This paper presents a machine learning (ML) approach to modeling radio signal strength in an open-air amphitheater environment. Towards this end, we have first collected field measurement data from an amphitheater located on the Malaysia Campus of the University of Nottingham at 900 MHz, 2.4 and 5.8 GHz. We have conducted the measurement campaign over the course of six months, with repeatability test done for each frequency for at least one cavea level to make sure the results are repeatable. These measurement data are plotted and analyzed before being fed to a ML model for training purposes. In particular, we have explored three options of the ML methods, namely, the Linear Regression method, the Random Forest method, and the Neural Network method; and finally settled on the Neural Network method for is superiority over the other two methods–it performs better when more input data were inserted to train continuously. In addition, we have run ray-tracing simulation to provide an extra layer of comparison to the ML-generated prediction results. Beyond this, we have expanded the ML model to account for a larger geometry of amphitheater. The output of this work is expected to enhance wireless communication reliability in amphitheaters, with potential benefits for event management, public safety, and entertainment industries.
{"title":"Machine-Learning-Empowered Propagation Measurement and Modeling for an Amphitheater","authors":"Soo Yong Lim;Chee Ren Ong;Jay Sern Chow;Khee Lee;Qi Ping Soo;Juinn-Horng Deng;Jeng-Kuang Hwang;Sheng-Kai Chen;Yan-Di Liu;Yu-Chien Wu;Hsiang-Chuan Hsien","doi":"10.1109/ACCESS.2026.3664462","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3664462","url":null,"abstract":"This paper presents a machine learning (ML) approach to modeling radio signal strength in an open-air amphitheater environment. Towards this end, we have first collected field measurement data from an amphitheater located on the Malaysia Campus of the University of Nottingham at 900 MHz, 2.4 and 5.8 GHz. We have conducted the measurement campaign over the course of six months, with repeatability test done for each frequency for at least one cavea level to make sure the results are repeatable. These measurement data are plotted and analyzed before being fed to a ML model for training purposes. In particular, we have explored three options of the ML methods, namely, the Linear Regression method, the Random Forest method, and the Neural Network method; and finally settled on the Neural Network method for is superiority over the other two methods–it performs better when more input data were inserted to train continuously. In addition, we have run ray-tracing simulation to provide an extra layer of comparison to the ML-generated prediction results. Beyond this, we have expanded the ML model to account for a larger geometry of amphitheater. The output of this work is expected to enhance wireless communication reliability in amphitheaters, with potential benefits for event management, public safety, and entertainment industries.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"26733-26741"},"PeriodicalIF":3.6,"publicationDate":"2026-02-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11396656","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146223673","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-16DOI: 10.1109/ACCESS.2026.3664744
Abdul Kadar Muhammad Masum;Md. Abul Kalam Azad;Chanda Rani Debi;Ramona Birǎu;Virgil Popescu;Costel Marian Ionascu
Early warning of systemic financial instability is a crucial issue for regulators and asset managers, especially in detecting flash crashes and concealed contagion paths, which cannot be detected using conventional surveillance mechanisms. In this context, conventional surveillance mechanisms refer to linear correlation based monitoring, volatility threshold rules, and static network centrality measures that rely on fixed topological summaries. Traditional regulatory mechanisms are based on a set of static measures of topology, such as Degree Centrality, PageRank, and matrices of linear correlations that assess the size of assets in terms of positioning in a financial network, systematically ignoring small and highly leveraged institutions that are sometimes crucial bridges in financial networks. We propose a novel geometric deep learning architecture, the Neuro Ricci Flow, a geometric framework that combines Financial Returns and ESG Momentum into a single Riemannian manifold, which conceptualizes systemic risk as real world changes in market topology and not statistical volatility, as commonly interpreted. Financial returns and ESG momentum are selected as universally observable and high frequency market indicators that directly encode dynamic interactions and stress propagation in financial systems. The framework is designed to detect market level structural instability and contagion dynamics, rather than to directly infer macroeconomic phenomena such as currency debasement or monetary policy driven effects. We consider a Neural Ordinary Differential Equation to learn the underlying physics of Ricci Flow, with dynamical simulation of market manifold evolution to subsequently find hyperbolic singularities, in which an area of extreme negative curvature indicates structural rupture in the network. Empirical valuation shows better predictive power with 100 percent recall of the geometric risk indicator with five percent baseline risk. The deeper meaning further elaborates on the practical implications of this framework, which include accurate alarm mechanisms for central banks, assisted by Ricci Curvature Analysis and geometric techniques for immunizing the portfolios of asset managers in need of protection against contagion. We discover that geometric deep learning has provided a more sensitive paradigm of systemic risk models compared to traditional topological models.
{"title":"From Topology to Geometry: A Neural Ricci Flow Framework for Predicting Flash Crashes and Contagion","authors":"Abdul Kadar Muhammad Masum;Md. Abul Kalam Azad;Chanda Rani Debi;Ramona Birǎu;Virgil Popescu;Costel Marian Ionascu","doi":"10.1109/ACCESS.2026.3664744","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3664744","url":null,"abstract":"Early warning of systemic financial instability is a crucial issue for regulators and asset managers, especially in detecting flash crashes and concealed contagion paths, which cannot be detected using conventional surveillance mechanisms. In this context, conventional surveillance mechanisms refer to linear correlation based monitoring, volatility threshold rules, and static network centrality measures that rely on fixed topological summaries. Traditional regulatory mechanisms are based on a set of static measures of topology, such as Degree Centrality, PageRank, and matrices of linear correlations that assess the size of assets in terms of positioning in a financial network, systematically ignoring small and highly leveraged institutions that are sometimes crucial bridges in financial networks. We propose a novel geometric deep learning architecture, the Neuro Ricci Flow, a geometric framework that combines Financial Returns and ESG Momentum into a single Riemannian manifold, which conceptualizes systemic risk as real world changes in market topology and not statistical volatility, as commonly interpreted. Financial returns and ESG momentum are selected as universally observable and high frequency market indicators that directly encode dynamic interactions and stress propagation in financial systems. The framework is designed to detect market level structural instability and contagion dynamics, rather than to directly infer macroeconomic phenomena such as currency debasement or monetary policy driven effects. We consider a Neural Ordinary Differential Equation to learn the underlying physics of Ricci Flow, with dynamical simulation of market manifold evolution to subsequently find hyperbolic singularities, in which an area of extreme negative curvature indicates structural rupture in the network. Empirical valuation shows better predictive power with 100 percent recall of the geometric risk indicator with five percent baseline risk. The deeper meaning further elaborates on the practical implications of this framework, which include accurate alarm mechanisms for central banks, assisted by Ricci Curvature Analysis and geometric techniques for immunizing the portfolios of asset managers in need of protection against contagion. We discover that geometric deep learning has provided a more sensitive paradigm of systemic risk models compared to traditional topological models.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"26912-26934"},"PeriodicalIF":3.6,"publicationDate":"2026-02-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11396506","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146223647","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-13DOI: 10.1109/ACCESS.2026.3664497
Yuning Song
Fault diagnosis of drainage pumps is essential for ensuring the safe and stable operation of drainage systems. However, vibration signals acquired from drainage pumps under practical operating conditions are often characterized by strong non-stationarity, unavoidable environmental interference, and fault features distributed across multiple temporal scales, which poses significant challenges to accurate multi-class fault identification. To tackle these challenges, this paper proposes a multi-scale residual attention convolutional neural network (MRANet) for drainage pump fault diagnosis. A multi-scale residual attention feature extraction (MRAFE) module is designed to jointly model fault-related information at different temporal scales by means of parallel convolutions and residual connections, while attention mechanisms are incorporated within each scale branch to enhance discriminative fault features. Furthermore, a complementary multi-level feature integration (CMFI) strategy is developed to effectively integrate features extracted at different network depths, enabling end-to-end fault identification directly from raw one-dimensional vibration signals. Extensive experiments are conducted on a drainage pump vibration dataset containing eight operating conditions. Experimental results show that MRANet achieves an accuracy of 95.43%, outperforming several benchmark models. These results verify the effectiveness and robustness of the proposed method for multi-class fault diagnosis of drainage pumps.
{"title":"A Multi-Scale Residual Attention Network for Drainage Pump Fault Diagnosis","authors":"Yuning Song","doi":"10.1109/ACCESS.2026.3664497","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3664497","url":null,"abstract":"Fault diagnosis of drainage pumps is essential for ensuring the safe and stable operation of drainage systems. However, vibration signals acquired from drainage pumps under practical operating conditions are often characterized by strong non-stationarity, unavoidable environmental interference, and fault features distributed across multiple temporal scales, which poses significant challenges to accurate multi-class fault identification. To tackle these challenges, this paper proposes a multi-scale residual attention convolutional neural network (MRANet) for drainage pump fault diagnosis. A multi-scale residual attention feature extraction (MRAFE) module is designed to jointly model fault-related information at different temporal scales by means of parallel convolutions and residual connections, while attention mechanisms are incorporated within each scale branch to enhance discriminative fault features. Furthermore, a complementary multi-level feature integration (CMFI) strategy is developed to effectively integrate features extracted at different network depths, enabling end-to-end fault identification directly from raw one-dimensional vibration signals. Extensive experiments are conducted on a drainage pump vibration dataset containing eight operating conditions. Experimental results show that MRANet achieves an accuracy of 95.43%, outperforming several benchmark models. These results verify the effectiveness and robustness of the proposed method for multi-class fault diagnosis of drainage pumps.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"34585-34599"},"PeriodicalIF":3.6,"publicationDate":"2026-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11395939","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147362271","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-13DOI: 10.1109/ACCESS.2026.3664454
Xing Wang;Hongxiang Xu;Tao Gu;Chengcheng Xu
Accurate and early diagnosis of Alzheimer’s disease (AD) is critical for timely intervention, yet existing multimodal approaches typically require precisely paired magnetic resonance imaging (MRI) and positron emission tomography (PET) scans, a condition that is rarely met in clinical practice due to missing modalities, protocol heterogeneity, and logistical constraints. To overcome these limitations, we propose a Cross-Modal Adversarial Network (CMA-Net) that enables effective AD diagnosis using unpaired MRI and PET images during training and supports flexible inference with either modality alone. The core of our framework is a Multimodal Adaptive Convolutional Neural Network (MA-CNN), which adopts dual-branch parallel processing with modality-specific Multi-Modal Convolution (MM Conv) blocks and integrates multi-scale attention mechanisms to extract discriminative features while preserving the intrinsic characteristics of each imaging modality. Extensive experiments on two publicly available, unpaired AD datasets demonstrate that our method achieves state-of-the-art performance, yielding classification accuracies of $96.27pm 0.27$ % on MRI and $94.92pm 0.19$ % on PET. These results underscore the potential of CMA-Net to facilitate robust, multimodal AD diagnosis without the stringent requirement of image pairing, thereby enhancing its applicability in real-world clinical settings.
{"title":"A Cross-Modal Adversarial Network for Alzheimer’s Disease Diagnosis Using Unpaired MRI and PET Imaging","authors":"Xing Wang;Hongxiang Xu;Tao Gu;Chengcheng Xu","doi":"10.1109/ACCESS.2026.3664454","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3664454","url":null,"abstract":"Accurate and early diagnosis of Alzheimer’s disease (AD) is critical for timely intervention, yet existing multimodal approaches typically require precisely paired magnetic resonance imaging (MRI) and positron emission tomography (PET) scans, a condition that is rarely met in clinical practice due to missing modalities, protocol heterogeneity, and logistical constraints. To overcome these limitations, we propose a Cross-Modal Adversarial Network (CMA-Net) that enables effective AD diagnosis using unpaired MRI and PET images during training and supports flexible inference with either modality alone. The core of our framework is a Multimodal Adaptive Convolutional Neural Network (MA-CNN), which adopts dual-branch parallel processing with modality-specific Multi-Modal Convolution (MM Conv) blocks and integrates multi-scale attention mechanisms to extract discriminative features while preserving the intrinsic characteristics of each imaging modality. Extensive experiments on two publicly available, unpaired AD datasets demonstrate that our method achieves state-of-the-art performance, yielding classification accuracies of <inline-formula> <tex-math>$96.27pm 0.27$ </tex-math></inline-formula>% on MRI and <inline-formula> <tex-math>$94.92pm 0.19$ </tex-math></inline-formula>% on PET. These results underscore the potential of CMA-Net to facilitate robust, multimodal AD diagnosis without the stringent requirement of image pairing, thereby enhancing its applicability in real-world clinical settings.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"26935-26952"},"PeriodicalIF":3.6,"publicationDate":"2026-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11395969","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146223656","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-12DOI: 10.1109/ACCESS.2026.3664214
Roy Hershkovitz;Yossi Oren
Sensor spoofing attacks are a serious threat to mobile phones, as they can manipulate sensor readings to subvert the behavior of applications that rely on these readings. Previous work has shown how machine learning defenses provide effective protection against sensor spoofing attacks without hardware modification. Unfortunately, these defenses require changes to the applications themselves. In this paper, we present $textsf {SDIOS}$ (Sensor Defense in the Operating System), an approach that engineers sensor spoofing protection into the operating system level, without requiring any modifications to the applications. At its core, $textsf {SDIOS}$ incorporates an autoencoder based on a Gramian Angular Field (GAF) representation of the sensor readings. We describe the design and implementation of $textsf {SDIOS}$ , and evaluate its performance and compatibility on a variety of devices. Our results show that $textsf {SDIOS}$ is able to detect and prevent sensor spoofing attacks in real time, while retaining compatibility with existing applications, but that its performance impact is significant, especially on resource-constrained devices where the machine learning pipeline is run on the central processing unit (CPU).
{"title":"Engineering Sensor Spoofing Protection Into the Android Operating System","authors":"Roy Hershkovitz;Yossi Oren","doi":"10.1109/ACCESS.2026.3664214","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3664214","url":null,"abstract":"Sensor spoofing attacks are a serious threat to mobile phones, as they can manipulate sensor readings to subvert the behavior of applications that rely on these readings. Previous work has shown how machine learning defenses provide effective protection against sensor spoofing attacks without hardware modification. Unfortunately, these defenses require changes to the applications themselves. In this paper, we present <inline-formula> <tex-math>$textsf {SDIOS}$ </tex-math></inline-formula> (Sensor Defense in the Operating System), an approach that engineers sensor spoofing protection into the operating system level, without requiring any modifications to the applications. At its core, <inline-formula> <tex-math>$textsf {SDIOS}$ </tex-math></inline-formula> incorporates an autoencoder based on a Gramian Angular Field (GAF) representation of the sensor readings. We describe the design and implementation of <inline-formula> <tex-math>$textsf {SDIOS}$ </tex-math></inline-formula>, and evaluate its performance and compatibility on a variety of devices. Our results show that <inline-formula> <tex-math>$textsf {SDIOS}$ </tex-math></inline-formula> is able to detect and prevent sensor spoofing attacks in real time, while retaining compatibility with existing applications, but that its performance impact is significant, especially on resource-constrained devices where the machine learning pipeline is run on the central processing unit (CPU).","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"26899-26911"},"PeriodicalIF":3.6,"publicationDate":"2026-02-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11394777","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146223633","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-11DOI: 10.1109/ACCESS.2026.3663875
Zhenbo Zhang;Zhiguo Feng;Aiqi Long;Zhiyu Wang;Xingqiang Tian;Wei Xiang;Zhenyin Tu
Multi-scale detection is vital for autonomous driving. In sparse scenarios such as highways, where targets often appear small, distant, and susceptible to substantial background interference, traditional models suffer from feature distortion, leading to missed detections and compromised safety. This study proposes S2-DETR to address these challenges, featuring an innovative hierarchical Sparse-to-Spatial Attention Mechanism (S2AM), which synergistically integrates a Dynamic Sparse Attention Module (DSAM) for coarse-grained sparse feature enhancement with a Spatial Attention Module (SAM) for fine-grained refinement. This design is particularly effective for enhancing the representation of small and hard-to-detect targets in complex visual environments. We further designed a Cross-scale Attention Pyramid Module (CAPM) that embeds S2AM within a dual-path architecture inspired by Feature Pyramid Networks and Path Aggregation Networks, replacing RT-DETR’s original fusion module to optimize multi-scale feature representation. Extensive ablation studies validated our S2AM and CAPM designs. Comparative experiments confirmed S2-DETR’s superiority: on public and self-built sparse datasets, it achieved accuracy improvements of 8.7% and 14.1%, respectively, over its RT-DETR baseline, with only a 7.5% speed trade-off. These results establish an improved accuracy-speed balance, notably for challenging small and multi-scale targets. The source code will be released on GitHub to foster further research in traffic participant detection for autonomous driving.
{"title":"S2-DETR: Hierarchical Sparse-to-Spatial Attention Enhanced DETR for Traffic Participant Detection in Sparse Autonomous Driving Scene","authors":"Zhenbo Zhang;Zhiguo Feng;Aiqi Long;Zhiyu Wang;Xingqiang Tian;Wei Xiang;Zhenyin Tu","doi":"10.1109/ACCESS.2026.3663875","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3663875","url":null,"abstract":"Multi-scale detection is vital for autonomous driving. In sparse scenarios such as highways, where targets often appear small, distant, and susceptible to substantial background interference, traditional models suffer from feature distortion, leading to missed detections and compromised safety. This study proposes S2-DETR to address these challenges, featuring an innovative hierarchical Sparse-to-Spatial Attention Mechanism (S2AM), which synergistically integrates a Dynamic Sparse Attention Module (DSAM) for coarse-grained sparse feature enhancement with a Spatial Attention Module (SAM) for fine-grained refinement. This design is particularly effective for enhancing the representation of small and hard-to-detect targets in complex visual environments. We further designed a Cross-scale Attention Pyramid Module (CAPM) that embeds S2AM within a dual-path architecture inspired by Feature Pyramid Networks and Path Aggregation Networks, replacing RT-DETR’s original fusion module to optimize multi-scale feature representation. Extensive ablation studies validated our S2AM and CAPM designs. Comparative experiments confirmed S2-DETR’s superiority: on public and self-built sparse datasets, it achieved accuracy improvements of 8.7% and 14.1%, respectively, over its RT-DETR baseline, with only a 7.5% speed trade-off. These results establish an improved accuracy-speed balance, notably for challenging small and multi-scale targets. The source code will be released on GitHub to foster further research in traffic participant detection for autonomous driving.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"29475-29492"},"PeriodicalIF":3.6,"publicationDate":"2026-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11393635","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147292793","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-10DOI: 10.1109/ACCESS.2026.3663494
Sachini P. Somathilaka;Nadun T. Senarathna;H. M. Wijekoon Banda;K. T. M. Udayanga Hemapala
The growing integration of variable renewable energy (VRE) sources into small-scale, isolated, low-inertia power systems has intensified frequency fluctuations, placing increased stress on conventional high-inertia generators and their governors. This study addresses the urgent need for a cost-effective, intermediate solution to mitigate these fluctuations, particularly in systems with limited access to high-capital technologies such as battery storage. A modified Particle Swarm Optimization (PSO) algorithm is proposed, integrating a non-linear inertia weight function and a storming mechanism to overcome the local optimum limitations of conventional PSO. The algorithm was applied to the Laxapana Hydro Complex in Sri Lanka, a key power generation and frequency regulation facility, and an ideal example of a small-scale isolated power system. Real-world system data were used in simulations conducted in MATLAB and PSSE environments. Simulation results across three operational scenarios: general generation fluctuations, solar intermittency, and wind intermittency, demonstrate that the optimized governor settings significantly reduce long-term mechanical movement while maintaining system stability. Specifically, under solar intermittency, governor actuator exhibited approximately a 20% reduction in movement compared to the baseline configuration. This reduction minimizes wear and tear, leading to lower maintenance costs and improved long-term reliability. These findings highlight the potential of the proposed method as a low-cost, adaptable solution for improving governor performance in renewable-integrated systems, offering a pathway toward more reliable and cost-effective power system operations.
{"title":"Enhanced Particle Swarm Optimization for Minimizing Governor Actuation in Hydropower Plants Under Renewable Energy Intermittency","authors":"Sachini P. Somathilaka;Nadun T. Senarathna;H. M. Wijekoon Banda;K. T. M. Udayanga Hemapala","doi":"10.1109/ACCESS.2026.3663494","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3663494","url":null,"abstract":"The growing integration of variable renewable energy (VRE) sources into small-scale, isolated, low-inertia power systems has intensified frequency fluctuations, placing increased stress on conventional high-inertia generators and their governors. This study addresses the urgent need for a cost-effective, intermediate solution to mitigate these fluctuations, particularly in systems with limited access to high-capital technologies such as battery storage. A modified Particle Swarm Optimization (PSO) algorithm is proposed, integrating a non-linear inertia weight function and a storming mechanism to overcome the local optimum limitations of conventional PSO. The algorithm was applied to the Laxapana Hydro Complex in Sri Lanka, a key power generation and frequency regulation facility, and an ideal example of a small-scale isolated power system. Real-world system data were used in simulations conducted in MATLAB and PSSE environments. Simulation results across three operational scenarios: general generation fluctuations, solar intermittency, and wind intermittency, demonstrate that the optimized governor settings significantly reduce long-term mechanical movement while maintaining system stability. Specifically, under solar intermittency, governor actuator exhibited approximately a 20% reduction in movement compared to the baseline configuration. This reduction minimizes wear and tear, leading to lower maintenance costs and improved long-term reliability. These findings highlight the potential of the proposed method as a low-cost, adaptable solution for improving governor performance in renewable-integrated systems, offering a pathway toward more reliable and cost-effective power system operations.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"27005-27022"},"PeriodicalIF":3.6,"publicationDate":"2026-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11389791","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146223648","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-10DOI: 10.1109/ACCESS.2026.3663468
Xiaoyue Gu;Yuhao Ren;Fabing Duan;Derek Abbott
Motivated by the principle of stochastic resonance, we investigate the noise-boosted activations within both channel attention mechanisms of convolutional networks and gated linear unit (GLU)-based feedforward networks (FFNs) of Vision Transformers (ViTs) under attention-based visual processing frameworks. Specifically, we replace conventional ReLU or ReLU-based GLU (ReGLU) activations with noise-boosted variants, which incorporate learnable noise scale parameters during training. Experiments on the CIFAR-10 and STL-10 image classifications, Kvasir-SEG medical image segmentation, and Cityscapes semantic segmentation show significant improvements over conventional baselines across diverse attention architectures. The learnable noise scale parameters in activations converge to non-zero values after training, demonstrating the existence of stochastic resonance in deep attention mechanisms. These results indicate that controlled noise injection can enhance information transfer efficiency of neural networks, and establish a coherent framework that connects the theoretical principle of stochastic resonance with its practical applicability in attention-based visual processing.
{"title":"Enhancing Attention-Based Visual Processing With Noise-Boosted Activation Functions","authors":"Xiaoyue Gu;Yuhao Ren;Fabing Duan;Derek Abbott","doi":"10.1109/ACCESS.2026.3663468","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3663468","url":null,"abstract":"Motivated by the principle of stochastic resonance, we investigate the noise-boosted activations within both channel attention mechanisms of convolutional networks and gated linear unit (GLU)-based feedforward networks (FFNs) of Vision Transformers (ViTs) under attention-based visual processing frameworks. Specifically, we replace conventional ReLU or ReLU-based GLU (ReGLU) activations with noise-boosted variants, which incorporate learnable noise scale parameters during training. Experiments on the CIFAR-10 and STL-10 image classifications, Kvasir-SEG medical image segmentation, and Cityscapes semantic segmentation show significant improvements over conventional baselines across diverse attention architectures. The learnable noise scale parameters in activations converge to non-zero values after training, demonstrating the existence of stochastic resonance in deep attention mechanisms. These results indicate that controlled noise injection can enhance information transfer efficiency of neural networks, and establish a coherent framework that connects the theoretical principle of stochastic resonance with its practical applicability in attention-based visual processing.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"26720-26732"},"PeriodicalIF":3.6,"publicationDate":"2026-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11389769","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146223632","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-10DOI: 10.1109/ACCESS.2026.3663213
Mohammed Lataifeh;Zulaiha Afrah Sadakathullah Shaduly;Naveed Ahmed
This study investigates how different interaction modalities influence user performance and experience in virtual reality environments. We compared three commonly used modalities, including controller-based input, hand tracking, and eye gaze, as well as a combined modality that integrated multiple inputs. We conducted an empirical user study with 30 participants (10 males, 20 females), in which each participant completed two tasks: arranging toruses on a vertical pole and stacking virtual blocks in predefined positions. We evaluated the efficiency, effectiveness, and user preference for each modality through quantitative and qualitative measures. Quantitative data included task completion time, error rates, and responses to closed-ended questions regarding the preferred modalities, whereas qualitative data included survey responses to the open-ended questions. Statistical analysis revealed that the controller modality resulted in faster task completion times with the least error rate, whereas the eye gaze took significantly longer completion times with the highest error rates. The majority of participants preferred using the controller for its efficiency and effectiveness, highlighting that interaction modality plays a crucial role in determining user performance and experience in basic object manipulation tasks in VR.
{"title":"Usability Evaluation of Interaction Modalities for 3-D Object Manipulation in Immersive Virtual Reality Systems","authors":"Mohammed Lataifeh;Zulaiha Afrah Sadakathullah Shaduly;Naveed Ahmed","doi":"10.1109/ACCESS.2026.3663213","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3663213","url":null,"abstract":"This study investigates how different interaction modalities influence user performance and experience in virtual reality environments. We compared three commonly used modalities, including controller-based input, hand tracking, and eye gaze, as well as a combined modality that integrated multiple inputs. We conducted an empirical user study with 30 participants (10 males, 20 females), in which each participant completed two tasks: arranging toruses on a vertical pole and stacking virtual blocks in predefined positions. We evaluated the efficiency, effectiveness, and user preference for each modality through quantitative and qualitative measures. Quantitative data included task completion time, error rates, and responses to closed-ended questions regarding the preferred modalities, whereas qualitative data included survey responses to the open-ended questions. Statistical analysis revealed that the controller modality resulted in faster task completion times with the least error rate, whereas the eye gaze took significantly longer completion times with the highest error rates. The majority of participants preferred using the controller for its efficiency and effectiveness, highlighting that interaction modality plays a crucial role in determining user performance and experience in basic object manipulation tasks in VR.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"26702-26719"},"PeriodicalIF":3.6,"publicationDate":"2026-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11389780","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146223679","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-09DOI: 10.1109/ACCESS.2026.3660678
Uichan Kim;Atom O. Watanabe;Woo-Su Kim;Youngwoo Kim
This article proposes dispersion analysis for an efficient design of electromagnetic bandgap (EBG) structures in glass interposers to suppress power/ground noise coupling. The proposed method considered design parameters such as patch, through glass via (TGV), and defected structure which can be efficiently realized in glass interposers without additional fabrication steps. The unit cell of the EBG structure is modeled into transmission lines for an efficient dispersion analysis to estimate the power/ground noise suppression band. Impacts of capacitance associated with the patch, inductance of TGVs, and defected structure on the noise suppression bands are characterized for an efficient bandgap design. Various test vehicles are fabricated and measured to validate the proposed design methodology based on the dispersion analysis. The measured results showed good correlation with estimated noise suppression bandgaps which verifies the proposed design methodology based on dispersion analysis. The proposed method is further applied to design an EBG structure including multiple noise suppression bands for broadband noise suppression. The proposed design is validated using the 3-dimensional electromagnetic simulation. Effectiveness of the proposed EBG structure in the glass interposer on signal and power integrity is verified by analyzing noise propagation in the power delivery network and coupling to the TGV channel.
{"title":"Design of Electromagnetic Bandgap Structures in Glass Interposers Based on Dispersion Analysis for Signal and Power Integrity Improvement","authors":"Uichan Kim;Atom O. Watanabe;Woo-Su Kim;Youngwoo Kim","doi":"10.1109/ACCESS.2026.3660678","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3660678","url":null,"abstract":"This article proposes dispersion analysis for an efficient design of electromagnetic bandgap (EBG) structures in glass interposers to suppress power/ground noise coupling. The proposed method considered design parameters such as patch, through glass via (TGV), and defected structure which can be efficiently realized in glass interposers without additional fabrication steps. The unit cell of the EBG structure is modeled into transmission lines for an efficient dispersion analysis to estimate the power/ground noise suppression band. Impacts of capacitance associated with the patch, inductance of TGVs, and defected structure on the noise suppression bands are characterized for an efficient bandgap design. Various test vehicles are fabricated and measured to validate the proposed design methodology based on the dispersion analysis. The measured results showed good correlation with estimated noise suppression bandgaps which verifies the proposed design methodology based on dispersion analysis. The proposed method is further applied to design an EBG structure including multiple noise suppression bands for broadband noise suppression. The proposed design is validated using the 3-dimensional electromagnetic simulation. Effectiveness of the proposed EBG structure in the glass interposer on signal and power integrity is verified by analyzing noise propagation in the power delivery network and coupling to the TGV channel.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"22533-22544"},"PeriodicalIF":3.6,"publicationDate":"2026-02-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11386810","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175633","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：