly to the server used for data storage for their subsequent processing as well as the transmission of the flight plan from the QGroundControl application to the drone. View this paper
  • Issues are regarded as officially published after their release is announced to the table of contents alert mailing list.
  • You may sign up for e-mail alerts to receive table of contents of newly released issues.
  • PDF is the official format for papers published in both, html and pdf forms. To view the papers in pdf format, click on the "PDF Full-text" link, and use the free Adobe Reader to open them.
Order results
Result details
Section
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
24 pages, 3548 KiB  
Article
Adapting CLIP for Action Recognition via Dual Semantic Supervision and Temporal Prompt Reparameterization
by Lujuan Deng, Jieqing Tan and Fangmei Liu
Electronics 2024, 13(16), 3348; https://doi.org/10.3390/electronics13163348 - 22 Aug 2024
Viewed by 361
Abstract
The contrastive vision–language pre-trained model CLIP, driven by large-scale open-vocabulary image–text pairs, has recently demonstrated remarkable zero-shot generalization capabilities in diverse downstream image tasks, which has made numerous models dominated by the “image pre-training followed by fine-tuning” paradigm exhibit promising results on standard [...] Read more.
The contrastive vision–language pre-trained model CLIP, driven by large-scale open-vocabulary image–text pairs, has recently demonstrated remarkable zero-shot generalization capabilities in diverse downstream image tasks, which has made numerous models dominated by the “image pre-training followed by fine-tuning” paradigm exhibit promising results on standard video benchmarks. However, as models scale up, full fine-tuning adaptive strategy for specific tasks becomes difficult in terms of training and storage. In this work, we propose a novel method that adapts CLIP to the video domain for efficient recognition without destroying the original pre-trained parameters. Specifically, we introduce temporal prompts to realize the object of reasoning about the dynamic content of videos for pre-trained models that lack temporal cues. Then, by replacing the direct learning style of prompt vectors with a lightweight reparameterization encoder, the model can be adapted to domain-specific adjustment to learn more generalizable representations. Furthermore, we predefine a Chinese label dictionary to enhance video representation by co-supervision of Chinese and English semantics. Extensive experiments on video action recognition benchmarks show that our method achieves competitive or even better performance than most existing methods with fewer trainable parameters in both general and few-shot recognition scenarios. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>General workflow diagram for VLMs based on contrastive learning.</p> Full article ">Figure 2
<p>Overview of the overall framework. Our proposed method contains three branches: video encoder, Chinese text encoder, and CLIP text encoder.</p> Full article ">Figure 3
<p>An illustration of temporal prompts. The final temporal prompts consist of global prompts, interchange prompts, and CLS tokens mapped with the attention module.</p> Full article ">Figure 4
<p>(<b>a</b>) Shows the structural details of the reparameterization encoder <math display="inline"><semantics> <mrow> <mi>β</mi> <mrow> <mo>(</mo> <mo>⋅</mo> <mo>)</mo> </mrow> </mrow> </semantics></math>; (<b>b</b>) illustrates the detailed information on ST-Block in <math display="inline"><semantics> <mrow> <mi>β</mi> <mrow> <mo>(</mo> <mo>⋅</mo> <mo>)</mo> </mrow> </mrow> </semantics></math>.</p> Full article ">Figure 5
<p>Ablation research using ViT-B/16 as the backbone on the HMDB-51 and UCF-101 datasets. The experiments in this chapter all use single-view testing. (<b>a</b>) The effectiveness of our proposed key components is proved step by step; (<b>b</b>) illustrates the impact of the number of sampling frames on the model; (<b>c</b>) shows the performance of learnable global prompts with different lengths. Note: TP is the abbreviation for Temporal Prompts; CN.Branch is the abbreviation for Chinese Label Text Encoder Branch.</p> Full article ">Figure 6
<p>Ablation study of training efficiency. (<b>a</b>) Parameter efficiency. Comparison of the number of trainable parameters between our method and Vita, which also freezes the visual backbone, as well as BIKE and XCLIP, which fine-tune the visual backbone; (<b>b</b>) training time efficiency. Our model still performs well with fewer training epochs. Note: the asterisk in Figure (<b>a</b>) only represents our method and has no special meaning, while the circle represents other methods.</p> Full article ">Figure 7
<p>We illustrate some behavioral actions such as “Clap”, “Cartwheel”, and “Fencing” on the original video frames and the attention map without and with our proposed method. It can be observed that our method focuses on distinguishable motion information and some key regions. Note: red indicates the areas that the model focuses on, while green mainly represents the background or some less important areas.</p> Full article ">Figure 8
<p>Performance comparison when data are extremely scarce.</p> Full article ">
0 pages, 15090 KiB  
Article
Multi-Agent Collaborative Path Planning Algorithm with Multiple Meeting Points
by Jianlin Mao, Zhigang He, Dayan Li, Ruiqi Li, Shufan Zhang and Niya Wang
Electronics 2024, 13(16), 3347; https://doi.org/10.3390/electronics13163347 (registering DOI) - 22 Aug 2024
Viewed by 425
Abstract
Traditional multi-agent path planning algorithms often lead to path overlap and excessive energy consumption when dealing with cooperative tasks due to the single-agent-single-task configuration. For this reason, the “many-to-one” cooperative planning method has been proposed, which, although improved, still faces challenges in the [...] Read more.
Traditional multi-agent path planning algorithms often lead to path overlap and excessive energy consumption when dealing with cooperative tasks due to the single-agent-single-task configuration. For this reason, the “many-to-one” cooperative planning method has been proposed, which, although improved, still faces challenges in the vast search space for meeting points and unreasonable task handover locations. This paper proposes the Cooperative Dynamic Priority Safe Interval Path Planning with a multi-meeting-point and single-meeting-point solving mode switching (Co-DPSIPPms) algorithm to achieve multi-agent path planning with task handovers at multiple or single meeting points. First, the initial priority is set based on the positional relationships among agents within the cooperative group, and the improved Fermat point method is used to locate multiple meeting points quickly. Second, considering that agents must pick up sub-tasks or conduct task handovers midway, a segmented path planning strategy is proposed to ensure that cooperative agents can efficiently and accurately complete task handovers. Finally, an automatic switching strategy between multi-meeting-point and single-meeting-point solving modes is designed to ensure the algorithm’s success rate. Tests show that Co-DPSIPPms outperforms existing algorithms in 1-to-1 and m-to-1 cooperative tasks, demonstrating its efficiency and practicality. Full article
(This article belongs to the Special Issue Path Planning for Mobile Robots, 2nd Edition)
►▼ Show Figures

Figure 1

Figure 1
<p>Principle of the SIPP algorithm.</p> Full article ">Figure 2
<p>Principle of the CBS algorithm.</p> Full article ">Figure 3
<p>Schematic diagram of the MAPF problem.</p> Full article ">Figure 4
<p>Collaborative form of single-meeting-point (Co-<math display="inline"> <semantics> <mrow> <msub> <mrow> <mi mathvariant="normal">D</mi> <mi mathvariant="normal">P</mi> <mi mathvariant="normal">S</mi> <mi mathvariant="normal">I</mi> <mi mathvariant="normal">P</mi> <mi mathvariant="normal">P</mi> </mrow> <mrow> <mi mathvariant="normal">s</mi> </mrow> </msub> </mrow> </semantics> </math>).</p> Full article ">Figure 5
<p>Collaborative form of multi-meeting-point (Co-<math display="inline"> <semantics> <mrow> <msub> <mrow> <mi mathvariant="normal">D</mi> <mi mathvariant="normal">P</mi> <mi mathvariant="normal">S</mi> <mi mathvariant="normal">I</mi> <mi mathvariant="normal">P</mi> <mi mathvariant="normal">P</mi> </mrow> <mrow> <mi mathvariant="normal">m</mi> </mrow> </msub> </mrow> </semantics> </math>).</p> Full article ">Figure 6
<p>Improved Fermat point method for obtaining single or specific subgroup meeting points.</p> Full article ">Figure 7
<p>Schematic diagram of solving multiple meeting points within a collaborative group.</p> Full article ">Figure 8
<p>Schematic diagram of the meeting points of subgroups solved within the segment (Squares of the same color indicate the three coordinate points that need to be involved in solving a confluence mi).</p> Full article ">Figure 9
<p>Comparison of task handover within a subgroup at one or two merging locations (Lines of the same color with arrows indicate the positional points involved in solving for the corresponding confluences).</p> Full article ">Figure 10
<p>Schematic diagram of segmented path planning for a collaborative task handover agent.</p> Full article ">Figure 11
<p>Schematic diagram of switching between multi-meeting-point and single-meeting-point solving modes.</p> Full article ">Figure 12
<p>Flow chart of the Co-<math display="inline"> <semantics> <mrow> <msub> <mrow> <mi mathvariant="normal">D</mi> <mi mathvariant="normal">P</mi> <mi mathvariant="normal">S</mi> <mi mathvariant="normal">I</mi> <mi mathvariant="normal">P</mi> <mi mathvariant="normal">P</mi> </mrow> <mrow> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">s</mi> </mrow> </msub> </mrow> </semantics> </math> algorithm.</p> Full article ">Figure 13
<p>Nest, Empty, Maze, and Warehouse benchmark maps.</p> Full article ">Figure 13 Cont.
<p>Nest, Empty, Maze, and Warehouse benchmark maps.</p> Full article ">Figure 14
<p>Path diagram of a single group collaborative task (<math display="inline"> <semantics> <mrow> <mi>n</mi> </mrow> </semantics> </math> = 10).</p> Full article ">Figure 15
<p>The success rates of the Co-CBS and Co-<math display="inline"> <semantics> <mrow> <msub> <mrow> <mi mathvariant="normal">D</mi> <mi mathvariant="normal">P</mi> <mi mathvariant="normal">S</mi> <mi mathvariant="normal">I</mi> <mi mathvariant="normal">P</mi> <mi mathvariant="normal">P</mi> </mrow> <mrow> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">s</mi> </mrow> </msub> </mrow> </semantics> </math> algorithms in the 1-to-1 collaboration.</p> Full article ">Figure 16
<p>Success rates of three algorithms in the 9-to-1 collaboration.</p> Full article ">Figure 17
<p>The impact of gradually increasing the number of single-group collaborative agents in the empty map on the Co-<math display="inline"> <semantics> <mrow> <msub> <mrow> <mi mathvariant="normal">D</mi> <mi mathvariant="normal">P</mi> <mi mathvariant="normal">S</mi> <mi mathvariant="normal">I</mi> <mi mathvariant="normal">P</mi> <mi mathvariant="normal">P</mi> </mrow> <mrow> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">s</mi> </mrow> </msub> </mrow> </semantics> </math> and Co-<math display="inline"> <semantics> <mrow> <msub> <mrow> <mi mathvariant="normal">D</mi> <mi mathvariant="normal">P</mi> <mi mathvariant="normal">S</mi> <mi mathvariant="normal">I</mi> <mi mathvariant="normal">P</mi> <mi mathvariant="normal">P</mi> </mrow> <mrow> <mi mathvariant="normal">s</mi> </mrow> </msub> </mrow> </semantics> </math> algorithms.</p> Full article ">Figure 18
<p>CoCube Robot.</p> Full article ">
23 pages, 1445 KiB  
Article
Dynamic Edge-Based High-Dimensional Data Aggregation with Differential Privacy
by Qian Chen, Zhiwei Ni, Xuhui Zhu, Moli Lyu, Wentao Liu and Pingfan Xia
Electronics 2024, 13(16), 3346; https://doi.org/10.3390/electronics13163346 - 22 Aug 2024
Viewed by 428
Abstract
Edge computing enables efficient data aggregation for services like data sharing and analysis in distributed IoT applications. However, uploading dynamic high-dimensional data to an edge server for efficient aggregation is challenging. Additionally, there is the significant risk of privacy leakage associated with direct [...] Read more.
Edge computing enables efficient data aggregation for services like data sharing and analysis in distributed IoT applications. However, uploading dynamic high-dimensional data to an edge server for efficient aggregation is challenging. Additionally, there is the significant risk of privacy leakage associated with direct such data uploading. Therefore, we propose an edge-based differential privacy data aggregation method leveraging progressive UMAP with a dynamic time window based on LSTM (EDP-PUDL). Firstly, a model of the dynamic time window based on a long short-term memory (LSTM) network was developed to divide dynamic data. Then, progressive uniform manifold approximation and projection (UMAP) with differential privacy was performed to reduce the dimension of the window data while preserving privacy. The privacy budget was determined by the data volume and the attribute’s Shapley value, adding DP noise. Finally, the privacy analysis and experimental comparisons demonstrated that EDP-PUDL ensures user privacy while achieving superior aggregation efficiency and availability compared to other algorithms used for dynamic high-dimensional data aggregation. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>IoT data aggregation based on edge computing.</p> Full article ">Figure 2
<p>The EDP-PUDL process.</p> Full article ">Figure 3
<p>The structures of LSTM cell and stacked LSTM model.</p> Full article ">Figure 4
<p>An example of UE selection.</p> Full article ">Figure 5
<p>Variance in Adult dataset.</p> Full article ">Figure 6
<p><math display="inline"><semantics> <mrow> <mi>T</mi> <mi>A</mi> <mi>V</mi> <mi>D</mi> </mrow> </semantics></math> with various privacy budgets on the Adult and TPC-E datasets.</p> Full article ">Figure 7
<p><math display="inline"><semantics> <mrow> <mi>T</mi> <mi>L</mi> <mn>2</mn> <mi>E</mi> <mi>r</mi> <mi>r</mi> <mi>o</mi> <mi>r</mi> </mrow> </semantics></math> with various privacy budgets on the ACS and NLTCS datasets.</p> Full article ">Figure 8
<p><math display="inline"><semantics> <mrow> <mi>T</mi> <mi>A</mi> <mi>V</mi> <mi>D</mi> </mrow> </semantics></math> with various initial window sizes on the Adult and TPC-E datasets.</p> Full article ">Figure 9
<p><math display="inline"><semantics> <mrow> <mi>T</mi> <mi>L</mi> <mn>2</mn> <mi>E</mi> <mi>r</mi> <mi>r</mi> <mi>o</mi> <mi>r</mi> </mrow> </semantics></math> with various initial window sizes on the ACS and NLTCS datasets.</p> Full article ">Figure 10
<p>SVM misclassification rates on the Adult and TPC-E datasets.</p> Full article ">Figure 11
<p>SVM misclassification rates on the ACS and NLTCS datasets.</p> Full article ">
15 pages, 270 KiB  
Article
Bi-Level Orthogonal Multi-Teacher Distillation
by Shuyue Gong and Weigang Wen
Electronics 2024, 13(16), 3345; https://doi.org/10.3390/electronics13163345 - 22 Aug 2024
Viewed by 397
Abstract
Multi-teacher knowledge distillation is a powerful technique that leverages diverse information sources from multiple pre-trained teachers to enhance student model performance. However, existing methods often overlook the challenge of effectively transferring knowledge to weaker student models. To address this limitation, we propose BOMD [...] Read more.
Multi-teacher knowledge distillation is a powerful technique that leverages diverse information sources from multiple pre-trained teachers to enhance student model performance. However, existing methods often overlook the challenge of effectively transferring knowledge to weaker student models. To address this limitation, we propose BOMD (Bi-level Optimization for Multi-teacher Distillation), a novel approach that combines bi-level optimization with multiple orthogonal projections. Our method employs orthogonal projections to align teacher feature representations with the student’s feature space while preserving structural properties. This alignment is further reinforced through a dedicated feature alignment loss. Additionally, we utilize bi-level optimization to learn optimal weighting factors for combining knowledge from heterogeneous teachers, treating the weights as upper-level variables and the student’s parameters as lower-level variables. Extensive experiments on multiple benchmark datasets demonstrate the effectiveness and flexibility of BOMD. Our method achieves state-of-the-art performance on the CIFAR-100 benchmark for multi-teacher knowledge distillation across diverse scenarios, consistently outperforming existing approaches. BOMD shows significant improvements for both homogeneous and heterogeneous teacher ensembles, even when distilling to compact student models. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>A schematic overview of our BOMD. During the training phase, BOMD utilizes the feature distillation (<b>left</b>) and logit distillation (<b>right</b>) between teacher–student models.</p> Full article ">
21 pages, 3115 KiB  
Article
Phishing Webpage Detection via Multi-Modal Integration of HTML DOM Graphs and URL Features Based on Graph Convolutional and Transformer Networks
by Jun-Ho Yoon, Seok-Jun Buu and Hae-Jung Kim
Electronics 2024, 13(16), 3344; https://doi.org/10.3390/electronics13163344 - 22 Aug 2024
Viewed by 672
Abstract
Detecting phishing webpages is a critical task in the field of cybersecurity, with significant implications for online safety and data protection. Traditional methods have primarily relied on analyzing URL features, which can be limited in capturing the full context of phishing attacks. In [...] Read more.
Detecting phishing webpages is a critical task in the field of cybersecurity, with significant implications for online safety and data protection. Traditional methods have primarily relied on analyzing URL features, which can be limited in capturing the full context of phishing attacks. In this study, we propose an innovative approach that integrates HTML DOM graph modeling with URL feature analysis using advanced deep learning techniques. The proposed method leverages Graph Convolutional Networks (GCNs) to model the structure of HTML DOM graphs, combined with Convolutional Neural Networks (CNNs) and Transformer Networks to capture the character and word sequence features of URLs, respectively. These multi-modal features are then integrated using a Transformer network, which is adept at selectively capturing the interdependencies and complementary relationships between different feature sets. We evaluated our approach on a real-world dataset comprising URL and HTML DOM graph data collected from 2012 to 2024. This dataset includes over 80 million nodes and edges, providing a robust foundation for testing. Our method demonstrated a significant improvement in performance, achieving a 7.03 percentage point increase in classification accuracy compared to state-of-the-art techniques. Additionally, we conducted ablation tests to further validate the effectiveness of individual features in our model. The results validate the efficacy of integrating HTML DOM structure and URL features using deep learning. Our framework significantly enhances phishing detection capabilities, providing a more accurate and comprehensive solution to identifying malicious webpages. Full article
(This article belongs to the Special Issue Network Security and Cryptography Applications)
►▼ Show Figures

Figure 1

Figure 1
<p>Visualization of key features for phishing and benign classification using three models: Select K Best, Random Forest, and LassoCV (<b>a</b>–<b>c</b>) show URL character-level analysis, (<b>d</b>–<b>f</b>) show URL word-level analysis, and (<b>g</b>–<b>i</b>) show HTML tag name analysis.</p> Full article ">Figure 2
<p>Comparison of HTML DOM structures between (<b>a</b>) a benign website and (<b>b</b>) a phishing website with similar URLs.</p> Full article ">Figure 3
<p>Overview of the proposed phishing detection framework integrating HTML DOM graphs and URL features.</p> Full article ">Figure 4
<p>The ensemble classification model combining CNN, Transformer, and GCN components for phishing detection.</p> Full article ">Figure 5
<p>t-SNE Visualization of Feature Integration for Graph-Based and Non-Graph-Based Models.</p> Full article ">Figure 6
<p>Comparative HTML DOM Graph Visualizations for Benign and Phishing Case Analysis. (<b>a</b>) DOM structure of the benign URL “<a href="https://ninecloud.ae/our-services/interior-exterior-paint-contractors-in-dubai/" target="_blank">https://ninecloud.ae/our-services/interior-exterior-paint-contractors-in-dubai/</a> (accessed on 19 August 2024),” showing a clear and simple structure, correctly classified as benign. (<b>b</b>) DOM structure of the phishing URL “<a href="https://pub-88f64e013ca94e82aa5d15393134722c.r2.dev/logs.html" target="_blank">https://pub-88f64e013ca94e82aa5d15393134722c.r2.dev/logs.html</a> (accessed on 19 August 2024),” showing a complex and suspicious structure, correctly classified as phishing. (<b>c</b>) DOM structure of the benign URL “<a href="https://rilm.am/wp-content/uploads/2022/07/12e47ac82164e89a8c15f399384-e6572.pdf" target="_blank">https://rilm.am/wp-content/uploads/2022/07/12e47ac82164e89a8c15f399384-e6572.pdf</a> (accessed on 19 August 2024),” showing a complex structure, incorrectly classified as phishing. (<b>d</b>) DOM structure of the phishing URL “<a href="https://amangroup.co/gy/linkedin_/" target="_blank">https://amangroup.co/gy/linkedin_/</a> (accessed on 19 August 2024),” showing a relatively simple structure, incorrectly classified as benign.</p> Full article ">
35 pages, 1125 KiB  
Review
Review of Smart-Home Security Using the Internet of Things
by George Vardakis, George Hatzivasilis, Eleftheria Koutsaki and Nikos Papadakis
Electronics 2024, 13(16), 3343; https://doi.org/10.3390/electronics13163343 - 22 Aug 2024
Viewed by 1696
Abstract
As the Internet of Things (IoT) continues to revolutionize the way we interact with our living spaces, the concept of smart homes has become increasingly prevalent. However, along with the convenience and connectivity offered by IoT-enabled devices in smart homes comes a range [...] Read more.
As the Internet of Things (IoT) continues to revolutionize the way we interact with our living spaces, the concept of smart homes has become increasingly prevalent. However, along with the convenience and connectivity offered by IoT-enabled devices in smart homes comes a range of security challenges. This paper explores the landscape of smart-home security. In contrast to similar surveys, this study also examines the particularities of popular categories of smart devices, like home assistants, TVs, AR/VR, locks, sensors, etc. It examines various security threats and vulnerabilities inherent in smart-home ecosystems, including unauthorized access, data breaches, and device tampering. Additionally, the paper discusses existing security mechanisms and protocols designed to mitigate these risks, such as encryption, authentication, and intrusion-detection systems. Furthermore, it highlights the importance of user awareness and education in maintaining the security of smart-home environments. Finally, the paper proposes future research directions and recommendations for enhancing smart-home security with IoT, including the development of robust security best practices and standards, improved device authentication methods, and more effective intrusion-detection techniques. By addressing these challenges, the potential of IoT-enabled smart homes to enhance convenience and efficiency while ensuring privacy, security, and cyber-resilience can be realized. Full article
(This article belongs to the Special Issue New Challenges in Information Security and Privacy and Cyber Resilience)
►▼ Show Figures

Figure 1

Figure 1
<p>Basic building blocks of IoT ecosystems.</p> Full article ">Figure 2
<p>Smart-home threats and countermeasures [<a href="#B30-electronics-13-03343" class="html-bibr">30</a>].</p> Full article ">
20 pages, 2056 KiB  
Article
A Deep Learning Approach for Fault-Tolerant Data Fusion Applied to UAV Position and Orientation Estimation
by Majd Saied, Abbas Mishi, Clovis Francis and Ziad Noun
Electronics 2024, 13(16), 3342; https://doi.org/10.3390/electronics13163342 - 22 Aug 2024
Viewed by 394
Abstract
This work introduces a novel fault-tolerance technique for data fusion in Unmanned Aerial Vehicles (UAVs), designed to address sensor faults through a deep learning-based framework. Unlike traditional methods that rely on hardware redundancy, our approach leverages Long Short-Term Memory (LSTM) networks for state [...] Read more.
This work introduces a novel fault-tolerance technique for data fusion in Unmanned Aerial Vehicles (UAVs), designed to address sensor faults through a deep learning-based framework. Unlike traditional methods that rely on hardware redundancy, our approach leverages Long Short-Term Memory (LSTM) networks for state estimation and a moving average (MA) algorithm for fault detection. The novelty of our technique lies in its dual strategy: utilizing LSTMs to analyze residuals and detect errors, while the MA algorithm identifies faulty sensors by monitoring variations in sensor data. This method allows for effective error correction and system recovery by replacing faulty measurements with reliable ones, eliminating the need for a fault-free prediction model. The approach has been validated through offline testing on real sensor data from a hexarotor UAV with simulated faults, demonstrating its efficacy in maintaining robust UAV operations without resorting to redundant hardware solutions. Full article
(This article belongs to the Special Issue Advanced Technologies in Unmanned Aerial Vehicle (UAV)-Based Communications)
►▼ Show Figures

Figure 1

Figure 1
<p>Fault-tolerant data fusion scheme using duplication–comparison.</p> Full article ">Figure 2
<p>Implementation of fault-tolerance architecture on UAV system.</p> Full article ">Figure 3
<p>LSTM neural network-based data fusion for attitude estimation.</p> Full article ">Figure 4
<p>LSTM neural network-based data fusion for position and altitude estimation.</p> Full article ">Figure 5
<p>Overview of the test environment.</p> Full article ">Figure 6
<p>Data fusion results on fault-free case: attitude states.</p> Full article ">Figure 6 Cont.
<p>Data fusion results on fault-free case: attitude states.</p> Full article ">Figure 7
<p>Data fusion results on fault-free case during low dynamic motion: attitude states.</p> Full article ">Figure 8
<p>Data fusion results on fault-free case during high dynamic motion: attitude states.</p> Full article ">Figure 9
<p>Residuals after an additive fault on Mag1 (<b>upper</b> figures for Block 1, <b>lower</b> figures for Block 2). The vertical red arrows indicate the fault injection time. The horizontal red lines represent the thresholds.</p> Full article ">Figure 10
<p>Attitude moving averages after an additive fault on Mag1.The horizontal red lines represent the thresholds.</p> Full article ">Figure 11
<p>The residual tests after Mag1 fault. The horizontal red lines represent the thresholds.</p> Full article ">Figure 12
<p>The outputs of the fusion blocks after Gyro2 fault.The vertical red arrows indicate the fault injection time.</p> Full article ">Figure 13
<p>Effects of sensor fault types on residuals. The horizontal red lines represent the thresholds.</p> Full article ">
34 pages, 2908 KiB  
Article
A Hybrid Contrast and Texture Masking Model to Boost High Efficiency Video Coding Perceptual Rate-Distortion Performance
by Javier Ruiz Atencia, Otoniel López-Granado, Manuel Pérez Malumbres, Miguel Martínez-Rach, Damian Ruiz Coll, Gerardo Fernández Escribano and Glenn Van Wallendael
Electronics 2024, 13(16), 3341; https://doi.org/10.3390/electronics13163341 - 22 Aug 2024
Viewed by 360
Abstract
As most of the videos are destined for human perception, many techniques have been designed to improve video coding based on how the human visual system perceives video quality. In this paper, we propose the use of two perceptual coding techniques, namely contrast [...] Read more.
As most of the videos are destined for human perception, many techniques have been designed to improve video coding based on how the human visual system perceives video quality. In this paper, we propose the use of two perceptual coding techniques, namely contrast masking and texture masking, jointly operating under the High Efficiency Video Coding (HEVC) standard. These techniques aim to improve the subjective quality of the reconstructed video at the same bit rate. For contrast masking, we propose the use of a dedicated weighting matrix for each block size (from 4×4 up to 32×32), unlike the HEVC standard, which only defines an 8×8 weighting matrix which it is upscaled to build the 16×16 and 32×32 weighting matrices (a 4×4 weighting matrix is not supported). Our approach achieves average Bjøntegaard Delta-Rate (BD-rate) gains of between 2.5% and 4.48%, depending on the perceptual metric and coding mode used. On the other hand, we propose a novel texture masking scheme based on the classification of each coding unit to provide an over-quantization depending on the coding unit texture level. Thus, for each coding unit, its mean directional variance features are computed to feed a support vector machine model that properly predicts the texture type (plane, edge, or texture). According to this classification, the block’s energy, the type of coding unit, and its size, an over-quantization value is computed as a QP offset (DQP) to be applied to this coding unit. By applying both techniques in the HEVC reference software, an overall average of 5.79% BD-rate gain is achieved proving their complementarity. Full article
(This article belongs to the Special Issue Recent Advances in Image/Video Compression and Coding)
►▼ Show Figures

Figure 1

Figure 1
<p>Default HEVC quantization weighting matrices.</p> Full article ">Figure 2
<p>Contrast sensitivity function. The red curve represents the original CSF as defined by Equation (1), while the blue dashed curve represents the flattened CSF, with spatial frequencies below the peak sensitivity saturated.</p> Full article ">Figure 3
<p>Proposed 4 × 4 quantization weighting matrices for intra- and interprediction modes.</p> Full article ">Figure 4
<p>Rate-distortion curves comparing our proposed CSF with the default implemented in the HEVC standard using different perceptual metrics. (<b>a</b>,<b>b</b>) correspond to the BQTerrace sequence of class B, while (<b>c</b>,<b>d</b>) correspond to the ChinaSpeed sequence of class F.</p> Full article ">Figure 5
<p>Samples of manually classified blocks (left-hand side) and their associated polar diagram of the MDV metric (right-hand side). From top to bottom: <math display="inline"><semantics> <mrow> <mn>8</mn> <mo>×</mo> <mn>8</mn> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mn>16</mn> <mo>×</mo> <mn>16</mn> </mrow> </semantics></math>, and <math display="inline"><semantics> <mrow> <mn>32</mn> <mo>×</mo> <mn>32</mn> </mrow> </semantics></math> block sizes; from left- to right-hand side: plain, edge, and texture blocks.</p> Full article ">Figure 6
<p>(<b>a</b>) Scatter plot of manually classified <math display="inline"><semantics> <mrow> <mn>16</mn> <mo>×</mo> <mn>16</mn> </mrow> </semantics></math> blocks (training dataset), and (<b>b</b>) the classification results provided by the trained SVM model (testing dataset).</p> Full article ">Figure 7
<p>Example of block classification for the first frame of sequence BasketballDrill, using optimal SVM models for each block size.</p> Full article ">Figure 8
<p>Box and whisker plot of the block energy (<math display="inline"><semantics> <mi>ε</mi> </semantics></math>) distribution by size and texture classification.</p> Full article ">Figure 9
<p>Representation of Equation (6) for two sets of function parameter, (<b>red</b>) <math display="inline"><semantics> <mrow> <mi>M</mi> <mi>i</mi> <mi>n</mi> <msub> <mi>E</mi> <mn>1</mn> </msub> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>M</mi> <mi>a</mi> <mi>x</mi> <msub> <mi>E</mi> <mn>1</mn> </msub> </mrow> </semantics></math> and <math display="inline"><semantics> <mrow> <mi>M</mi> <mi>a</mi> <mi>x</mi> <mi>Q</mi> <mi>S</mi> <mi>t</mi> <mi>e</mi> <msub> <mi>p</mi> <mn>1</mn> </msub> </mrow> </semantics></math> and (<b>blue</b>) <math display="inline"><semantics> <mrow> <mi>M</mi> <mi>i</mi> <mi>n</mi> <msub> <mi>E</mi> <mn>2</mn> </msub> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>M</mi> <mi>a</mi> <mi>x</mi> <msub> <mi>E</mi> <mn>2</mn> </msub> </mrow> </semantics></math>, and <math display="inline"><semantics> <mrow> <mi>M</mi> <mi>a</mi> <mi>x</mi> <mi>Q</mi> <mi>S</mi> <mi>t</mi> <mi>e</mi> <msub> <mi>p</mi> <mn>2</mn> </msub> </mrow> </semantics></math>. <math display="inline"><semantics> <mrow> <mo>Δ</mo> <mi>Q</mi> <mi>S</mi> <mi>t</mi> <mi>e</mi> <msub> <mi>p</mi> <mrow> <mi>i</mi> <mo>,</mo> <mi>j</mi> </mrow> </msub> </mrow> </semantics></math> is different for each set.</p> Full article ">Figure 10
<p>Flowchart of candidate selection for brute-force analysis of perceptually optimal parameters. The Ps in energy range boxes refer to the percentile.</p> Full article ">Figure 11
<p>BD-rate curves (MS-SSIM metric) for PeopleOnStreet video test sequence over the <math display="inline"><semantics> <mrow> <mi>M</mi> <mi>a</mi> <mi>x</mi> <mi>Q</mi> <mi>S</mi> <mi>t</mi> <mi>e</mi> <mi>p</mi> </mrow> </semantics></math> parameter when modifying texture blocks of size 8. Each curve represents a different block energy range (<math display="inline"><semantics> <mrow> <mi>M</mi> <mi>i</mi> <mi>n</mi> <mi>E</mi> </mrow> </semantics></math> and <math display="inline"><semantics> <mrow> <mi>M</mi> <mi>a</mi> <mi>x</mi> <mi>E</mi> </mrow> </semantics></math>).</p> Full article ">Figure 12
<p>Rate-distortion curves of the first frame of the BQSquare sequence, comparing our proposed contrast masking (red line) and contrast and texture masking (yellow line) with the HM reference coding (blue line), using the (<b>a</b>) SSIM, (<b>b</b>) MS-SSIM, and (<b>c</b>) PSNR-HVS-M perceptual metrics.</p> Full article ">Figure 13
<p>Visual comparison of the first frame of the BQSquare sequence encoded at <math display="inline"><semantics> <mrow> <mi>Q</mi> <mi>P</mi> <mo>=</mo> <mn>22</mn> </mrow> </semantics></math>. (<b>a</b>) HM reference-encoded frame; (<b>b</b>) frame encoded with contrast and texture masking.</p> Full article ">Figure 13 Cont.
<p>Visual comparison of the first frame of the BQSquare sequence encoded at <math display="inline"><semantics> <mrow> <mi>Q</mi> <mi>P</mi> <mo>=</mo> <mn>22</mn> </mrow> </semantics></math>. (<b>a</b>) HM reference-encoded frame; (<b>b</b>) frame encoded with contrast and texture masking.</p> Full article ">Figure A1
<p>Traffic 2560 × 1600 30 fps Class A.</p> Full article ">Figure A2
<p>PeopleOnStreet 2560 × 1600 30 fps Class A.</p> Full article ">Figure A3
<p>NebutaFestival 2560 × 1600 60 fps Class A.</p> Full article ">Figure A4
<p>SteamLocomotiveTrain 2560 × 1600 60 fps Class A.</p> Full article ">Figure A5
<p>Kimono 1920 × 1080 24 fps Class B.</p> Full article ">Figure A6
<p>ParkScene 1920 × 1080 24 fps Class B.</p> Full article ">Figure A7
<p>Cactus 1920 × 1080 50 fps Class B.</p> Full article ">Figure A8
<p>BQTerrace 1920 × 1080 60 fps Class B.</p> Full article ">Figure A9
<p>BasketballDrive 1920 × 1080 50 fps Class B.</p> Full article ">Figure A10
<p>RaceHorses 832 × 480 30 fps Class C.</p> Full article ">Figure A11
<p>BQMall 832 × 480 60 fps Class C.</p> Full article ">Figure A12
<p>PartyScene 832 × 480 50 fps Class C.</p> Full article ">Figure A13
<p>BasketballDrill 832 × 480 50 fps Class C.</p> Full article ">Figure A14
<p>RaceHorses 416 × 240 30 fps Class D.</p> Full article ">Figure A15
<p>BQSquare 416 × 240 60 fps Class D.</p> Full article ">Figure A16
<p>BlowingBubbles 416 × 240 50 fps Class D.</p> Full article ">Figure A17
<p>BasketballPass 416 × 240 50 fps Class D.</p> Full article ">Figure A18
<p>FourPeople 1280 × 720 60 fps Class E.</p> Full article ">Figure A19
<p>Johnny 1280 × 720 60 fps Class E.</p> Full article ">Figure A20
<p>KristenAndSara 1280 × 720 60 fps Class E.</p> Full article ">Figure A21
<p>BasketballDrillText 832 × 480 50 fps Class F.</p> Full article ">Figure A22
<p>ChinaSpeed 1024 × 768 30 fps Class F.</p> Full article ">Figure A23
<p>SlideEditing 1280 × 720 30 fps Class F.</p> Full article ">Figure A24
<p>SlideShow 1280 × 720 20 fps Class F.</p> Full article ">
22 pages, 7018 KiB  
Article
PAN: Improved PointNet++ for Pavement Crack Information Extraction
by Jiakai Fan, Weidong Song, Jinhe Zhang, Shangyu Sun, Guohui Jia and Guang Jin
Electronics 2024, 13(16), 3340; https://doi.org/10.3390/electronics13163340 - 22 Aug 2024
Viewed by 355
Abstract
Maintenance and repair of expressways are becoming increasingly important due to the growing frequency of their use. Accurate pavement crack information extraction helps with routine maintenance and reduces the risk of traffic accidents. The traditional 2D crack image detection method has limitations and [...] Read more.
Maintenance and repair of expressways are becoming increasingly important due to the growing frequency of their use. Accurate pavement crack information extraction helps with routine maintenance and reduces the risk of traffic accidents. The traditional 2D crack image detection method has limitations and cannot effectively obtain depth information. Three-dimensional crack extraction from 3D point cloud has become a new solution that can capture pavement crack information more comprehensively and accurately. However, the existing algorithms are not effective in the feature extraction of cracks due to the different and irregular shapes and sizes of pavement cracks and interference from the external environment. To solve this, a new method for detecting pavement cracks in point clouds, namely point attention net (PAN), is herein proposed. It uses a two-branch attention fusion module to focus on space and feature information in the cloud and capture features of crack points at different scales. It also uses the Poly Loss function to solve the imbalance of foreground and background points in pavement point cloud data. Experiments on the LNTU-RDD-LiDAR dataset were carried out to verify the effectiveness of the proposed method. Compared with the traditional method and the latest point cloud segmentation technology, the performance indexes of mIoU, Acc, F1, and Rec achieved significant improvement, reaching 75.4%, 91.5%, 75.4%, and 67.1%, respectively. Full article
(This article belongs to the Special Issue Fault Detection Technology Based on Deep Learning)
►▼ Show Figures

Figure 1

Figure 1
<p>PAN model structure, using PointNet++ as the baseline code and adding PC-Parallel modules in the early and late stages of the network. Adding the PC-Parallel module in the early stage can increase the correlation and encoding between local geometric features; adding PC in the later-stage PC-Parallel module can enable better feature interaction in all aspects of the object and expand the original receptive field.</p> Full article ">Figure 2
<p>PC-Parallel module structure. The two attention modules work together to improve the point cloud model’s understanding of different channel and position combinations and enhance complex structure modeling. Global structure understanding enables more accurate segmentation and better model adaptability, and dynamic weight adjustment promotes feature learning for generalization performance.</p> Full article ">Figure 3
<p>LNTU-RDD-LiDAR pavement crack point cloud dataset. (<b>a</b>) Generate a gray value image based on the z value of the point cloud. (<b>b</b>) Generate a depth value image based on the z value of the point cloud. (<b>c</b>) Corresponds to the true value of the manually labeled road crack.</p> Full article ">Figure 4
<p>Visual comparison results and corresponding detailed observations of pavement crack information extraction between PAN and PointNet, PointNet++, PointTransformer, and PointMLP methods on the LNTU-RDD-LiDAR dataset.</p> Full article ">Figure 5
<p>The PC-Parallel module on LNTU-RDD-LiDAR dataset tested the results of visual comparison and detailed observation of the ablation experimental method.</p> Full article ">Figure 6
<p>Visual comparison and detailed observation of ablation experiments with Poly Loss test on LNTU-RDD-LiDAR dataset.</p> Full article ">
33 pages, 1322 KiB  
Review
Outlier Detection in Streaming Data for Telecommunications and Industrial Applications: A Survey
by Roland N. Mfondoum, Antoni Ivanov, Pavlina Koleva, Vladimir Poulkov and Agata Manolova
Electronics 2024, 13(16), 3339; https://doi.org/10.3390/electronics13163339 - 22 Aug 2024
Viewed by 504
Abstract
Streaming data are present all around us. From traditional radio systems streaming audio to today’s connected end-user devices constantly sending information or accessing services, data are flowing constantly between nodes across various networks. The demand for appropriate outlier detection (OD) methods in the [...] Read more.
Streaming data are present all around us. From traditional radio systems streaming audio to today’s connected end-user devices constantly sending information or accessing services, data are flowing constantly between nodes across various networks. The demand for appropriate outlier detection (OD) methods in the fields of fault detection, special events detection, and malicious activities detection and prevention is not only persistent over time but increasing, especially with the recent developments in Telecommunication systems such as Fifth Generation (5G) networks facilitating the expansion of the Internet of Things (IoT). The process of selecting a computationally efficient OD method, adapted for a specific field and accounting for the existence of empirical data, or lack thereof, is non-trivial. This paper presents a thorough survey of OD methods, categorized by the applications they are implemented in, the basic assumptions that they use according to the characteristics of the streaming data, and a summary of the emerging challenges, such as the evolving structure and nature of the data and their dimensionality and temporality. A categorization of commonly used datasets in the context of streaming data is produced to aid data source identification for researchers in this field. Based on this, guidelines for OD method selection are defined, which consider flexibility and sample size requirements and facilitate the design of such algorithms in Telecommunications and other industries. Full article
(This article belongs to the Special Issue Knowledge Engineering and Data Mining, 3rd Edition)
►▼ Show Figures

Figure 1

Figure 1
<p>Literature review strategy.</p> Full article ">Figure 2
<p>Structure of this paper.</p> Full article ">
18 pages, 3413 KiB  
Review
Green Energy Management in Manufacturing Based on Demand Prediction by Artificial Intelligence—A Review
by Izabela Rojek, Dariusz Mikołajewski, Adam Mroziński and Marek Macko
Electronics 2024, 13(16), 3338; https://doi.org/10.3390/electronics13163338 - 22 Aug 2024
Viewed by 830
Abstract
Energy efficiency in production systems and processes is a key global research topic, especially in light of the Green Deal, Industry 4.0/5.0 paradigms, and rising energy prices. Research on improving the energy efficiency of production based on artificial intelligence (AI) analysis brings promising [...] Read more.
Energy efficiency in production systems and processes is a key global research topic, especially in light of the Green Deal, Industry 4.0/5.0 paradigms, and rising energy prices. Research on improving the energy efficiency of production based on artificial intelligence (AI) analysis brings promising solutions, and the digital transformation of industry towards green energy is slowly becoming a reality. New production planning rules, the optimization of the use of the Industrial Internet of Things (IIoT), industrial cyber-physical systems (ICPSs), and the effective use of production data and their optimization with AI bring further opportunities for sustainable, energy-efficient production. The aim of this study is to systematically evaluate and quantify the research results, trends, and research impact on energy management in production based on AI-based demand forecasting. The value of the research includes the broader use of AI which will reduce the impact of the observed environmental and economic problems in the areas of reducing energy consumption, forecasting accuracy, and production efficiency. In addition, the demand for Green AI technologies in creating sustainable solutions, reducing the impact of AI on the environment, and improving the accuracy of forecasts, including in the area of optimization of electricity storage, will increase. A key emerging research trend in green energy management in manufacturing is the use of AI-based demand forecasting to optimize energy consumption, reduce waste, and increase sustainability. An innovative perspective that leverages AI’s ability to accurately forecast energy demand allows manufacturers to align energy consumption with production schedules, minimizing excess energy consumption and emissions. Advanced machine learning (ML) algorithms can integrate real-time data from various sources, such as weather patterns and market demand, to improve forecast accuracy. This supports both sustainability and economic efficiency. In addition, AI-based demand forecasting can enable more dynamic and responsive energy management systems, paving the way for smarter, more resilient manufacturing processes. The paper’s contribution goes beyond mere description, making analyses, comparisons, and generalizations based on the leading current literature, logical conclusions from the state-of-the-art, and the authors’ knowledge and experience in renewable energy, AI, and mechatronics. Full article
(This article belongs to the Special Issue Advanced Industry 4.0/5.0: Intelligence and Automation)
►▼ Show Figures

Figure 1

Figure 1
<p>Bibliometric analysis procedure.</p> Full article ">Figure 2
<p>A flow chart of the review process using some points of PRISMA.</p> Full article ">Figure 3
<p>Annual publication trend (WoS).</p> Full article ">Figure 4
<p>Annual citation trend (WoS).</p> Full article ">Figure 5
<p>Categories of publications (WoS).</p> Full article ">Figure 6
<p>Countries of research (WoS).</p> Full article ">Figure 7
<p>Annual publication trend (Scopus).</p> Full article ">Figure 8
<p>Publications by type (Scopus).</p> Full article ">Figure 9
<p>Publications by area (Scopus).</p> Full article ">Figure 10
<p>Publications by country (Scopus).</p> Full article ">Figure 11
<p>Summary of the literature review (Scopus and WoS, n = 26).</p> Full article ">
20 pages, 4443 KiB  
Article
Transient Synchronous Stability Analysis of Grid-Following Converter Considering Outer-Loop Control with Current Limiting
by Leke Chen, Lin Zhu, Yang Liu, Nan Ye, Yonghao Hu and Lin Guan
Electronics 2024, 13(16), 3337; https://doi.org/10.3390/electronics13163337 - 22 Aug 2024
Viewed by 351
Abstract
With the evolution of modern power systems, inverter-based resources have become increasingly prevalent. As critical energy conversion interfaces, grid-following converters exhibit dynamic performances, presenting challenges for system security and stability. This paper focuses on the transient synchronization stability of converters after disturbances, highlighting [...] Read more.
With the evolution of modern power systems, inverter-based resources have become increasingly prevalent. As critical energy conversion interfaces, grid-following converters exhibit dynamic performances, presenting challenges for system security and stability. This paper focuses on the transient synchronization stability of converters after disturbances, highlighting differences in mechanisms compared to synchronous generators. Although previous studies on the transient synchronization stability of converters have been conducted, they primarily concentrate on the dynamics of the phase-locked loop, with limited consideration of the effects of outer-loop control. This has created a cognitive bottleneck in understanding the transient synchronization mechanisms of converters. To address these challenges, this paper models a grid-following voltage source converter system, incorporating detailed converter control strategies and current-limiting control. The stability regions of the stable equilibrium point under various fault severities are first analyzed. Then, the impacts of outer-loop control, including PI control and current-limiting control, on transient synchronization are examined. The study systematically elucidates the influence of outer-loop control on the transient synchronization stability of converters. Finally, the validity of the proposed theory is confirmed through simulations conducted in PSCAD/EMTDC. Full article
(This article belongs to the Special Issue Section Collection Series: New Horizons and Recent Advances in Power Electronics)
►▼ Show Figures

Figure 1

Figure 1
<p>Circuit and control diagram of VSC.</p> Full article ">Figure 2
<p>Outer-loop current-limiting control diagram of the q-axis prior current limiting strategy.</p> Full article ">Figure 3
<p>Change in SEPs of the second-order model before and after disturbance.</p> Full article ">Figure 4
<p>Schematic diagram of phase change.</p> Full article ">Figure 5
<p>Changes in SEP before and after disturbance. (<b>a</b>) Slight and moderate dips in grid voltage; (<b>b</b>) severe dip in grid voltage with equilibrium point; (<b>c</b>) no equilibrium point.</p> Full article ">Figure 6
<p>Phasor diagram of VSC when the grid voltage drops slightly.</p> Full article ">Figure 7
<p>Phasor diagram of VSC with a medium voltage drop.</p> Full article ">Figure 8
<p>Phasor diagram of VSC with a severe voltage drop. (<b>a</b>) Existence of equilibrium point; (<b>b</b>) absence of equilibrium point.</p> Full article ">Figure 9
<p>Time-domain simulation verification with or without considering outer-loop control. (<b>a</b>) Virtual power angle; (<b>b</b>) d-axis reference current; (<b>c</b>) q-axis reference current.</p> Full article ">Figure 10
<p>Simulation verification of SEP changes. (<b>a</b>) Scenario 1 and Scenario 2; (<b>b</b>) Scenario 3 and Scenario 4.</p> Full article ">Figure 10 Cont.
<p>Simulation verification of SEP changes. (<b>a</b>) Scenario 1 and Scenario 2; (<b>b</b>) Scenario 3 and Scenario 4.</p> Full article ">Figure 11
<p>Simulation verification of system dynamics considering outer-loop control. (<b>a</b>) Reference current; (<b>b</b>) output active power; (<b>c</b>) output reactive power.</p> Full article ">
15 pages, 4982 KiB  
Article
Research on DC Electric Shock Protection Method Based on Sliding Curvature Accumulation Quantity
by Hongzhang Zhu, Chuanping Wu, Yao Xie, Yang Zhou, Xiujin Liao and Jian Li
Electronics 2024, 13(16), 3336; https://doi.org/10.3390/electronics13163336 - 22 Aug 2024
Viewed by 400
Abstract
To address the limitations of current DC residual current protection methods, which primarily rely on the amplitude of DC residual current for fault detection and fail to safeguard against electric shocks at two points on the same side in DC Isolated Terra (IT) [...] Read more.
To address the limitations of current DC residual current protection methods, which primarily rely on the amplitude of DC residual current for fault detection and fail to safeguard against electric shocks at two points on the same side in DC Isolated Terra (IT) System systems, this paper introduces a novel protection method based on DC electric shock features. This paper first analyzes the sliding curvature accumulation and peak rise time features of DC basic residual current, load mutation current, and animal body electric shock current under multi-factor conditions. The analysis shows that sliding curvature accumulation in the range of 0.1 ≤ K ≤ 1 and a peak rise time of Δt ≥ 20 ms can effectively distinguish animal body electric shock. Then, based on this electric shock’s distinctive characteristics, an approach for identifying types of electric shock is developed. Finally, a DC residual current protective device (DC-RCD) is designed. The prototype test results demonstrate that the DC-RCD has an action time of ts < 70 ms. The proposed method accurately provides protection against electric shocks and effectively addresses the issue of inadequate protection when two fault points occur on the same side within an IT system. This approach holds significant reference value for the development of next-generation DC-RCDs. Full article
(This article belongs to the Section Power Electronics)
►▼ Show Figures

Figure 1

Figure 1
<p>Experimental platform for LVDC electric shock.</p> Full article ">Figure 2
<p>Current waveforms of animal and non-animal bodies. (<b>a</b>) Live pig electric shock, (<b>b</b>) Load burst, (<b>c</b>) Basic residual current.</p> Full article ">Figure 3
<p>Schematic diagram of a two-point fault in a DC IT grounding system. (<b>a</b>) Two-point opposite-side faults. (<b>b</b>) Two-point same-side faults.</p> Full article ">Figure 4
<p>Schematic diagram of curve curvature calculation principle.</p> Full article ">Figure 5
<p>Pig electric shock current feature. (<b>a</b>) Waveform after filtering, (<b>b</b>) Sliding curvature accumulation.</p> Full article ">Figure 6
<p>Load burst current feature. (<b>a</b>) Waveform after filtering, (<b>b</b>) Sliding curvature accumulation.</p> Full article ">Figure 7
<p>Basic residual current feature. (<b>a</b>) Waveform after filtering, (<b>b</b>) Sliding curvature accumulation.</p> Full article ">Figure 8
<p>Peak value of sliding current accumulation under different conditions. (<b>a</b>) Different voltage levels, (<b>b</b>) Different skin surface humidity levels, (<b>c</b>) Different ways of electric shock, (<b>d</b>) Different weights, (<b>e</b>) Electric shock path: right forelimb-left hindlimb and right forelimb-right hindlimb, (<b>f</b>) Electric shock path: left forelimb-left hindlimb, left forelimb-right hindlimb.</p> Full article ">Figure 9
<p>Peak value of sliding current accumulation at different currents. (<b>a</b>) Load burst, (<b>b</b>) Basic residual current.</p> Full article ">Figure 10
<p>Rise time of animal and non-animal bodies.</p> Full article ">Figure 11
<p>Algorithm for electric shock type recognition.</p> Full article ">Figure 12
<p>DC-RCD prototype design scheme. (<b>a</b>) DC-RCD system schematic design, (<b>b</b>) Physical drawing of DC-RCD prototype.</p> Full article ">Figure 13
<p>Dynamic waveform of DC voltage/fault current test under no-electric-shock ground fault condition. (<b>a</b>) The method proposed in this paper, (<b>b</b>) Amplitude identification method.</p> Full article ">Figure 14
<p>Dynamic waveform of DC voltage/electric shock current test under two-point opposite-side fault conditions. (<b>a</b>) The method proposed in this paper, (<b>b</b>) Amplitude identification method.</p> Full article ">Figure 15
<p>Dynamic waveform of DC voltage/electric shock current test under two-point same-side fault conditions. (<b>a</b>) The method proposed in this paper, (<b>b</b>) Amplitude identification method.</p> Full article ">
11 pages, 1849 KiB  
Article
Improved Segmentation of Cellular Nuclei Using UNET Architectures for Enhanced Pathology Imaging
by Simão Castro, Vitor Pereira and Rui Silva
Electronics 2024, 13(16), 3335; https://doi.org/10.3390/electronics13163335 - 22 Aug 2024
Viewed by 397
Abstract
Medical imaging is essential for pathology diagnosis and treatment, enhancing decision making and reducing costs, but despite various computational methodologies proposed to improve imaging modalities, further optimization is needed for broader acceptance. This study explores deep learning (DL) methodologies for classifying and segmenting [...] Read more.
Medical imaging is essential for pathology diagnosis and treatment, enhancing decision making and reducing costs, but despite various computational methodologies proposed to improve imaging modalities, further optimization is needed for broader acceptance. This study explores deep learning (DL) methodologies for classifying and segmenting pathological imaging data, optimizing models to accurately predict and generalize from training to new data. Different CNN and U-Net architectures are implemented for segmentation tasks, with their performance evaluated on histological image datasets using enhanced pre-processing techniques such as resizing, normalization, and data augmentation. These are trained, parameterized, and optimized using metrics such as accuracy, the DICE coefficient, and intersection over union (IoU). The experimental results show that the proposed method improves the efficiency of cell segmentation compared to networks, such as U-NET and W-UNET. The results show that the proposed pre-processing has improved the IoU from 0.9077 to 0.9675, about 7% better results; also, the values of the DICE coefficient obtained improved from 0.9215 to 0.9916, about 7% better results, surpassing the results reported in the literature. Full article
(This article belongs to the Special Issue Real-Time Computer Vision)
►▼ Show Figures

Figure 1

Figure 1
<p>U-NET architecture: the feature maps on each layer are represented by the blue boxes, the arrows represent the different operations, and the white boxes represent the cropped feature maps from the contraction path.</p> Full article ">Figure 2
<p>Workflow depicting the steps developed and executed in this study.</p> Full article ">Figure 3
<p>Graphs of the metrics obtained during the learning phase in image detection and segmentation for W-UNET architectures. (<b>A</b>) Graph of loss values vs. number of iterations (Epochs); (<b>B</b>) Graph of accuracy values vs. number of iterations (Epochs); (<b>C</b>) Graph of DICE coefficient values vs. number of iterations (Epochs); (<b>D</b>) Graph of IoU values vs. number of iterations (Epochs); (<b>E</b>) Graph of recall values vs. number of iterations (Epochs); (<b>F</b>) Graph of precision values vs. number of iterations (Epochs) obtained with the W-UNET and UNET++ architectures, respectively.</p> Full article ">Figure 4
<p>Images resulting from segmentation tests obtained with W-UNET architecture for the DSB2018 dataset.</p> Full article ">
17 pages, 1642 KiB  
Article
Leveraging Time-Critical Computation and AI Techniques for Task Offloading in Internet of Vehicles Network Applications
by Peifeng Liang, Wenhe Chen, Honghui Fan and Hongjin Zhu
Electronics 2024, 13(16), 3334; https://doi.org/10.3390/electronics13163334 - 22 Aug 2024
Viewed by 367
Abstract
Vehicular fog computing (VFC) is an innovative computing paradigm with an exceptional ability to improve the vehicles’ capacity to manage computation-intensive applications with both low latency and energy consumption. Moreover, more and more Artificial Intelligence (AI) technologies are applied in task offloading on [...] Read more.
Vehicular fog computing (VFC) is an innovative computing paradigm with an exceptional ability to improve the vehicles’ capacity to manage computation-intensive applications with both low latency and energy consumption. Moreover, more and more Artificial Intelligence (AI) technologies are applied in task offloading on the Internet of Vehicles (IoV). Focusing on the problems of computing latency and energy consumption, in this paper, we propose an AI-based Vehicle-to-Everything (V2X) model for tasks and resource offloading model for an IoV network, which ensures reliable low-latency communication, efficient task offloading in the IoV network by using a Software-Defined Vehicular-based FC (SDV-F) architecture. To fit to time-critical data transmission task distribution, the proposed model reduces unnecessary task allocation at the fog computing layer by proposing an AI-based task-allocation algorithm in the IoV layer to implement the task allocation of each vehicle. By applying AI technologies such as reinforcement learning (RL), Markov decision process, and deep learning (DL), the proposed model intelligently makes decision on maximizing resource utilization at the fog layer and minimizing the average end-to-end delay of time-critical IoV applications. The experiment demonstrates the proposed model can efficiently distribute the fog layer tasks while minimizing the delay. Full article
(This article belongs to the Special Issue AI in Information Processing and Real-Time Communication)
►▼ Show Figures

Figure 1

Figure 1
<p>The IoV’s layered architectures in [<a href="#B9-electronics-13-03334" class="html-bibr">9</a>].</p> Full article ">Figure 2
<p>Base architecture of the proposed system.</p> Full article ">Figure 3
<p>Illustration of the Intelligent Data Acquisition layer.</p> Full article ">Figure 4
<p>Framework of reinforcement learning.</p> Full article ">Figure 5
<p>Illustration of the Intelligent Data Acquisition layer.</p> Full article ">Figure 6
<p>Architecture of the U-Net used in the DQN.</p> Full article ">Figure 7
<p>Attention module of the U-Net.</p> Full article ">Figure 8
<p>Comparison of energy consumption and computation delay for different DNN architectures. (<b>a</b>) Energy consumption. (<b>b</b>) Computation delay.</p> Full article ">Figure 9
<p>Comparison of the performance of our proposed model and other baseline models in different time slots. (<b>a</b>) Energy shortfall. (<b>b</b>) Average latency. (<b>c</b>) Energy consumption.</p> Full article ">Figure 10
<p>Comparison of the performance of the proposed model and other baseline models with different number of IoVs. (<b>a</b>) Energy shortfall. (<b>b</b>) Average latency. (<b>c</b>) Energy consumption. (<b>d</b>) Average overload probability.</p> Full article ">Figure 11
<p>Comparison of the computation efficiency with different speeds when varying the number of vehicles. (<b>a</b>) Speed of 30 km/h. (<b>b</b>) Speed of 60 km/h. (<b>c</b>) Speed of 80 km/h.</p> Full article ">
17 pages, 8979 KiB  
Article
Determining the Optimal Window Duration to Enhance Emotion Recognition Based on Galvanic Skin Response and Photoplethysmography Signals
by Marcos F. Bamonte, Marcelo Risk and Victor Herrero
Electronics 2024, 13(16), 3333; https://doi.org/10.3390/electronics13163333 - 22 Aug 2024
Viewed by 442
Abstract
Automatic emotion recognition using portable sensors is gaining attention due to its potential use in real-life scenarios. Existing studies have not explored Galvanic Skin Response and Photoplethysmography sensors exclusively for emotion recognition using nonlinear features with machine learning (ML) classifiers such as Random [...] Read more.
Automatic emotion recognition using portable sensors is gaining attention due to its potential use in real-life scenarios. Existing studies have not explored Galvanic Skin Response and Photoplethysmography sensors exclusively for emotion recognition using nonlinear features with machine learning (ML) classifiers such as Random Forest, Support Vector Machine, Gradient Boosting Machine, K-Nearest Neighbor, and Decision Tree. In this study, we proposed a genuine window sensitivity analysis on a continuous annotation dataset to determine the window duration and percentage of overlap that optimize the classification performance using ML algorithms and nonlinear features, namely, Lyapunov Exponent, Approximate Entropy, and Poincaré indices. We found an optimum window duration of 3 s with 50% overlap and achieved accuracies of 0.75 and 0.74 for both arousal and valence, respectively. In addition, we proposed a Strong Labeling Scheme that kept only the extreme values of the labels, which raised the accuracy score to 0.94 for arousal. Under certain conditions mentioned, traditional ML models offer a good compromise between performance and low computational cost. Our results suggest that well-known ML algorithms can still contribute to the field of emotion recognition, provided that window duration, overlap percentage, and nonlinear features are carefully selected. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>Scatter plot of the mean annotation data, labeled according to the types of videos (each point corresponds to a video clip). Adapted from “A dataset of continuous affect annotations and physiological signals for emotion analysis” by [<a href="#B14-electronics-13-03333" class="html-bibr">14</a>]. Licensed under a Creative Commons Attribution 4.0 International License. Changes made: cropped the scatter plot and centered the legends for the video types in this figure. <a href="https://www.nature.com/articles/s41597-019-0209-0" target="_blank">https://www.nature.com/articles/s41597-019-0209-0</a> (accessed on 1 August 2024).</p> Full article ">Figure 2
<p>Number of points for each class in the CASE dataset. The label range from 1 to 9 was split into data bins of 0.1. (<b>a</b>) Arousal. (<b>b</b>) Valence.</p> Full article ">Figure 3
<p>Segment labels mapping: from continuous multivalued labels to one value annotation per segment. (<b>a</b>) Clear majority. (<b>b</b>) No clear majority. (<b>c</b>) No clear majority: extreme case.</p> Full article ">Figure 4
<p>Convolutional Neural Network with a Single-Layer Perceptron Classifier (CNN-SLP). First published in IFMBE Proceedings, Volume 1, pages 23–35, 2024 by Springer Nature [<a href="#B22-electronics-13-03333" class="html-bibr">22</a>].</p> Full article ">Figure 5
<p>Mean Accuracy performance for different window duration sizes and percentages of overlap, employing Random Forest (RF). The vertical dotted black lines represent the maximum possible window size associated with a representative median (see <a href="#electronics-13-03333-f006" class="html-fig">Figure 6</a>). (<b>a</b>) Arousal. (<b>b</b>) Valence.</p> Full article ">Figure 6
<p>Mean Dominance ground truth <math display="inline"><semantics> <msub> <mi>D</mi> <mrow> <mi>G</mi> <mi>T</mi> </mrow> </msub> </semantics></math> label for different window duration sizes and percentages of overlap. Standard deviations are shown as error bands around the average Dominance, following the color convention of the overlap legend. The vertical dotted black lines represent the maximum possible window size that meets the <math display="inline"><semantics> <mrow> <msub> <mi>D</mi> <mrow> <mi>G</mi> <mi>T</mi> </mrow> </msub> <mo>−</mo> <mi>σ</mi> <mrow> <mo>(</mo> <msub> <mi>D</mi> <mrow> <mi>G</mi> <mi>T</mi> </mrow> </msub> <mo>)</mo> </mrow> <mo>≥</mo> <mn>0.5</mn> </mrow> </semantics></math> criterion. (<b>a</b>) Arousal. (<b>b</b>) Valence.</p> Full article ">Figure 7
<p>Number of window instances for different window sizes and percentages of overlap.</p> Full article ">Figure 8
<p>Features domain performance comparison for optimum window size (<math display="inline"><semantics> <mrow> <msub> <mi>W</mi> <mrow> <mi>s</mi> <mi>i</mi> <mi>z</mi> <mi>e</mi> </mrow> </msub> <mo>=</mo> <mn>3</mn> </mrow> </semantics></math>), employing Random Forest. (<b>a</b>) Arousal. (<b>b</b>) Valence.</p> Full article ">Figure 9
<p>Labeling scheme performance comparison for optimum window size (<math display="inline"><semantics> <mrow> <msub> <mi>W</mi> <mrow> <mi>s</mi> <mi>i</mi> <mi>z</mi> <mi>e</mi> </mrow> </msub> <mo>=</mo> <mn>3</mn> </mrow> </semantics></math>), employing Random Forest. (<b>a</b>) Arousal. (<b>b</b>) Valence.</p> Full article ">Figure 10
<p>Median label of two adjacent windows for different overlap situations (window size <math display="inline"><semantics> <mrow> <msub> <mi>W</mi> <mrow> <mi>s</mi> <mi>i</mi> <mi>z</mi> <mi>e</mi> </mrow> </msub> <mo>=</mo> <mn>25</mn> </mrow> </semantics></math> s). (<b>a</b>) Overlap <math display="inline"><semantics> <mrow> <msub> <mi>O</mi> <mrow> <mi>l</mi> <mi>a</mi> <mi>p</mi> </mrow> </msub> <mo>=</mo> <mn>0</mn> <mo>%</mo> </mrow> </semantics></math>; (<b>b</b>) <math display="inline"><semantics> <mrow> <msub> <mi>O</mi> <mrow> <mi>l</mi> <mi>a</mi> <mi>p</mi> </mrow> </msub> <mo>=</mo> <mn>25</mn> <mo>%</mo> </mrow> </semantics></math>.</p> Full article ">
17 pages, 29032 KiB  
Article
Real-Time Dense Visual SLAM with Neural Factor Representation
by Weifeng Wei, Jie Wang, Xiaolong Xie, Jie Liu and Pengxiang Su
Electronics 2024, 13(16), 3332; https://doi.org/10.3390/electronics13163332 - 22 Aug 2024
Viewed by 572
Abstract
Developing a high-quality, real-time, dense visual SLAM system poses a significant challenge in the field of computer vision. NeRF introduces neural implicit representation, marking a notable advancement in visual SLAM research. However, existing neural implicit SLAM methods suffer from long runtimes and face [...] Read more.
Developing a high-quality, real-time, dense visual SLAM system poses a significant challenge in the field of computer vision. NeRF introduces neural implicit representation, marking a notable advancement in visual SLAM research. However, existing neural implicit SLAM methods suffer from long runtimes and face challenges when modeling complex structures in scenes. In this paper, we propose a neural implicit dense visual SLAM method that enables high-quality real-time reconstruction even on a desktop PC. Firstly, we propose a novel neural scene representation, encoding the geometry and appearance information of the scene as a combination of the basis and coefficient factors. This representation allows for efficient memory usage and the accurate modeling of high-frequency detail regions. Secondly, we introduce feature integration rendering to significantly improve rendering speed while maintaining the quality of color rendering. Extensive experiments on synthetic and real-world datasets demonstrate that our method achieves an average improvement of more than 60% for Depth L1 and ATE RMSE compared to existing state-of-the-art methods when running at 9.8 Hz on a desktop PC with a 3.20 GHz Intel Core i9-12900K CPU and a single NVIDIA RTX 3090 GPU. This remarkable advancement highlights the crucial importance of our approach in the field of dense visual SLAM. Full article
(This article belongs to the Special Issue Advances of Artificial Intelligence and Vision Applications)
►▼ Show Figures

Figure 1

Figure 1
<p><b>Overview.</b> (1) Scene representation: We use two different sets of factor grids to represent the scene geometry and appearance, respectively. In order to simplify our overview, we use the symbol ∗ to denote both geometry, <span class="html-italic">g</span>, and appearance, <span class="html-italic">a</span>, e.g., <math display="inline"><semantics> <mrow> <msub> <mi>b</mi> <mo>∗</mo> </msub> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> </mrow> </semantics></math> can be either <math display="inline"><semantics> <mrow> <msub> <mi>b</mi> <mi>g</mi> </msub> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> </mrow> </semantics></math> or <math display="inline"><semantics> <mrow> <msub> <mi>b</mi> <mi>a</mi> </msub> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> </mrow> </semantics></math>. For sample points along the ray, we query the basis and coefficient factors for depth and feature integration rendering. (2) Mapping process: We uniformly sample pixels from a selected set of keyframes and jointly optimize scene representation and camera poses of these keyframes. (3) Tracking process: Factor grids and MLPs remain fixed, and only the camera pose of each input frame in the RGB-D stream is estimated.</p> Full article ">Figure 2
<p>Comparison of qualitative results for reconstruction using the Replica dataset [<a href="#B28-electronics-13-03332" class="html-bibr">28</a>].</p> Full article ">Figure 3
<p>Comparison of qualitative results for reconstruction using the Replica dataset [<a href="#B28-electronics-13-03332" class="html-bibr">28</a>]. We visualize the untextured meshes.</p> Full article ">Figure 4
<p>Comparison of qualitative results for reconstruction using ScanNet [<a href="#B29-electronics-13-03332" class="html-bibr">29</a>].</p> Full article ">Figure 5
<p>Qualitative reconstruction using the ScanNet dataset [<a href="#B29-electronics-13-03332" class="html-bibr">29</a>]. We visualize the untextured meshes.</p> Full article ">Figure 6
<p>Camera trajectories of (<b>a</b>) NICE-SLAM [<a href="#B6-electronics-13-03332" class="html-bibr">6</a>], (<b>b</b>) Co-SLAM [<a href="#B8-electronics-13-03332" class="html-bibr">8</a>], (<b>c</b>) ESLAM [<a href="#B7-electronics-13-03332" class="html-bibr">7</a>] and (<b>d</b>) our method for scene0207 from the ScanNet dataset [<a href="#B29-electronics-13-03332" class="html-bibr">29</a>].</p> Full article ">Figure 7
<p>Comparison of the relationship between map size and memory usage changes. We select some scenes from the Replica [<a href="#B28-electronics-13-03332" class="html-bibr">28</a>] and ScanNet [<a href="#B29-electronics-13-03332" class="html-bibr">29</a>] datasets for evaluation.</p> Full article ">Figure 8
<p>Ablation results for feature integration rendering for Replica [<a href="#B28-electronics-13-03332" class="html-bibr">28</a>] room0.</p> Full article ">
13 pages, 7269 KiB  
Article
A High-Quality and Space-Efficient Design for Memristor Emulation
by Atul Kumar and Bhartendu Chaturvedi
Electronics 2024, 13(16), 3331; https://doi.org/10.3390/electronics13163331 - 22 Aug 2024
Viewed by 332
Abstract
The paper presents a new design for a compact memristor emulator that uses a single active component and a grounded capacitor. This design incorporates a current backward transconductance amplifier as the active element, enabling the emulation of both grounded and floating memristors in [...] Read more.
The paper presents a new design for a compact memristor emulator that uses a single active component and a grounded capacitor. This design incorporates a current backward transconductance amplifier as the active element, enabling the emulation of both grounded and floating memristors in incremental and decremental modes. The paper provides an in-depth analysis of the circuit, covering ideal, non-ideal, and parasitic factors. The theoretical performance of the memristor emulator is confirmed through post-layout simulations with 180 nm generic process design kit (gpdk) technology, demonstrating its capability to operate at low voltages (±1 V) with minimal power consumption. Additionally, the emulator shows strong performance under variations in process, voltage, and temperature (PVT) and functions effectively at a frequency of 2 MHz. Experimental validation using commercially available integrated circuits further supports the proposed design. Full article
(This article belongs to the Section Circuit and Signal Processing)
►▼ Show Figures

Figure 1

Figure 1
<p>(<b>a</b>) CBTA symbol; (<b>b</b>) CMOS implementation of CBTA [<a href="#B36-electronics-13-03331" class="html-bibr">36</a>].</p> Full article ">Figure 2
<p>Parasitic model of CBTA.</p> Full article ">Figure 3
<p>Proposed decremental/incremental memristor emulator.</p> Full article ">Figure 4
<p>Non-ideal model of the proposed memristor emulator.</p> Full article ">Figure 5
<p>Layout of CBTA.</p> Full article ">Figure 6
<p>Transient responses for input voltage and current at (<b>a</b>) 200 kHz and (<b>b</b>) 500 kHz.</p> Full article ">Figure 7
<p>Pinch characteristics of the proposed decremental memristor (<b>a</b>) at different frequencies and (<b>b</b>) for a constant product of <span class="html-italic">f</span> and <span class="html-italic">C</span>.</p> Full article ">Figure 8
<p>MCS results for frequency variations of input current signal for (<b>a</b>) MOS transistors’ mismatch and (<b>b</b>) the process variation.</p> Full article ">Figure 9
<p>Pinch characteristics at different (<b>a</b>) temperatures and (<b>b</b>) supply voltages.</p> Full article ">Figure 10
<p>Pinch characteristics at different higher frequencies for <span class="html-italic">C</span> = 5 pF.</p> Full article ">Figure 11
<p>Simulation results of incremental memristor (<b>a</b>) transient responses for input voltage and current at 500 kHz. (<b>b</b>) Pinch characteristics at different frequencies.</p> Full article ">Figure 12
<p>Transient responses for input voltage and current which confirm the volatility.</p> Full article ">Figure 13
<p>Experimental realization of CBTA used to realize the grounded ME.</p> Full article ">Figure 14
<p>Experimental setup.</p> Full article ">Figure 15
<p>Pinch characteristics at different temperatures: (<b>a</b>) 1 kHz, (<b>b</b>) 10 kHz, (<b>c</b>) 50 kHz.</p> Full article ">Figure 15 Cont.
<p>Pinch characteristics at different temperatures: (<b>a</b>) 1 kHz, (<b>b</b>) 10 kHz, (<b>c</b>) 50 kHz.</p> Full article ">
17 pages, 3683 KiB  
Article
Depth Image Rectification Based on an Effective RGB–Depth Boundary Inconsistency Model
by Hao Cao, Xin Zhao, Ang Li and Meng Yang
Electronics 2024, 13(16), 3330; https://doi.org/10.3390/electronics13163330 - 22 Aug 2024
Viewed by 384
Abstract
Depth image has been widely involved in various tasks of 3D systems with the advancement of depth acquisition sensors in recent years. Depth images suffer from serious distortions near object boundaries due to the limitations of depth sensors or estimation methods. In this [...] Read more.
Depth image has been widely involved in various tasks of 3D systems with the advancement of depth acquisition sensors in recent years. Depth images suffer from serious distortions near object boundaries due to the limitations of depth sensors or estimation methods. In this paper, a simple method is proposed to rectify the erroneous object boundaries of depth images with the guidance of reference RGB images. First, an RGB–Depth boundary inconsistency model is developed to measure whether collocated pixels in depth and RGB images belong to the same object. The model extracts the structures of RGB and depth images, respectively, by Gaussian functions. The inconsistency of two collocated pixels is then statistically determined inside large-sized local windows. In this way, pixels near object boundaries of depth images are identified to be erroneous when they are inconsistent with collocated ones in RGB images. Second, a depth image rectification method is proposed by embedding the model into a simple weighted mean filter (WMF). Experiment results on two datasets verify that the proposed method well improves the RMSE and SSIM of depth images by 2.556 and 0.028, respectively, compared with recent optimization-based and learning-based methods. Full article
(This article belongs to the Special Issue Recent Advancements in Signal and Vision Analysis)
►▼ Show Figures

Figure 1

Figure 1
<p>Weight maps of RGB and depth images when two collocated pixels are consistent: (<b>a</b>) RGB image, (<b>b</b>) weight map of RGB, (<b>c</b>) depth image, and (<b>d</b>) weight map of depth image. The points in red denote two collocated pixels at the location <span class="html-italic">i</span> in RGB and depth images. White and black regions indicate pixels with large and small weight values, respectively.</p> Full article ">Figure 2
<p>Weight maps of RGB and depth images when two collocated pixels are inconsistent: (<b>a</b>) RGB image, (<b>b</b>) weight map of RGB image, (<b>c</b>) depth image, and (<b>d</b>) weight map of depth image.</p> Full article ">Figure 3
<p>Diagram of the weights of RGB and depth images: (<b>a</b>) improvement of RGB weights; (<b>b</b>) improvement of depth weights.</p> Full article ">Figure 4
<p>The overall framework of our method.</p> Full article ">Figure 5
<p>Enlarged examples of rectified depth images: (<b>a</b>) RGB image, (<b>b</b>) raw depth, (<b>c</b>) AR [<a href="#B36-electronics-13-03330" class="html-bibr">36</a>], (<b>d</b>) RWLS [<a href="#B37-electronics-13-03330" class="html-bibr">37</a>], (<b>e</b>) EIEF [<a href="#B41-electronics-13-03330" class="html-bibr">41</a>], (<b>f</b>) DGCNN [<a href="#B29-electronics-13-03330" class="html-bibr">29</a>], (<b>g</b>) DKN [<a href="#B30-electronics-13-03330" class="html-bibr">30</a>], and (<b>h</b>) proposed.</p> Full article ">Figure 6
<p>Visual results of on whole depth images from the MPEG dataset: (<b>a</b>) RGB image, (<b>b</b>) raw depth, (<b>c</b>) AR [<a href="#B36-electronics-13-03330" class="html-bibr">36</a>], (<b>d</b>) RWLS [<a href="#B37-electronics-13-03330" class="html-bibr">37</a>], (<b>e</b>) EIEF [<a href="#B41-electronics-13-03330" class="html-bibr">41</a>], (<b>f</b>) DGCNN [<a href="#B29-electronics-13-03330" class="html-bibr">29</a>], (<b>g</b>) DKN [<a href="#B30-electronics-13-03330" class="html-bibr">30</a>], (<b>h</b>) DM [<a href="#B52-electronics-13-03330" class="html-bibr">52</a>], (<b>i</b>) SGN [<a href="#B53-electronics-13-03330" class="html-bibr">53</a>], and (<b>j</b>) proposed.</p> Full article ">Figure 7
<p>Visual results on whole depth images from the Middlebury2014 dataset: (<b>a</b>) RGB image, (<b>b</b>) raw depth, (<b>c</b>) AR [<a href="#B36-electronics-13-03330" class="html-bibr">36</a>], (<b>d</b>) RWLS [<a href="#B37-electronics-13-03330" class="html-bibr">37</a>], (<b>e</b>) EIEF [<a href="#B41-electronics-13-03330" class="html-bibr">41</a>], (<b>f</b>) DGCNN [<a href="#B29-electronics-13-03330" class="html-bibr">29</a>], (<b>g</b>) DKN [<a href="#B30-electronics-13-03330" class="html-bibr">30</a>], (<b>h</b>) DM [<a href="#B52-electronics-13-03330" class="html-bibr">52</a>], (<b>i</b>) SGN [<a href="#B53-electronics-13-03330" class="html-bibr">53</a>], and (<b>j</b>) proposed.</p> Full article ">Figure 8
<p>Analysis of parameters in our model.</p> Full article ">Figure 9
<p>Ablation study of boundary inconsistency model: (<b>a</b>) RGB, (<b>b</b>), w/o BIM, and (<b>c</b>) with BIM.</p> Full article ">
24 pages, 1749 KiB  
Article
Improved African Vulture Optimization Algorithm Based on Random Opposition-Based Learning Strategy
by Xingsheng Kuang, Junfa Hou, Xiaotong Liu, Chengming Lin, Zhu Wang and Tianlei Wang
Electronics 2024, 13(16), 3329; https://doi.org/10.3390/electronics13163329 - 22 Aug 2024
Viewed by 364
Abstract
This paper proposes an improved African vulture optimization algorithm (IROAVOA), which integrates the random opposition-based learning strategy and disturbance factor to solve problems such as the relatively weak global search capability and the poor ability to balance exploration and exploitation stages. IROAVOA is [...] Read more.
This paper proposes an improved African vulture optimization algorithm (IROAVOA), which integrates the random opposition-based learning strategy and disturbance factor to solve problems such as the relatively weak global search capability and the poor ability to balance exploration and exploitation stages. IROAVOA is divided into two parts. Firstly, the random opposition-based learning strategy is introduced in the population initialization stage to improve the diversity of the population, enabling the algorithm to more comprehensively explore the potential solution space and improve the convergence speed of the algorithm. Secondly, the disturbance factor is introduced at the exploration stage to increase the randomness of the algorithm, effectively avoiding falling into the local optimal solution and allowing a better balance of the exploration and exploitation stages. To verify the effectiveness of the proposed algorithm, comprehensive testing was conducted using the 23 benchmark test functions, the CEC2019 test suite, and two engineering optimization problems. The algorithm was compared with seven state-of-the-art metaheuristic algorithms in benchmark test experiments and compared with five algorithms in engineering optimization experiments. The experimental results indicate that IROAVOA achieved better mean and optimal values in all test functions and achieved significant improvement in convergence speed. It can also solve engineering optimization problems better than the other five algorithms. Full article
(This article belongs to the Section Artificial Intelligence)
►▼ Show Figures

Figure 1

Figure 1
<p>African vulture optimization algorithm flowchart.</p> Full article ">Figure 2
<p><span class="html-italic">w</span> change trend chart.</p> Full article ">Figure 3
<p>ROBL one-dimensional solution space.</p> Full article ">Figure 4
<p>Improved African vulture optimization algorithm flowchart.</p> Full article ">Figure 5
<p>Convergence curve of unimodal optimization function.</p> Full article ">Figure 6
<p>Convergence curve of multimodal optimization functions.</p> Full article ">Figure 6 Cont.
<p>Convergence curve of multimodal optimization functions.</p> Full article ">Figure 7
<p>Convergence curves of 23 sets of benchmark test functions between the IROAVOA and other comparative optimization algorithms.</p> Full article ">Figure 7 Cont.
<p>Convergence curves of 23 sets of benchmark test functions between the IROAVOA and other comparative optimization algorithms.</p> Full article ">Figure 8
<p>Pressure vessel design problem.</p> Full article ">Figure 9
<p>Welded beam design problem.</p> Full article ">
9 pages, 4326 KiB  
Communication
A Highly Integrated Millimeter-Wave Circularly Polarized Wide-Angle Scanning Antenna Unit
by Guishan Yuan, Sai Guo, Kan Wang and Jiawen Xu
Electronics 2024, 13(16), 3328; https://doi.org/10.3390/electronics13163328 - 22 Aug 2024
Viewed by 392
Abstract
This paper introduces a novel, small-sized, highly integrated, circularly polarized wide-angle scanning antenna using substrate-integrated waveguide (SIW) technology at millimeter-wave frequencies. The antenna unit addresses requirements for high data transmission rates, wide spatial coverage, and strong interference resistance in communication systems. By integrating [...] Read more.
This paper introduces a novel, small-sized, highly integrated, circularly polarized wide-angle scanning antenna using substrate-integrated waveguide (SIW) technology at millimeter-wave frequencies. The antenna unit addresses requirements for high data transmission rates, wide spatial coverage, and strong interference resistance in communication systems. By integrating radiating square waveguides, circular polarizers, filters, and matching loads, the antenna enhances out-of-band suppression, eliminates cross-polarization, and reduces manufacturing complexity and costs. Utilizing this antenna unit as a component, a 4 × 4 phased array antenna with a two-dimensional ±60° scanning capability is designed and simulated. The simulation and measurement results confirm that the phased array antenna achieves the desired scan range with a gain reduction of less than 3.9 dB. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>Schematic diagram of the SIW square waveguide.</p> Full article ">Figure 2
<p>The structure of the radiating square waveguide incorporating the circular polarizer.</p> Full article ">Figure 3
<p>(<b>a</b>) Model of the SIW filter structure. (<b>b</b>) The field distribution of the SIW filter structure.</p> Full article ">Figure 4
<p>Model of the matching load structure.</p> Full article ">Figure 5
<p>Model of the antenna unit structure from different perspectives: (<b>a</b>) Perspective; (<b>b</b>) Front view; (<b>c</b>) Side view.</p> Full article ">Figure 6
<p>Electrical performance of the filter: (<b>a</b>) VSWR; (<b>b</b>) insertion loss.</p> Full article ">Figure 7
<p>The VSWR of the matching load.</p> Full article ">Figure 8
<p>The active VSWR for θ = 0°, 20°, 40°, and 60°: (<b>a</b>) φ = 0°; (<b>b</b>) φ = 90°.</p> Full article ">Figure 9
<p>The illustration of the array layout.</p> Full article ">Figure 10
<p>The simulation of the active radiation pattern at frequency <span class="html-italic">f</span><sub>0</sub> GHz.</p> Full article ">Figure 11
<p>The fabricated antenna array.</p> Full article ">Figure 12
<p>The comparison of simulated and measured results of the scanning pattern at the center frequency.</p> Full article ">
22 pages, 16164 KiB  
Article
Reducing Noise and Impact of High-Frequency Torque Ripple Caused by Injection Voltages by Using Self-Regulating Random Model Algorithm for SynRMs Sensorless Speed Control
by Yibo Guo, Lingyun Pan, Yang Yang, Yimin Gong and Xiaolei Che
Electronics 2024, 13(16), 3327; https://doi.org/10.3390/electronics13163327 - 22 Aug 2024
Viewed by 471
Abstract
For the sensorless control in a low-speed range of synchronous reluctance motors (SynRMs), injecting random high-frequency (HF) square-wave-type voltages has become a widely used and technologically mature method. It can solve the noise problem of traditional injection signal methods. However, all injection signal [...] Read more.
For the sensorless control in a low-speed range of synchronous reluctance motors (SynRMs), injecting random high-frequency (HF) square-wave-type voltages has become a widely used and technologically mature method. It can solve the noise problem of traditional injection signal methods. However, all injection signal methods will cause problems such as torque ripple, which causes speed fluctuations. This article proposes a self-regulating random model algorithm for the random injection signal method, which includes a quantity adaptive module for adding additional random processes, an evaluation module for evaluating torque deviation degree, and an updated model module that is used to receive signals from the other two modules and complete model changes and output random model elements. The main function of this algorithm is to create a model that updates to suppress the evaluation value deviation based on the evaluation situation and outputs an optimal sequence of random numbers, thereby limiting speed bias always in a small range; this can reduce unnecessary changes in the output value of the speed regulator. The feasibility and effectiveness of the proposed algorithm and control method have been demonstrated in experiments based on a 5-kW synchronous reluctance motor. Full article
(This article belongs to the Special Issue Power Electronics in Renewable Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>The schematic diagram of the injection voltage of the <span class="html-italic">d</span>-axis.</p> Full article ">Figure 2
<p>Configuration of adaptive speed estimator.</p> Full article ">Figure 3
<p>PLL for calculating electrical angle.</p> Full article ">Figure 4
<p>Using adaptive velocity estimator and PLL to obtain motor speed and electrical angle.</p> Full article ">Figure 5
<p>The schematic diagram of the eight basic injection voltages and corresponding induced currents in <span class="html-italic">d</span>-axis. (<b>a</b>) <math display="inline"><semantics> <mrow> <mi>φ</mi> <mo>=</mo> <mn>0</mn> <mo>°</mo> </mrow> </semantics></math>. (<b>b</b>) <math display="inline"><semantics> <mrow> <mi>φ</mi> <mo>=</mo> <mn>45</mn> <mo>°</mo> </mrow> </semantics></math>. (<b>c</b>) <math display="inline"><semantics> <mrow> <mi>φ</mi> <mo>=</mo> <mn>90</mn> <mo>°</mo> </mrow> </semantics></math>. (<b>d</b>) <math display="inline"><semantics> <mrow> <mi>φ</mi> <mo>=</mo> <mn>135</mn> <mo>°</mo> </mrow> </semantics></math>. (<b>e</b>) <math display="inline"><semantics> <mrow> <mi>φ</mi> <mo>=</mo> <mn>180</mn> <mo>°</mo> </mrow> </semantics></math>. (<b>f</b>) <math display="inline"><semantics> <mrow> <mi>φ</mi> <mo>=</mo> <mn>225</mn> <mo>°</mo> </mrow> </semantics></math>. (<b>g</b>) <math display="inline"><semantics> <mrow> <mi>φ</mi> <mo>=</mo> <mn>270</mn> <mo>°</mo> </mrow> </semantics></math>. (<b>h</b>) <math display="inline"><semantics> <mrow> <mi>φ</mi> <mo>=</mo> <mn>315</mn> <mo>°</mo> </mrow> </semantics></math>.</p> Full article ">Figure 6
<p>Reference diagram for calculating the impact of injected voltage on speed. (<b>a</b>) No.1. (<b>b</b>) No.2.</p> Full article ">Figure 7
<p>The variation in SynRM <span class="html-italic">dq</span>-axis inductance due to saturation effect.</p> Full article ">Figure 8
<p>The summed-up value of the speed bias of using RVIM according to <a href="#electronics-13-03327-t002" class="html-table">Table 2</a>.</p> Full article ">Figure 9
<p>Overall diagram of self-regulating random model algorithm.</p> Full article ">Figure 10
<p>Principle diagram of the quantity adaptive module.</p> Full article ">Figure 11
<p>Update Model Module. (<b>a</b>) Overall logic diagram. (<b>b</b>) Schematic diagram of model state changes.</p> Full article ">Figure 12
<p>Flowchart of using the self-regulating random model algorithm to output injection voltage.</p> Full article ">Figure 13
<p>Block diagram of speed sensorless control system for SynRMs using self-regulating random model algorithm.</p> Full article ">Figure 14
<p>PSD results of the induced current of three different control methods.</p> Full article ">Figure 15
<p>Experimental platform display diagram.</p> Full article ">Figure 16
<p>The output signal from quantity adaptive module and the number of active elements.</p> Full article ">Figure 17
<p>The evaluation result value and the injection voltage number.</p> Full article ">Figure 18
<p>Sampling results of current, speed, and torque during motor running at 100 rpm. (<b>a</b>) SRRMM with no load. (<b>b</b>) RVIM with no load. (<b>c</b>) SRRMM with 50% rated load. (<b>d</b>) RVIM with 50% rated load. (<b>e</b>) SRRMM with rated load. (<b>f</b>) RVIM with rated load.</p> Full article ">Figure 19
<p>Sampling results of current and speed during motor start-up to 100 rpm. (<b>a</b>) SRRMM with no load. (<b>b</b>) RVIM with no load. (<b>c</b>) SRRMM with rated load. (<b>d</b>) RVIM with rated load.</p> Full article ">Figure 20
<p>Sampling results of current and speed of the motor from −100 rpm to 100 rpm. (<b>a</b>) SRRMM with no load. (<b>b</b>) RVIM with no load. (<b>c</b>) SRRMM with rated load. (<b>d</b>) RVIM with rated load.</p> Full article ">Figure 21
<p>Sampling results of current and speed during acceleration and deceleration of the motor between 100 rpm and 150 rpm. (<b>a</b>) SRRMM with no-load. (<b>b</b>) RVIM with no-load. (<b>c</b>) SRRMM with rated load. (<b>d</b>) RVIM with rated load.</p> Full article ">Figure 22
<p>Collecting the noise level generated by the motor using Smart Sensor AS804B.</p> Full article ">Figure 23
<p>Record of voltage injection times for each number.</p> Full article ">Figure 24
<p>During steady state of the motor, phase current image, corresponding PSD calculation results, and Fourier transform results display diagram. (<b>a</b>) OVIM with 2.5 kHz injection voltage with no load. (<b>b</b>) OVIM with 1.25 kHz injection voltage with no load. (<b>c</b>) RVIM with no load. (<b>d</b>) SRRMM with no load. (<b>e</b>) OVIM with 2.5 kHz injection voltage with rated load. (<b>f</b>) OVIM with 1.25 kHz injection voltage with rated load. (<b>g</b>) RVIM with rated load. (<b>h</b>) SRRMM with rated load.</p> Full article ">
14 pages, 6513 KiB  
Article
An Improved SPWM Strategy for Effectively Reducing Total Harmonic Distortion
by Shaoru Zhang, Huixian Li, Yang Liu, Xiaoyan Liu, Qing Lv, Xiuju Du and Jielu Zhang
Electronics 2024, 13(16), 3326; https://doi.org/10.3390/electronics13163326 - 21 Aug 2024
Viewed by 466
Abstract
In the inverter circuit, the speed at which the MOSFET is impacted by the presence of a parasitic inductor within the printed circuit board (PCB) leads to a delay in the switching process. Furthermore, the parasitic inductor within the circuit can easily form [...] Read more.
In the inverter circuit, the speed at which the MOSFET is impacted by the presence of a parasitic inductor within the printed circuit board (PCB) leads to a delay in the switching process. Furthermore, the parasitic inductor within the circuit can easily form an LC oscillation with the parasitic capacitor of the MOSFET. These two issues result in an inconsistency between the actual output of the MOSFET and the driving signal waveform, leading to distortion in the sinusoidal pulse width modulation (SPWM) waveform and an increase in total harmonic distortion (THD). It is a common practice to mitigate gate oscillation by introducing a resistor at the gate of the MOSFET. However, elevating the resistance leads to deceleration in the charging process of the MOSFET’s parasitic capacitor, consequently causing an increase in the switching delay, and thereby increasing THD. Therefore, an effective strategy to reduce THD is proposed in this paper, while augmenting the gate resistance, computing the MOSFET switching delay, and applying corrective compensation. In this way, the inherent issues of the switch are addressed, resulting in inverter output waveforms that closely resemble sine waves and reduced THD. Through a combination of simulation and empirical experimentation, the efficacy of the proposed approach in significantly reducing THD in the inverter’s output waveform has been empirically substantiated. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>The conventional inverter.</p> Full article ">Figure 2
<p>Equivalent model of MOSFET. The red arrows indicate the direction of the current flow.</p> Full article ">Figure 3
<p>Waveforms of MOSFET: (<b>a</b>) When MOSFET is being turned ON. (<b>b</b>) When MOSFET is being turned OFF.</p> Full article ">Figure 4
<p>The driving circuit of MOSFET. The red arrows indicate the direction of the current flow.</p> Full article ">Figure 5
<p>Narrow pulses with different shapes but the same impulse: (<b>a</b>) Rectangular pulse. (<b>b</b>) Triangle pulse. (<b>c</b>) Sinusoidal half-wave pulse. (<b>d</b>) Unit pulse.</p> Full article ">Figure 6
<p>Principle of symmetric regular sampling method for unipolar SPWM wave. A represents the sampling points of the sinusoidal wave, and B represents the peak points of the triangular carrier wave.</p> Full article ">Figure 7
<p>Equivalent MOSFET simulation model considering delay and oscillation issues: (<b>a</b>) Equivalent MOSFET model. (<b>b</b>) Elevating the driving resistance to diminish oscillation.</p> Full article ">Figure 8
<p>The waveforms of the MOSFET under non-ideal conditions: (<b>a</b>) MOSFET ON-delay. (<b>b</b>) MOSFET OFF-delay.</p> Full article ">Figure 9
<p>Output waveforms of the inverter: (<b>a</b>) Near the zero crossing point. (<b>b</b>) Near the peak value.</p> Full article ">Figure 10
<p>THD of filtered output waveform at a frequency of 20 kHz: (<b>a</b>) SPWM. (<b>b</b>) The proposed method.</p> Full article ">Figure 11
<p>THD of filtered output waveform at a frequency of 50 kHz: (<b>a</b>) SPWM. (<b>b</b>) The proposed method.</p> Full article ">Figure 12
<p>THD of filtered output waveform at a frequency of 100 kHz: (<b>a</b>) SPWM. (<b>b</b>) The proposed method.</p> Full article ">Figure 13
<p>Experimental prototype of the proposed converter.</p> Full article ">Figure 14
<p>The filtered output waveform.</p> Full article ">
19 pages, 674 KiB  
Article
Zero-Shot Proxy with Incorporated-Score for Lightweight Deep Neural Architecture Search
by Thi-Trang Nguyen and Ji-Hyeong Han
Electronics 2024, 13(16), 3325; https://doi.org/10.3390/electronics13163325 - 21 Aug 2024
Viewed by 464
Abstract
Designing a high-performance neural network is a difficult task. Neural architecture search (NAS) methods aim to solve this process. However, the construction of a high-quality accuracy predictor, which is a key component of NAS, usually requires significant computation. Therefore, zero-shot proxy-based NAS methods [...] Read more.
Designing a high-performance neural network is a difficult task. Neural architecture search (NAS) methods aim to solve this process. However, the construction of a high-quality accuracy predictor, which is a key component of NAS, usually requires significant computation. Therefore, zero-shot proxy-based NAS methods have been actively and extensively investigated. In this work, we propose a new efficient zero-shot proxy, Incorporated-Score, to rank deep neural network architectures instead of using an accuracy predictor. The proposed Incorporated-Score proxy is generated by incorporating the zen-score and entropy information of the network, and it does not need to train any network. We then introduce an optimal NAS algorithm called Incorporated-NAS that targets the maximization of the Incorporated-Score of the neural network within the specified inference budgets. The experiments show that the network designed by Incorporated-NAS with Incorporated-Score outperforms the previously proposed Zen-NAS and achieves a new SOTAaccuracy on the CIFAR-10, CIFAR-100, and ImageNet datasets with a lightweight scale. Full article
(This article belongs to the Special Issue Towards Efficient and Reliable AI at the Edge)
►▼ Show Figures

Figure 1

Figure 1
<p>The overview of the proposed Incorporated-NAS. Incorporated-NAS searches for the best network architecture in the search space (<math display="inline"><semantics> <mi mathvariant="script">A</mi> </semantics></math>) based on the architecture generator and the accuracy predictor without any training.</p> Full article ">Figure 2
<p>Value of zen score and the effectiveness entropy value (<math display="inline"><semantics> <msub> <mi>E</mi> <mrow> <mi>s</mi> <mi>c</mi> <mi>o</mi> <mi>r</mi> <mi>e</mi> </mrow> </msub> </semantics></math>) through <math display="inline"><semantics> <mrow> <mi>T</mi> <mo>=</mo> </mrow> </semantics></math> 96,000 iterations in the NAS process when the model size constraint is less than 1 M in search space A (defined in <a href="#sec5dot1dot1-electronics-13-03325" class="html-sec">Section 5.1.1</a>). Black lines are trend lines of the two scores in (<b>a</b>,<b>b</b>).</p> Full article ">Figure 3
<p>Values of Incorporated-Score-l, Incorporated-Score-s, and zen-score through <math display="inline"><semantics> <mrow> <mi>T</mi> <mo>=</mo> </mrow> </semantics></math> 96,000 iterations in the NAS process in search space I. Black lines are trend lines in (<b>b</b>,<b>d</b>).</p> Full article ">
20 pages, 3209 KiB  
Article
IPLog: An Efficient Log Parsing Method Based on Few-Shot Learning
by Shuxian Liu, Libo Yun, Shuaiqi Nie, Guiheng Zhang and Wei Li
Electronics 2024, 13(16), 3324; https://doi.org/10.3390/electronics13163324 - 21 Aug 2024
Viewed by 379
Abstract
Log messages from enterprise-level software systems contain crucial runtime details. Engineers can convert log messages into structured data through log parsing, laying the foundation for downstream tasks such as log anomaly detection. Existing log parsing schemes usually underperform in production environments for several [...] Read more.
Log messages from enterprise-level software systems contain crucial runtime details. Engineers can convert log messages into structured data through log parsing, laying the foundation for downstream tasks such as log anomaly detection. Existing log parsing schemes usually underperform in production environments for several reasons: first, they often ignore the semantics of log messages; second, they are often not adapted to different systems, and their performance varies greatly; and finally, they are difficult to adapt to the complexity and variety of log formats in the real environment. In response to the limitations of current approaches, we introduce IPLog (Intelligent Parse Log), a parsing method designed to address these issues. IPLog samples a limited set of log samples based on the distribution of templates in the system’s historical logs, and allows the model to make full use of the small number of log samples to recognize common patterns of keywords and parameters through few-shot learning, and thus can be easily adapted to different systems. In addition, IPLog can further improve the grouping accuracy of log templates through a novel manual feedback merge query strategy based on the longest common prefix, thus enhancing the model’s adaptability to handle complex log formats in production environments. We conducted experiments on four newly released public log datasets, and the experimental results show that IPLog can achieve an average grouping accuracy (GA) of 0.987 and parsing accuracy (PA) of 0.914 on the four public datasets, which are the best among the mainstream parsing schemes. These results demonstrate that IPLog is effective for log parsing tasks. Full article
(This article belongs to the Section Computer Science & Engineering)
►▼ Show Figures

Figure 1

Figure 1
<p>Example of log parsing.</p> Full article ">Figure 2
<p>Overview of IPLog (where the top gray box represents the steps in the offline training phase of IPLog, the blue box is the process of prompt tuning, the yellow box represents the steps of online parsing, and the bottom green box is the postprocessing process).</p> Full article ">Figure 3
<p>Offline training phase (<b>left</b>) and online parsing phase (<b>right</b>) of IPLog, where the offline-trained model can be directly used for online parsing.</p> Full article ">Figure 4
<p>Manual feedback merge query.</p> Full article ">Figure 5
<p>Comparison of parsing results of different parsing methods for ten randomly selected log messages in HDFS (green log templates represent correct parsing).</p> Full article ">Figure 5 Cont.
<p>Comparison of parsing results of different parsing methods for ten randomly selected log messages in HDFS (green log templates represent correct parsing).</p> Full article ">Figure 6
<p>Comparison of different sampling algorithms.</p> Full article ">
10 pages, 786 KiB  
Article
Design and Implementation of a Printed Circuit Model for a Wideband Circularly Polarized Bow-Tie Antenna
by Matthew J. Dodd and Atef Z. Elsherbeni
Electronics 2024, 13(16), 3323; https://doi.org/10.3390/electronics13163323 - 21 Aug 2024
Viewed by 331
Abstract
A crossed bow-tie antenna design for S- and C-Band (2.44–7.62 GHz) with a peak gain of 7.29 dBi is presented to achieve wideband radiation efficiency greater than 90% and circular polarization with a single feed point. The polarization of the antenna is modeled [...] Read more.
A crossed bow-tie antenna design for S- and C-Band (2.44–7.62 GHz) with a peak gain of 7.29 dBi is presented to achieve wideband radiation efficiency greater than 90% and circular polarization with a single feed point. The polarization of the antenna is modeled by the input admittance of crossed bow-ties, and the model predictions are validated by experiments. A wideband matching network is designed to be tightly integrated with the antenna and produce a 103% impedance bandwidth. The matching network is decomposed into an equivalent circuit model, and an analysis is presented to demonstrate the principles of the matching network design. A prototype of the optimized antenna design is fabricated and measured to validate the analysis. Full article
(This article belongs to the Section Microwave and Wireless Communications)
►▼ Show Figures

Figure 1

Figure 1
<p>(<b>a</b>) TEM mode supported by the tapered transmission line formed in the bow-tie antenna. (<b>b</b>) Geometric parameters for analyzing the input impedance of the bow-tie antenna model. The darker shade of orange represents metallization on the top side of a dielectric substrate and the lighter shade represents metallization on the bottom side.</p> Full article ">Figure 2
<p>Input admittance of two crossed dipoles with <math display="inline"><semantics> <mrow> <mi>k</mi> <msub> <mi>a</mi> <mi>x</mi> </msub> <mo>=</mo> <mn>0.22</mn> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>k</mi> <msub> <mi>a</mi> <mi>y</mi> </msub> <mo>=</mo> <mn>0.19</mn> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <msub> <mi>α</mi> <mi>x</mi> </msub> <mo>=</mo> <mn>17.1</mn> <mo>°</mo> </mrow> </semantics></math>, and <math display="inline"><semantics> <mrow> <msub> <mi>α</mi> <mi>y</mi> </msub> <mo>=</mo> <mn>12.0</mn> <mo>°</mo> </mrow> </semantics></math>.</p> Full article ">Figure 3
<p>Simulated <math display="inline"><semantics> <msub> <mi>S</mi> <mn>11</mn> </msub> </semantics></math> results of initial design (antenna 1) compared to that of optimized design (antenna 2).</p> Full article ">Figure 4
<p>(<b>a</b>) The optimized matching network consists of an inductor radius (R), a capacitor gap (G), a transmission line length (L), and a transmission line width (W). (<b>b</b>) The equivalent circuit of the matching network.</p> Full article ">Figure 5
<p>‘Matched Real’ and ‘Matched Imaginary’ show the input impedance of the equivalent circuit model with extracted values to match the full wave simulations from HFSS, which are plotted as ‘Antenna 2 Real’ and ‘Antenna 2 Imaginary’.</p> Full article ">Figure 6
<p>(<b>a</b>) Dimensions of the optimized antenna with matching network. (<b>b</b>) Photograph of the fabricated prototype antenna, including the antenna PCB, foam spacers, coaxial feed, and microstrip feed line to the SMA connector.</p> Full article ">Figure 7
<p>Surface current distribution at 5 GHz for the optimized antenna 2 design with a printed matching network.</p> Full article ">Figure 8
<p>(<b>a</b>) The simulated and measured <math display="inline"><semantics> <msub> <mi>S</mi> <mn>11</mn> </msub> </semantics></math> of the optimized antenna design compared to the simulated <math display="inline"><semantics> <msub> <mi>S</mi> <mn>11</mn> </msub> </semantics></math> of antenna 1. The −10 dB reference line for <math display="inline"><semantics> <msub> <mi>S</mi> <mn>11</mn> </msub> </semantics></math> is plotted in green. (<b>b</b>) The measured axial ratio of the prototype antenna.</p> Full article ">Figure 9
<p>Measured and simulated normalized radiation patterns at (<b>a</b>) 3 GHz, (<b>b</b>) 4 GHz, and (<b>c</b>) 5 GHz.</p> Full article ">
20 pages, 1490 KiB  
Article
Adaptive Clustering and Scheduling for UAV-Enabled Data Aggregation
by Tien-Dung Nguyen, Tien Pham Van, Duc-Tai Le and Hyunseung Choo
Electronics 2024, 13(16), 3322; https://doi.org/10.3390/electronics13163322 - 21 Aug 2024
Viewed by 374
Abstract
Using unmanned aerial vehicles (UAVs) is an effective way to gather data from Internet of Things (IoT) devices. To reduce data gathering time and redundancy, thereby enabling the timely response of state-of-the-art systems, one can partition a network into clusters and perform aggregation [...] Read more.
Using unmanned aerial vehicles (UAVs) is an effective way to gather data from Internet of Things (IoT) devices. To reduce data gathering time and redundancy, thereby enabling the timely response of state-of-the-art systems, one can partition a network into clusters and perform aggregation within each cluster. Existing works solved the UAV trajectory planning problem, in which the energy consumption and/or flight time of the UAV is the minimization objective. The aggregation scheduling within each cluster was neglected, and they assumed that data must be ready when the UAV arrives at the cluster heads (CHs). This paper addresses the minimum time aggregation scheduling problem in duty-cycled networks with a single UAV. We propose an adaptive clustering method that takes into account the trajectory and speed of the UAV. The transmission schedule of IoT devices and the UAV departure times are jointly computed so that (1) the UAV flies continuously throughout the shortest path among the CHs to minimize the hovering time and energy consumption, and (2) data are aggregated at each CH right before the UAV arrival, to maximize the data freshness. Intensive simulation shows that the proposed scheme reduces up to 35% of the aggregation delay compared to other benchmarking methods. Full article
(This article belongs to the Special Issue Ubiquitous Sensor Networks, 2nd Edition)
►▼ Show Figures

Figure 1

Figure 1
<p>An example of a duty-cycled network. Node <span class="html-italic">c</span> is active at slot 0 and each working period consists of 3 slots. Two senders <span class="html-italic">a</span> and <span class="html-italic">b</span> must take turns sending data to node <span class="html-italic">c</span> in two different working periods.</p> Full article ">Figure 2
<p>An illustration of the hovering and flying times from the docking station to the CHs <math display="inline"><semantics> <msub> <mrow> <mo>{</mo> <msub> <mi>c</mi> <mi>i</mi> </msub> <mo>}</mo> </mrow> <mrow> <mi>i</mi> <mo>∈</mo> <mo>[</mo> <mn>1</mn> <mo>,</mo> <mi>k</mi> <mo>]</mo> </mrow> </msub> </semantics></math>. Dashed arrows indicate the sequence of visits. The blue arrow indicates the moving direction of the UAV. The <math display="inline"><semantics> <msub> <mi>α</mi> <mi>i</mi> </msub> </semantics></math> is the hovering time at <math display="inline"><semantics> <msub> <mi>c</mi> <mi>i</mi> </msub> </semantics></math>, and <math display="inline"><semantics> <msub> <mi>ζ</mi> <mi>i</mi> </msub> </semantics></math> is the flying time from <math display="inline"><semantics> <msub> <mi>c</mi> <mi>i</mi> </msub> </semantics></math> to <math display="inline"><semantics> <msub> <mi>c</mi> <mrow> <mi>i</mi> <mo>+</mo> <mn>1</mn> </mrow> </msub> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>i</mi> <mo>∈</mo> <mo>[</mo> <mn>0</mn> <mo>,</mo> <mi>k</mi> <mo>]</mo> </mrow> </semantics></math>.</p> Full article ">Figure 3
<p>Sample UAV trajectory to three CHs. The arrival times at each CH are <math display="inline"><semantics> <mrow> <msub> <mi>D</mi> <mn>1</mn> </msub> <mo>&lt;</mo> <msub> <mi>D</mi> <mn>2</mn> </msub> <mo>&lt;</mo> <msub> <mi>D</mi> <mn>3</mn> </msub> </mrow> </semantics></math>. With different time allowances to collect data, <math display="inline"><semantics> <mrow> <mi>C</mi> <msub> <mi>H</mi> <mn>1</mn> </msub> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>C</mi> <msub> <mi>H</mi> <mn>2</mn> </msub> </mrow> </semantics></math>, and <math display="inline"><semantics> <mrow> <mi>C</mi> <msub> <mi>H</mi> <mn>3</mn> </msub> </mrow> </semantics></math> may have different cluster sizes. (<b>a</b>) UAV trajectory; (<b>b</b>) CHs and the corresponding sizes.</p> Full article ">Figure 4
<p>Reverse scheduling approach with a large <math display="inline"><semantics> <msub> <mi>D</mi> <mi>k</mi> </msub> </semantics></math>. The algorithm finds the transmissions in each time slot, from the latest to the earliest. The last schedulable transmissions are in time slot <span class="html-italic">t</span><sub>min</sub>. There is no transmission in time slots [1, <span class="html-italic">t</span><sub>min</sub><math display="inline"><semantics> <mrow> <mo>−</mo> <mn>1</mn> </mrow> </semantics></math>]. All the transmissions are scheduled in [<span class="html-italic">t</span><sub>min</sub>, <math display="inline"><semantics> <msub> <mi>D</mi> <mi>k</mi> </msub> </semantics></math>].</p> Full article ">Figure 5
<p>Performing ACS on a sample topology with two CHs <math display="inline"><semantics> <mrow> <msub> <mi>c</mi> <mn>1</mn> </msub> <mo>=</mo> <mi>B</mi> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <msub> <mi>c</mi> <mn>2</mn> </msub> <mo>=</mo> <mi>A</mi> </mrow> </semantics></math>. Working period length <math display="inline"><semantics> <mrow> <mi>T</mi> <mo>=</mo> <mn>5</mn> </mrow> </semantics></math>, and the aggregation deadlines <math display="inline"><semantics> <mrow> <msub> <mi>D</mi> <mn>1</mn> </msub> <mo>=</mo> <mn>17</mn> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <msub> <mi>D</mi> <mn>2</mn> </msub> <mo>=</mo> <mn>25</mn> </mrow> </semantics></math>. (<b>a</b>) Original topology; (<b>b</b>) <math display="inline"><semantics> <mrow> <mi>t</mi> <mo>=</mo> <mn>24</mn> </mrow> </semantics></math>; (<b>c</b>) <math display="inline"><semantics> <mrow> <mi>t</mi> <mo>=</mo> <mn>19</mn> </mrow> </semantics></math>; (<b>d</b>) <math display="inline"><semantics> <mrow> <mi>t</mi> <mo>=</mo> <mn>18</mn> </mrow> </semantics></math>; (<b>e</b>) <math display="inline"><semantics> <mrow> <mi>t</mi> <mo>=</mo> <mn>17</mn> </mrow> </semantics></math>; (<b>f</b>) <math display="inline"><semantics> <mrow> <mi>t</mi> <mo>=</mo> <mn>14</mn> </mrow> </semantics></math>; (<b>g</b>) <math display="inline"><semantics> <mrow> <mi>t</mi> <mo>=</mo> <mn>13</mn> </mrow> </semantics></math>; (<b>h</b>) final schedule.</p> Full article ">Figure 6
<p>The network is deliberately divided into four equal partitions. Each partition has one CH near the center point. The arrows point out the arrival sequence.</p> Full article ">Figure 7
<p>Data aggregation delay with <math display="inline"><semantics> <mrow> <mi>L</mi> <mo>=</mo> <mn>500</mn> </mrow> </semantics></math> m and <math display="inline"><semantics> <mrow> <mi>v</mi> <mo>=</mo> <mn>2</mn> </mrow> </semantics></math> m per time slot. (<b>a</b>) <math display="inline"><semantics> <mrow> <mi>T</mi> <mo>=</mo> <mn>20</mn> </mrow> </semantics></math> slots; (<b>b</b>) <math display="inline"><semantics> <mrow> <mi>T</mi> <mo>=</mo> <mn>50</mn> </mrow> </semantics></math> slots; (<b>c</b>) <math display="inline"><semantics> <mrow> <mi>T</mi> <mo>=</mo> <mn>100</mn> </mrow> </semantics></math> slots.</p> Full article ">Figure 8
<p>Networks of different sizes but with the same aggregation delay: 30 time slots. (<b>a</b>) Two nodes; (<b>b</b>) three nodes; (<b>c</b>) four nodes.</p> Full article ">Figure 9
<p>Data aggregation delay with <math display="inline"><semantics> <mrow> <mi>L</mi> <mo>=</mo> <mn>500</mn> </mrow> </semantics></math> m, <math display="inline"><semantics> <mrow> <mi>v</mi> <mo>=</mo> <mn>2</mn> </mrow> </semantics></math> m per time slot. (<b>a</b>) <math display="inline"><semantics> <mrow> <mi>N</mi> <mo>=</mo> <mn>500</mn> </mrow> </semantics></math>; (<b>b</b>) <math display="inline"><semantics> <mrow> <mi>N</mi> <mo>=</mo> <mn>700</mn> </mrow> </semantics></math>; (<b>c</b>) <math display="inline"><semantics> <mrow> <mi>N</mi> <mo>=</mo> <mn>900</mn> </mrow> </semantics></math>.</p> Full article ">Figure 10
<p>Node–cluster distribution when <math display="inline"><semantics> <mrow> <mi>T</mi> <mo>=</mo> <mn>100</mn> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>N</mi> <mo>=</mo> <mn>500</mn> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>L</mi> <mo>=</mo> <mn>500</mn> </mrow> </semantics></math>. Cluster 1: yellow triangles, Cluster 2: blue squares, Cluster 3: black stars, Cluster 4: red circles. (<b>a</b>) ICS scheme; (<b>b</b>) ACS scheme.</p> Full article ">Figure 11
<p>Impact of the UAV speed. <math display="inline"><semantics> <mrow> <mi>T</mi> <mo>=</mo> <mn>50</mn> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>N</mi> <mo>=</mo> <mn>700</mn> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>L</mi> <mo>=</mo> <mn>500</mn> </mrow> </semantics></math>.</p> Full article ">Figure 12
<p>Docking station departure time (<math display="inline"><semantics> <msub> <mi>α</mi> <mn>0</mn> </msub> </semantics></math>) with <math display="inline"><semantics> <mrow> <mi>N</mi> <mo>=</mo> <mn>700</mn> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>L</mi> <mo>=</mo> <mn>500</mn> </mrow> </semantics></math>. (<b>a</b>) <math display="inline"><semantics> <mrow> <mi>v</mi> <mo>=</mo> <mn>1</mn> </mrow> </semantics></math> m/time slot; (<b>b</b>) <math display="inline"><semantics> <mrow> <mi>v</mi> <mo>=</mo> <mn>2</mn> </mrow> </semantics></math> m/time slot.</p> Full article ">
13 pages, 5155 KiB  
Article
A Linear Regression-Based Methodology to Improve the Stability of a Low-Cost GPS Receiver Using the Precision Timing Signals from an Atomic Clock
by Shilpa Manandhar, Sneha Saravanan, Yu Song Meng and Yung Chuen Tan
Electronics 2024, 13(16), 3321; https://doi.org/10.3390/electronics13163321 - 21 Aug 2024
Viewed by 365
Abstract
The global positioning system (GPS) is widely known for its applications in navigation, timing, and positioning. However, its accuracy can be greatly impacted by the performance of its receiver clocks, especially for a low-cost receiver equipped with lower-grade clocks like crystal oscillators. The [...] Read more.
The global positioning system (GPS) is widely known for its applications in navigation, timing, and positioning. However, its accuracy can be greatly impacted by the performance of its receiver clocks, especially for a low-cost receiver equipped with lower-grade clocks like crystal oscillators. The objective of this study is to develop a model to improve the stability of a low-cost receiver. To achieve this, a machine-learning-based linear regression algorithm is proposed to predict the differences of the low-cost GPS receiver compared to the precision timing source. Experiments were conducted using low-cost receivers like Ublox and expensive receivers like Septentrio. The model was implemented and the clocks of low-cost receivers were steered. The outcomes demonstrate a notable enhancement in the stability of low-cost receivers after the corrections were applied. This improvement underscores the efficacy of the proposed model in enhancing the performance of low-cost GPS receivers. Consequently, these low-cost receivers can be cost-effectively utilized for various purposes, particularly in applications requiring the deployment of numerous GPS receivers to achieve extensive spatial coverage. Full article
(This article belongs to the Special Issue Application of Artificial Intelligence in the New Era of Communication Networks)
►▼ Show Figures

Figure 1

Figure 1
<p>Scatter plot of the position recorded by Ublox for a randomly chosen day in a static condition for single point positioning (SPP). The latitude and longitude are in meters. The red circle represents a radius of 10 m error in 2D positioning.</p> Full article ">Figure 2
<p>Hardware at Time and Frequency Lab of Singapore [located at Fusionopolis 2 (FP2)].</p> Full article ">Figure 3
<p>Flowchart of the system setup at the Time and Frequency Lab at FP2.</p> Full article ">Figure 4
<p>Map of Singapore showing the location of the receivers: Septentrio receiver at Time Lab in FP2 (in green triangle), Ublox receiver at NTU (in gold star), and Ublox receiver at Stirling (in orange star). The diagram also shows the flow of post-processing steps for a Ublox receiver.</p> Full article ">Figure 5
<p>Results (<b>a</b>) before and (<b>b</b>) after synchronization to the external reference. The legend in each figure represents the condition. The graph with red dots is for “before” and the graph with blue dots is for “after” the synchronization.</p> Full article ">Figure 6
<p>Boxplot to compare the data from Septentrio and Ublox receivers for (<b>a</b>) NTU, (<b>b</b>) Stirling Rd, and (<b>c</b>) Septentrio receiver. The x-axes for all plots are the number of days, as shown by <a href="#electronics-13-03321-t001" class="html-table">Table 1</a> and the y-axes for all the plots are the respective Refsys values.</p> Full article ">Figure 7
<p>Flowchart of the proposed machine-learning-based correction model.</p> Full article ">Figure 8
<p>Modified Allan deviation for the <math display="inline"><semantics> <mrow> <mi>R</mi> <mi>e</mi> <mi>f</mi> <mi>s</mi> <mi>y</mi> <msub> <mi>s</mi> <mrow> <mi>u</mi> <mi>b</mi> <mi>x</mi> </mrow> </msub> </mrow> </semantics></math> values. The red curve is for before the application of the correction and the blue curve is for after the correction has been applied.</p> Full article ">
29 pages, 5764 KiB  
Article
Evaluation and Improvement of the Flexibility of Biomass Blended Burning Units in a Virtual Power Plant
by Qiwei Zheng, Heng Chen, Kaijie Gou, Peiyuan Pan, Gang Xu and Guoqiang Zhang
Electronics 2024, 13(16), 3320; https://doi.org/10.3390/electronics13163320 - 21 Aug 2024
Viewed by 362
Abstract
Aiming at the problems of small thermal power units and biomass mixed combustion units with small generation loads and insufficient primary frequency modulation capability, which cannot be connected to the virtual power plant, this paper adopts a variety of flexibility retrofit methods for [...] Read more.
Aiming at the problems of small thermal power units and biomass mixed combustion units with small generation loads and insufficient primary frequency modulation capability, which cannot be connected to the virtual power plant, this paper adopts a variety of flexibility retrofit methods for the units and explores the peak load capability of the units. Then, multiple units are coupled, and the unit coupling scheme with better economy and environmental protection is screened using comprehensive evaluation indexes. While evaluating the peaking load space of multiple unit coupling, the units’ primary frequency regulation capability and new energy consumption capability are improved. According to the calculation results, the low-pressure cylinder zero-output retrofit has the largest peaking potential among different technical paths, in which unit #3 has 27.55 MW of peaking space. The compression heat pump decoupling retrofit has the best economy, in which the daily profit of unit #3 increases from 0.93 to 1.02 million CNY with an increase of 0.09 million CNY. After the unit has been retrofitted with steam extraction, the three units can be coupled to meet the national feed-in standards. The multiple unit coupling can accommodate up to 203.44 MW of other energy sources while meeting the standard. Full article
(This article belongs to the Special Issue Advanced Power Electronics and Sustainable Energy Systems: Recent Developments, Challenges and Future Perspectives)
►▼ Show Figures

Figure 1

Figure 1
<p>Schematic diagram of the thermal system of each unit. (<b>a</b>) Unit #1; (<b>b</b>) unit #2; (<b>c</b>) unit #3.</p> Full article ">Figure 2
<p>Ebsilon model of the case unit. (<b>a</b>) Unit #1; (<b>b</b>) Unit #2; (<b>c</b>) Unit #3.</p> Full article ">Figure 3
<p>Evaluation flowchart of flexibility comprehensive evaluation index.</p> Full article ">Figure 4
<p>Sensitivity analysis for unit #1.</p> Full article ">Figure 5
<p>Sensitivity analysis for unit #2.</p> Full article ">Figure 6
<p>Sensitivity analysis for unit #3.</p> Full article ">Figure 7
<p>Comparison of the minimum output capacity of each thermal electrolytic coupling retrofit scheme of unit #1 with the original unit.</p> Full article ">Figure 8
<p>Comparison of the minimum output capacity of each thermal electrolytic coupling retrofit scheme of unit #2 with the original unit.</p> Full article ">Figure 9
<p>Comparison of the minimum output capacity of each thermal electrolytic coupling retrofit scheme of unit #3 with the original unit.</p> Full article ">Figure 10
<p>Comparison of the minimum power output improvement percentage under zero-output retrofit in different case units.</p> Full article ">Figure 11
<p>Comparison of the maximum heat load increase in extraction steam under different flexibility retrofit of each units.</p> Full article ">Figure 12
<p>Change in heat load and daily profit before and after frequency modulation in 50%.</p> Full article ">Figure 13
<p>Load curve before and after PFR retrofit of unit #1 with load change rate folding line.</p> Full article ">Figure 14
<p>Load curve and load change rate folding line before and after PFR retrofit at 75% operating condition of unit #1.</p> Full article ">Figure 15
<p>Load curve and load change rate folding line before and after PFR retrofit for THA condition of unit #2.</p> Full article ">Figure 16
<p>Load curve and load change rate folding line before and after PFR retrofit for THA condition of unit #3.</p> Full article ">Figure 17
<p>Load curve and load change rate folding line before and after PFR retrofit at 75% operating condition of unit #3.</p> Full article ">Figure 18
<p>Unit #1 THA condition coupled with unit #2 THA condition coupled with unit #3 THA condition.</p> Full article ">Figure 19
<p>Unit #1 THA condition coupled with unit #2 THA condition coupled with unit #3 75% condition.</p> Full article ">Figure 20
<p>Coupling of unit #1 75% operating condition with unit #2 THA condition with unit #3 THA condition.</p> Full article ">
21 pages, 2071 KiB  
Article
Trajectory Control of Quadrotors via Spiking Neural Networks
by Yesim Oniz
Electronics 2024, 13(16), 3319; https://doi.org/10.3390/electronics13163319 - 21 Aug 2024
Viewed by 366
Abstract
In this study, a novel control scheme based on spiking neural networks (SNNs) has been proposed to accomplish the trajectory tracking of quadrotor unmanned aerial vehicles (UAVs). The update rules for the network parameters have been derived using the Lyapunov stability theorem. Three [...] Read more.
In this study, a novel control scheme based on spiking neural networks (SNNs) has been proposed to accomplish the trajectory tracking of quadrotor unmanned aerial vehicles (UAVs). The update rules for the network parameters have been derived using the Lyapunov stability theorem. Three different trajectories have been utilized in the simulated and experimental studies to verify the efficacy of the proposed control scheme. The acquired results have been compared with the responses obtained for proportional–integral–derivative (PID) and traditional neural network controllers. Simulated and experimental studies demonstrate that the proposed SNN-based controller is capable of providing better tracking accuracy and robust system response in the presence of disturbing factors. Full article
(This article belongs to the Special Issue Recent Advances in Modelling, Control and Navigation of Ground and Aerial Robots)
►▼ Show Figures

Figure 1

Figure 1
<p>The defined coordinate systems and Euler angles for the quadrotor. The coordinate frame <span class="html-italic">B</span> is attached to the body of the quadrotor and inertial coordinate system matches the initial position of the quadrotor before takeoff.</p> Full article ">Figure 2
<p>Structure of a fully connected spiking neural network with a single hidden layer. The inputs and outputs of an SNN consist of temporal information, which corresponds to the firing times of the spikes. Neurons in a layer are connected to the neurons of the proceeding layer through multiple synapses. Each synapse has a delay and weight. The network learns by adapting the weight values. In the upper right corner of the figure, a detailed view of synaptic connections has been given. In this figure, the presynaptic neurons are connected to the postsynaptic neurons with <span class="html-italic">k</span> synaptic connections. The spikes generated by neuron <span class="html-italic">i</span> are transmitted to neuron <span class="html-italic">j</span> by these synaptic connections, each having a different delay and weight. At the moment when the membrane potential of the neuron exceeds the threshold value <math display="inline"><semantics> <msub> <mi>V</mi> <mrow> <mi>T</mi> <mi>h</mi> </mrow> </msub> </semantics></math> due to the incoming pulses, a spike will be generated and propagated to the post-synaptic neurons.</p> Full article ">Figure 3
<p>The controller structure. The proposed control scheme utilizes three controllers to attain the trajectory tracking of the quadrotor. Each controller is connected to two inputs, deviation from the reference trajectory along <span class="html-italic">j</span>-axis, <math display="inline"><semantics> <mrow> <msub> <mi>e</mi> <mi>j</mi> </msub> <mo>,</mo> <mrow> <mo>(</mo> <mi>j</mi> <mo>=</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>,</mo> <mi>h</mi> <mo>)</mo> </mrow> </mrow> </semantics></math>, and the rate of change in this deviation, <math display="inline"><semantics> <msub> <mover accent="true"> <mi>e</mi> <mo>˙</mo> </mover> <mi>j</mi> </msub> </semantics></math>. These inputs are first converted into spike trains using the encoding scheme given in Equation (<a href="#FD21-electronics-13-03319" class="html-disp-formula">21</a>), and applied to the SNN network. The network parameters have been updated using Equations (<a href="#FD18-electronics-13-03319" class="html-disp-formula">18</a>) and (<a href="#FD19-electronics-13-03319" class="html-disp-formula">19</a>). The output of the network has been decoded into a real value using Equation (<a href="#FD22-electronics-13-03319" class="html-disp-formula">22</a>) and applied to the quadrotor system as a control command.</p> Full article ">Figure 4
<p>Target trajectories used in simulations and experiments. Three different reference trajectories have been generated. The first reference trajectory has been defined on <math display="inline"><semantics> <mrow> <mi>x</mi> <mi>y</mi> </mrow> </semantics></math>-plane and it requires sudden changes in the movements of the quadrotor. The other reference trajectories require simultaneous control of the quadrotor movement along each translational axis.</p> Full article ">Figure 5
<p>Simulation results in the absence of disturbing factors for (<b>a</b>) target trajectory 1, (<b>b</b>) target trajectory 2, and (<b>c</b>) target trajectory 3.</p> Full article ">Figure 6
<p>Simulation results in the presence of sinusoidal disturbing factors for (<b>a</b>) target trajectory 1, (<b>b</b>) target trajectory 2, and (<b>c</b>) target trajectory 3. To imitate the effect of a wind disturbance acting along x- and y-directions, a disturbance of <math display="inline"><semantics> <mrow> <mi>d</mi> <mo>(</mo> <mi>t</mi> <mo>)</mo> <mo>=</mo> <mn>0.2</mn> <mi>s</mi> <mi>i</mi> <mi>n</mi> <mo>(</mo> <mn>2.5</mn> <mi>π</mi> <mi>t</mi> <mo>)</mo> </mrow> </semantics></math> has been included in <span class="html-italic">x</span>- and <span class="html-italic">y</span>-axis.</p> Full article ">Figure 7
<p>Responses of the controllers along x-direction in the presence of sinusoidal disturbing factors acting in x- and y-directions: (<b>a</b>) target trajectory 1, (<b>b</b>) target trajectory 2, and (<b>c</b>) target trajectory 3.</p> Full article ">Figure 8
<p>Responses of the controllers along y-direction in the presence of sinusoidal disturbing factors acting in x- and y-directions: (<b>a</b>) target trajectory 1, (<b>b</b>) target trajectory 2, and (<b>c</b>) target trajectory 3.</p> Full article ">Figure 9
<p>Simulation results for discrete wind gust model for (<b>a</b>) target trajectory 1, (<b>b</b>) target trajectory 2, and (<b>c</b>) target trajectory 3. The disturbing factors acting in all three axes have been generated using the discrete wind gust model given in Equation (<a href="#FD24-electronics-13-03319" class="html-disp-formula">24</a>).</p> Full article ">Figure 10
<p>Experimental results for (<b>a</b>) target trajectory 1, (<b>b</b>) target trajectory 2, and (<b>c</b>) target trajectory 3. Reference trajectory and control commands have been transmitted to the quadrotor once the altitude reaches 75 cm.</p> Full article ">
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
 
Displaying articles 1-246
Previous Issue
Next Issue
Back to TopTop