Pub Date : 2025-12-30DOI: 10.1109/ACCESS.2025.3649761
Waseem Akram;Muhayy Ud Din;Tarek Taha;Irfan Hussain
Thruster allocation is critical for the reliable operation of underwater vehicles, particularly under actuator degradation, power limitations, and thermal stress. Existing methods, such as pseudo-inverse or standard quadratic programming (QP)-based approaches, mainly minimize allocation error or energy consumption but often overlook real-time degradation and resource constraints. In this paper, we propose an adaptive fault-tolerant thrust allocation framework integrated with a PID plus Sliding Mode Control (PID+SMC) law for robust trajectory tracking. The approach leverages convex optimization to simultaneously enforce: 1) residual-driven health adaptation that down-weights degraded thrusters online; 2) power-aware allocation ensuring operation within a global energy budget; and 3) thermal-aware constraints that actively prevent overheating. A lightweight residual filter continuously updates thruster health indices, enabling rapid reallocation under faults and efficiency loss. Simulation results across nominal, power-limited, thermal-limited, faulted, and combined scenarios show that the proposed method reduces trajectory tracking error by up to 4.3% and completely eliminates power and thermal violations compared to conventional baselines. This unified framework establishes a foundation for real-time, safety-aware thruster management in marine robotics.
{"title":"Adaptive Fault-Tolerant Thrust Allocation for Underwater Vehicles With Resource Constraints","authors":"Waseem Akram;Muhayy Ud Din;Tarek Taha;Irfan Hussain","doi":"10.1109/ACCESS.2025.3649761","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3649761","url":null,"abstract":"Thruster allocation is critical for the reliable operation of underwater vehicles, particularly under actuator degradation, power limitations, and thermal stress. Existing methods, such as pseudo-inverse or standard quadratic programming (QP)-based approaches, mainly minimize allocation error or energy consumption but often overlook real-time degradation and resource constraints. In this paper, we propose an adaptive fault-tolerant thrust allocation framework integrated with a PID plus Sliding Mode Control (PID+SMC) law for robust trajectory tracking. The approach leverages convex optimization to simultaneously enforce: 1) residual-driven health adaptation that down-weights degraded thrusters online; 2) power-aware allocation ensuring operation within a global energy budget; and 3) thermal-aware constraints that actively prevent overheating. A lightweight residual filter continuously updates thruster health indices, enabling rapid reallocation under faults and efficiency loss. Simulation results across nominal, power-limited, thermal-limited, faulted, and combined scenarios show that the proposed method reduces trajectory tracking error by up to 4.3% and completely eliminates power and thermal violations compared to conventional baselines. This unified framework establishes a foundation for real-time, safety-aware thruster management in marine robotics.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"1341-1357"},"PeriodicalIF":3.6,"publicationDate":"2025-12-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11318569","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145898248","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}
Parallel back-to-back converters are highly demanded in many high-power applications such as adjustable speed drive (ASD) systems, which reduce harmonics and improve the power factor and reliability compared to single two-level converters. It is evident that common-mode voltage (CMV) is the root cause of many challenges in ASD systems, such as shaft voltage and bearing damage, which may reduce equipment lifespan. On the other hand, Zero Sequence Circulating Current (ZSCC) leads to an additional current of switches which increases power loss and decreases the current capacity of converters. Simultaneous reduction of these two critical issues has to be considered in any switching strategy. In this regard, this paper presents a switching strategy based on a modified three-level space vector modulation scheme, which completely eliminates the common-mode voltage (CMV = 0 V). Moreover, the proposed switching sequence keeps the ZSCC within a low-amplitude and fully symmetric ripple, ensuring controlled circulating-current behavior without requiring any additional hardware. The method also generates a three-level line voltage and achieves an input-current THD of 3.92%. The simulation and experimental results confirm the effectiveness of the proposed approach.
{"title":"Common-Mode Voltage Elimination and Zero-Sequence Circulating Current Reduction of Parallel Back-to-Back Converters","authors":"Reza Farajpour;Mojtaba Sarparast;Jafar Adabi;Mohammad Rezanejad;Edris Pouresmaeil","doi":"10.1109/ACCESS.2025.3649536","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3649536","url":null,"abstract":"Parallel back-to-back converters are highly demanded in many high-power applications such as adjustable speed drive (ASD) systems, which reduce harmonics and improve the power factor and reliability compared to single two-level converters. It is evident that common-mode voltage (CMV) is the root cause of many challenges in ASD systems, such as shaft voltage and bearing damage, which may reduce equipment lifespan. On the other hand, Zero Sequence Circulating Current (ZSCC) leads to an additional current of switches which increases power loss and decreases the current capacity of converters. Simultaneous reduction of these two critical issues has to be considered in any switching strategy. In this regard, this paper presents a switching strategy based on a modified three-level space vector modulation scheme, which completely eliminates the common-mode voltage (CMV = 0 V). Moreover, the proposed switching sequence keeps the ZSCC within a low-amplitude and fully symmetric ripple, ensuring controlled circulating-current behavior without requiring any additional hardware. The method also generates a three-level line voltage and achieves an input-current THD of 3.92%. The simulation and experimental results confirm the effectiveness of the proposed approach.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"1256-1268"},"PeriodicalIF":3.6,"publicationDate":"2025-12-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11318331","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145898250","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 : 2025-12-30DOI: 10.1109/ACCESS.2025.3649668
Yousef Nejatbakhsh;Malihe Aliasgari
Accurate stock market prediction remains a critical yet challenging task due to the highly non-linear, volatile, and sentiment-driven nature of financial markets. In this paper, we present a hybrid deep learning framework that integrates long-short-term memory (LSTM) networks with Transformer-based attention mechanisms, sentiment analysis from financial news, and a privacy-preserving Federated Learning (FL) strategy. First, we benchmark traditional forecasting approaches, including ARIMA, SARIMAX, Prophet, Random Forest, and Support Vector Regression, against the baseline LSTM models. Our results show that LSTMs consistently outperform conventional methods in capturing temporal dependencies. To further enhance predictive accuracy, we incorporate Transformer attention to improve long-range dependency modeling and apply sentiment analysis using FinBERT-tone to embed market sentiment signals into the model. Finally, we simulate a Federated Learning environment, enabling decentralized model training without sharing raw financial data, thus addressing privacy concerns in the financial domain. Experimental results in ten major technology companies (Tesla, Apple, Amazon, Microsoft, Google, etc.) demonstrate that our hybrid model achieves superior short-term forecasting performance, with an average $R^{2}$ variance score of 0.91 across ten major technology companies and a trend precision of $65.36~%$ , demonstrating strong prediction performance for short-term stock forecasting. These findings highlight the potential of combining deep sequential models, attention mechanisms, and privacy-sensitive training strategies for robust and secure stock market forecasting.
{"title":"Enhancing Stock Market Prediction With Hybrid Deep Learning: Integrating LSTM, Transformer Attention, Federated Learning, and Sentiment Analysis","authors":"Yousef Nejatbakhsh;Malihe Aliasgari","doi":"10.1109/ACCESS.2025.3649668","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3649668","url":null,"abstract":"Accurate stock market prediction remains a critical yet challenging task due to the highly non-linear, volatile, and sentiment-driven nature of financial markets. In this paper, we present a hybrid deep learning framework that integrates long-short-term memory (LSTM) networks with Transformer-based attention mechanisms, sentiment analysis from financial news, and a privacy-preserving Federated Learning (FL) strategy. First, we benchmark traditional forecasting approaches, including ARIMA, SARIMAX, Prophet, Random Forest, and Support Vector Regression, against the baseline LSTM models. Our results show that LSTMs consistently outperform conventional methods in capturing temporal dependencies. To further enhance predictive accuracy, we incorporate Transformer attention to improve long-range dependency modeling and apply sentiment analysis using FinBERT-tone to embed market sentiment signals into the model. Finally, we simulate a Federated Learning environment, enabling decentralized model training without sharing raw financial data, thus addressing privacy concerns in the financial domain. Experimental results in ten major technology companies (Tesla, Apple, Amazon, Microsoft, Google, etc.) demonstrate that our hybrid model achieves superior short-term forecasting performance, with an average <inline-formula> <tex-math>$R^{2}$ </tex-math></inline-formula> variance score of 0.91 across ten major technology companies and a trend precision of <inline-formula> <tex-math>$65.36~%$ </tex-math></inline-formula>, demonstrating strong prediction performance for short-term stock forecasting. These findings highlight the potential of combining deep sequential models, attention mechanisms, and privacy-sensitive training strategies for robust and secure stock market forecasting.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"3926-3942"},"PeriodicalIF":3.6,"publicationDate":"2025-12-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11318561","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145929449","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 : 2025-12-30DOI: 10.1109/ACCESS.2025.3649751
Jeongwoo Son;Chansu Kim;Sang Hoon Kang
This paper proposes an electrical contact-based autonomous charging station for uncrewed aerial vehicles (UAVs) that reliably initiates charging regardless of landing position and orientation inaccuracies. Unlike existing UAV charging methods – which may suffer from efficiency losses due to wireless power transfer or require mechanical actuators, specially shaped structures, or diode bridges – the proposed autonomous charging station uses modular units with Hall-effect sensors to detect a magnet mounted on the UAV’s positive charging electrode. Thus, the proposed charging station was designed to allow direct electrical contact without rectifier diodes or actuators, reducing unnecessary losses. Across all 832 possible landing poses of the UAV, the power transfer efficiency exceeded 98.34% – surpassing the 91.02% reported in prior work; in outdoor repeated-flight tests, charging initiated and succeeded in all trials (30/30) with randomized landing positions and orientations. Preliminary field trials at a 765-kV substation demonstrated feasibility under elevated electromagnetic interference. These results highlight the robustness of the proposed system to substantial landing inaccuracies, providing a strong foundation for prolonged, unattended UAV missions in demanding real-world environments.
{"title":"Robust and Efficient Autonomous Charging Station for Uncrewed Aerial Vehicles Under Large Landing Inaccuracies","authors":"Jeongwoo Son;Chansu Kim;Sang Hoon Kang","doi":"10.1109/ACCESS.2025.3649751","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3649751","url":null,"abstract":"This paper proposes an electrical contact-based autonomous charging station for uncrewed aerial vehicles (UAVs) that reliably initiates charging regardless of landing position and orientation inaccuracies. Unlike existing UAV charging methods – which may suffer from efficiency losses due to wireless power transfer or require mechanical actuators, specially shaped structures, or diode bridges – the proposed autonomous charging station uses modular units with Hall-effect sensors to detect a magnet mounted on the UAV’s positive charging electrode. Thus, the proposed charging station was designed to allow direct electrical contact without rectifier diodes or actuators, reducing unnecessary losses. Across all 832 possible landing poses of the UAV, the power transfer efficiency exceeded 98.34% – surpassing the 91.02% reported in prior work; in outdoor repeated-flight tests, charging initiated and succeeded in all trials (30/30) with randomized landing positions and orientations. Preliminary field trials at a 765-kV substation demonstrated feasibility under elevated electromagnetic interference. These results highlight the robustness of the proposed system to substantial landing inaccuracies, providing a strong foundation for prolonged, unattended UAV missions in demanding real-world environments.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"4027-4037"},"PeriodicalIF":3.6,"publicationDate":"2025-12-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11318570","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145929348","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}
The inherent variability in human performance introduces stochastic perturbations into manufacturing environments, undermining the seamless coordination required for effective human-robot collaboration (HRC) systems. While human cognitive flexibility enhances adaptability, it simultaneously acts as a source of operational uncertainty, complicating the modeling and optimization of integrated robotic systems. Given these challenges, there is an urgent need to substantially expand the adaptability of robotic systems through real-time detection, algorithmic analysis and dynamic behavioral adjustments in response to human performance fluctuations. The systematic development of such systems capable of precisely detecting task-specific variations, analyzing them via advanced AI algorithms and adapting their behavior accordingly remains a critical focus of contemporary research. To evaluate progress in this domain, this study conducts a systematic literature review, synthesizing advancements across 124 publications and identifying underexplored research frontiers. The findings reveal a persistent misalignment between current technical capabilities and the requirements of adaptive collaboration in dynamic industrial environments. Key gaps include the absence of explainable AI frameworks for transparent decision-making, limited generalizability of adaptive control architectures and a lack of proactive strategies that anticipate rather than merely react to performance deviations.
{"title":"Adaptive Robotic Behavior in Industrial Human–Robot Collaboration: A Systematic Review of Taxonomies, Enabling Mechanisms, and Research Frontiers","authors":"Bsher Karbouj;Rajwinder Garha;Konstantin KeßLer;Jörg Krüger","doi":"10.1109/ACCESS.2025.3649702","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3649702","url":null,"abstract":"The inherent variability in human performance introduces stochastic perturbations into manufacturing environments, undermining the seamless coordination required for effective human-robot collaboration (HRC) systems. While human cognitive flexibility enhances adaptability, it simultaneously acts as a source of operational uncertainty, complicating the modeling and optimization of integrated robotic systems. Given these challenges, there is an urgent need to substantially expand the adaptability of robotic systems through real-time detection, algorithmic analysis and dynamic behavioral adjustments in response to human performance fluctuations. The systematic development of such systems capable of precisely detecting task-specific variations, analyzing them via advanced AI algorithms and adapting their behavior accordingly remains a critical focus of contemporary research. To evaluate progress in this domain, this study conducts a systematic literature review, synthesizing advancements across 124 publications and identifying underexplored research frontiers. The findings reveal a persistent misalignment between current technical capabilities and the requirements of adaptive collaboration in dynamic industrial environments. Key gaps include the absence of explainable AI frameworks for transparent decision-making, limited generalizability of adaptive control architectures and a lack of proactive strategies that anticipate rather than merely react to performance deviations.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"1398-1422"},"PeriodicalIF":3.6,"publicationDate":"2025-12-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11318881","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145898239","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 : 2025-12-30DOI: 10.1109/ACCESS.2025.3649485
Jeong Hyeon Park;Hong Seong Park
The security of Robot Operating System 2 (ROS 2) is crucial for ensuring the safety of individual robotic systems and the reliability of the environments in which they operate. Although Secure ROS 2 (SROS2) enhances security via Data Distribution Service (DDS) Security for authentication, access control, and data encryption, it has several limitations. The complexity of managing security artifacts and the degradation of communication performance are two primary concerns. To address these challenges, we propose the ROS 2 Security (ROS2Sec) module, which enhances ROS 2 security while minimizing communication performance degradation. ROS2Sec introduces centralized authentication management and group-based access control, thus reducing the number of security artifacts from seven for SROS2 to only three, thereby simplifying management. Additionally, ROS2Sec employs the Advanced Encryption Standard in Galois/Counter Mode at the ROS 2 message level to optimize data confidentiality while minimizing overhead. Experimental results demonstrate that the proposed ROS2Sec prevents unauthorized access by malicious ROS 2 nodes, reduces the mean communication latency by approximately 9% compared with that of SROS2, and maintains stable performance even as the number of subscribers increases. These findings confirm that ROS2Sec balances security and communication performance and is a practical solution for secure and efficient ROS 2-based robotic systems.
{"title":"Design and Implementation of ROS2 Security Module for Performance and Security Harmonization","authors":"Jeong Hyeon Park;Hong Seong Park","doi":"10.1109/ACCESS.2025.3649485","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3649485","url":null,"abstract":"The security of Robot Operating System 2 (ROS 2) is crucial for ensuring the safety of individual robotic systems and the reliability of the environments in which they operate. Although Secure ROS 2 (SROS2) enhances security via Data Distribution Service (DDS) Security for authentication, access control, and data encryption, it has several limitations. The complexity of managing security artifacts and the degradation of communication performance are two primary concerns. To address these challenges, we propose the ROS 2 Security (ROS2Sec) module, which enhances ROS 2 security while minimizing communication performance degradation. ROS2Sec introduces centralized authentication management and group-based access control, thus reducing the number of security artifacts from seven for SROS2 to only three, thereby simplifying management. Additionally, ROS2Sec employs the Advanced Encryption Standard in Galois/Counter Mode at the ROS 2 message level to optimize data confidentiality while minimizing overhead. Experimental results demonstrate that the proposed ROS2Sec prevents unauthorized access by malicious ROS 2 nodes, reduces the mean communication latency by approximately 9% compared with that of SROS2, and maintains stable performance even as the number of subscribers increases. These findings confirm that ROS2Sec balances security and communication performance and is a practical solution for secure and efficient ROS 2-based robotic systems.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"1287-1296"},"PeriodicalIF":3.6,"publicationDate":"2025-12-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11318342","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145898254","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 : 2025-12-30DOI: 10.1109/ACCESS.2025.3649764
Mohammed A. Bou-Rabee;Fajer M. Alelaj;Hussain Al-Sairfi
The Arabian Peninsula is one of the regions in the world with the highest potential for solar energy development in the sense that the levels of solar irradiance are very high, making it an ideal site to install photovoltaic (PV) systems. However, this huge potential is seriously undermined by enduring environmental problems, especially the high levels of airborne particulate matter and anthropogenic pollutants that together reduce the intensity of the sun and contribute to the rapid soiling of PV modules. Though the current literature has largely concentrated on the quantification of the technical performance instability of PV systems caused by soiling phenomena, the associated extensive analysis in terms of quantifying the technical losses into quantifiable economic and operational effects on a regional level is strikingly lacking in the literature. This study builds a combined machine learning system in order to quantitatively measure the economic losses attributable to air pollution for utility-scale, grid-connected PV systems across the Gulf Cooperation Council (GCC) member states. We developed and strictly tested a Random Forest regression model with a remarkable coefficient of determination (R 2) of 98.24 that was trained on a large dataset of meteorological parameters, real-time air pollution data (PM2.52.5, PM1010, SO22, NO22, O33, CO) and the most important operational features of solar panels that occurred between 2018 and 2020. With the help of this validated predictive model, we conducted an advanced counterfactual analysis by modeling the potential power production under hypothetical conditions of clean atmosphere benchmarks. We show that air pollution is one of the contributors to the loss of 8.5% to 12.3% of annual energy production in the region; or more simply, it is a significant loss that can translate to huge financial fines, which can significantly affect the economics of projects. In addition, we propose a new data-driven predictive cleaning scheduling algorithm that proves capable of cutting operational expenditures (OPEX) by up to 25 percent relative to traditional calendar-driven cleaning schedules. The findings are empirically based, critical to renewable energy investors, utility grid operators, and policymakers, and they categorically highlight the significant economic necessity to set up wide-ranging air pollution reduction measures, even as operation and maintenance (O&M) protocols in arid, dust-prone geographical settings run to their optimum.
{"title":"Quantifying the Economic Loss and Operational Implications of Air Pollution on Grid-Connected PV Systems in the Arabian Peninsula: A Machine Learning-Based Analysis","authors":"Mohammed A. Bou-Rabee;Fajer M. Alelaj;Hussain Al-Sairfi","doi":"10.1109/ACCESS.2025.3649764","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3649764","url":null,"abstract":"The Arabian Peninsula is one of the regions in the world with the highest potential for solar energy development in the sense that the levels of solar irradiance are very high, making it an ideal site to install photovoltaic (PV) systems. However, this huge potential is seriously undermined by enduring environmental problems, especially the high levels of airborne particulate matter and anthropogenic pollutants that together reduce the intensity of the sun and contribute to the rapid soiling of PV modules. Though the current literature has largely concentrated on the quantification of the technical performance instability of PV systems caused by soiling phenomena, the associated extensive analysis in terms of quantifying the technical losses into quantifiable economic and operational effects on a regional level is strikingly lacking in the literature. This study builds a combined machine learning system in order to quantitatively measure the economic losses attributable to air pollution for utility-scale, grid-connected PV systems across the Gulf Cooperation Council (GCC) member states. We developed and strictly tested a Random Forest regression model with a remarkable coefficient of determination (R 2) of 98.24 that was trained on a large dataset of meteorological parameters, real-time air pollution data (PM2.52.5, PM1010, SO22, NO22, O33, CO) and the most important operational features of solar panels that occurred between 2018 and 2020. With the help of this validated predictive model, we conducted an advanced counterfactual analysis by modeling the potential power production under hypothetical conditions of clean atmosphere benchmarks. We show that air pollution is one of the contributors to the loss of 8.5% to 12.3% of annual energy production in the region; or more simply, it is a significant loss that can translate to huge financial fines, which can significantly affect the economics of projects. In addition, we propose a new data-driven predictive cleaning scheduling algorithm that proves capable of cutting operational expenditures (OPEX) by up to 25 percent relative to traditional calendar-driven cleaning schedules. The findings are empirically based, critical to renewable energy investors, utility grid operators, and policymakers, and they categorically highlight the significant economic necessity to set up wide-ranging air pollution reduction measures, even as operation and maintenance (O&M) protocols in arid, dust-prone geographical settings run to their optimum.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"4180-4188"},"PeriodicalIF":3.6,"publicationDate":"2025-12-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11318572","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145929421","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 : 2025-12-29DOI: 10.1109/ACCESS.2025.3649006
Md Sadi Al Huda;Esrath Kanon;Md. Shahidul Khan Pappo;Md. Asraf Ali;Nasim Ahmed
Chronic Kidney Disease (CKD) is a serious health condition that progresses silently, often going undiagnosed until it reaches critical stages. Early detection is vital, but traditional diagnostic methods can be expensive, slow, and out of reach for many people. In this study, we introduce NefroAI, a real-time, explainable framework designed to predict CKD using both machine learning and deep learning techniques. We worked with three diverse datasets sourced from the UCI Machine Learning Repository and Kaggle to ensure strong generalization across populations. Our main contributions include applying thorough data preprocessing like missing value imputation, SMOTE for balancing, normalization, and using hybrid feature selection and hyperparameter tuning techniques. We also prioritized explainability, incorporating SHAP and LIME to make model decisions transparent, and deployed the model using Streamlit for easy real-time access with Real-time CKD prediction in Risk Indicators and a Download report. We implemented eight machine learning models and three deep learning models in this study. Among these models, Random Forest, K-Nearest Neighbors, and Support Vector Machine achieved 100.0% accuracy on Dataset A; SVM performed best on Dataset B with 98.7% accuracy; Random Forest, Decision Tree, K-Nearest Neighbors, AdaBoost, Logistic Regression, Naive Bayes, Support Vector Machine, XGBoost, ANN, LSTM, and RNN achieved 100% accuracy on Dataset C, showing strong consistency across datasets. While our results are promising, exploring collaborative learning across multiple data sources as a next step can enhance privacy and improve model generalizability.
{"title":"NefroAI: An Explainable and Real-Time Framework for Predicting Chronic Kidney Disease Using Diverse Machine Learning Models and Different Feature Selection Techniques","authors":"Md Sadi Al Huda;Esrath Kanon;Md. Shahidul Khan Pappo;Md. Asraf Ali;Nasim Ahmed","doi":"10.1109/ACCESS.2025.3649006","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3649006","url":null,"abstract":"Chronic Kidney Disease (CKD) is a serious health condition that progresses silently, often going undiagnosed until it reaches critical stages. Early detection is vital, but traditional diagnostic methods can be expensive, slow, and out of reach for many people. In this study, we introduce NefroAI, a real-time, explainable framework designed to predict CKD using both machine learning and deep learning techniques. We worked with three diverse datasets sourced from the UCI Machine Learning Repository and Kaggle to ensure strong generalization across populations. Our main contributions include applying thorough data preprocessing like missing value imputation, SMOTE for balancing, normalization, and using hybrid feature selection and hyperparameter tuning techniques. We also prioritized explainability, incorporating SHAP and LIME to make model decisions transparent, and deployed the model using Streamlit for easy real-time access with Real-time CKD prediction in Risk Indicators and a Download report. We implemented eight machine learning models and three deep learning models in this study. Among these models, Random Forest, K-Nearest Neighbors, and Support Vector Machine achieved 100.0% accuracy on Dataset A; SVM performed best on Dataset B with 98.7% accuracy; Random Forest, Decision Tree, K-Nearest Neighbors, AdaBoost, Logistic Regression, Naive Bayes, Support Vector Machine, XGBoost, ANN, LSTM, and RNN achieved 100% accuracy on Dataset C, showing strong consistency across datasets. While our results are promising, exploring collaborative learning across multiple data sources as a next step can enhance privacy and improve model generalizability.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"10939-10976"},"PeriodicalIF":3.6,"publicationDate":"2025-12-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11316603","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146026413","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 : 2025-12-29DOI: 10.1109/ACCESS.2025.3649109
Rahul Deva;Arvind Dagur
Chest radiographs are widely used to clinically diagnose thoracic diseases such as lung opacity, pneumonia, and coronavirus disease 2019 due to their cost-effectiveness and easy access. However, accurate interpretation remains challenging because of overlapping anatomical structures, low contrast, and subtle disease manifestations. Classical deep learning models are effective but often exhibit overfitting, and weak interpretability, which restricts their clinical applicability. This paper presents an attention-driven hybrid framework that integrates contrast-limited adaptive histogram equalization-based enhancement with a dual-backbone architecture combining a vision transformer and a residual network. The vision transformer captures global contextual dependencies, while the residual network extracts local discriminative features. The concatenated representations are classified using a multi-layer perceptron and optimized end-to-end with the AdamW optimizer and a step learning rate scheduler. To improve transparency, gradient-weighted class activation mapping is used to highlight disease-relevant regions in chest radiographs. Experimental evaluation highlights that the proposed framework achieves 98.54% classification accuracy, outperforming state-of-the-art models, including EfficientNet 96.20%, DenseNet 97.50%, and a fine-tuned vision transformer 95.79%. To ensure generalizability, cross-dataset validation was conducted and trained on COVID-19 Radiography Dataset. The same model was tested on an independent COVID–Pneumonia–Normal chest X-ray dataset which achieved an accuracy of 94.07% and demonstrated good performance over heterogeneous imaging sources. These findings confirm the proposed framework’s robustness, interpretability, and suitability for real-time clinical decision support in both pandemic and routine diagnostic settings.
{"title":"ViT-ResNet Fusion: An Explainable Hybrid Framework for High-Accuracy Multiclass Lung Disease Classification in Chest X-Rays","authors":"Rahul Deva;Arvind Dagur","doi":"10.1109/ACCESS.2025.3649109","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3649109","url":null,"abstract":"Chest radiographs are widely used to clinically diagnose thoracic diseases such as lung opacity, pneumonia, and coronavirus disease 2019 due to their cost-effectiveness and easy access. However, accurate interpretation remains challenging because of overlapping anatomical structures, low contrast, and subtle disease manifestations. Classical deep learning models are effective but often exhibit overfitting, and weak interpretability, which restricts their clinical applicability. This paper presents an attention-driven hybrid framework that integrates contrast-limited adaptive histogram equalization-based enhancement with a dual-backbone architecture combining a vision transformer and a residual network. The vision transformer captures global contextual dependencies, while the residual network extracts local discriminative features. The concatenated representations are classified using a multi-layer perceptron and optimized end-to-end with the AdamW optimizer and a step learning rate scheduler. To improve transparency, gradient-weighted class activation mapping is used to highlight disease-relevant regions in chest radiographs. Experimental evaluation highlights that the proposed framework achieves 98.54% classification accuracy, outperforming state-of-the-art models, including EfficientNet 96.20%, DenseNet 97.50%, and a fine-tuned vision transformer 95.79%. To ensure generalizability, cross-dataset validation was conducted and trained on COVID-19 Radiography Dataset. The same model was tested on an independent COVID–Pneumonia–Normal chest X-ray dataset which achieved an accuracy of 94.07% and demonstrated good performance over heterogeneous imaging sources. These findings confirm the proposed framework’s robustness, interpretability, and suitability for real-time clinical decision support in both pandemic and routine diagnostic settings.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"1358-1372"},"PeriodicalIF":3.6,"publicationDate":"2025-12-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11316598","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145898183","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 : 2025-12-29DOI: 10.1109/ACCESS.2025.3649265
Haotian Ji;Israel Mendonça;Masayoshi Aritsugi
In the field of Computer Vision (CV), assistive navigation systems for visually impaired individuals have garnered significant attention in recent years. However, most existing solutions rely on wearable devices or mobile platforms, which often face limitations in cost, deployment, and robustness in complex outdoor environments. This paper proposes a practical approach to public space recognition for the visually impaired. The proposed recognition method can be widely applied to existing public video surveillance systems. We created a specialized image dataset tailored for recognition by visually impaired individuals. At the same time, for recognition tasks involving nighttime environments and partially occluded targets, we propose a composite framework that integrates various grouped convolutions and image enhancement networks. Compared to the original baseline detection models, our models outperform on our created dataset by 1.5% average precision (e.g., from 94.1% to 95.6% for YOLOv8x), while reducing parameters by up to 35.7% (e.g., from 56.9M to 36.6M for YOLOv11x). Furthermore, our models also achieve over 0.8% AP and a parameter reduction exceeding 10% compared to the original baseline models on ExDARK dataset.
{"title":"Multi-Scene Dataset and Object Detector for Outside Blind Individual Identification","authors":"Haotian Ji;Israel Mendonça;Masayoshi Aritsugi","doi":"10.1109/ACCESS.2025.3649265","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3649265","url":null,"abstract":"In the field of Computer Vision (CV), assistive navigation systems for visually impaired individuals have garnered significant attention in recent years. However, most existing solutions rely on wearable devices or mobile platforms, which often face limitations in cost, deployment, and robustness in complex outdoor environments. This paper proposes a practical approach to public space recognition for the visually impaired. The proposed recognition method can be widely applied to existing public video surveillance systems. We created a specialized image dataset tailored for recognition by visually impaired individuals. At the same time, for recognition tasks involving nighttime environments and partially occluded targets, we propose a composite framework that integrates various grouped convolutions and image enhancement networks. Compared to the original baseline detection models, our models outperform on our created dataset by 1.5% average precision (e.g., from 94.1% to 95.6% for YOLOv8x), while reducing parameters by up to 35.7% (e.g., from 56.9M to 36.6M for YOLOv11x). Furthermore, our models also achieve over 0.8% AP and a parameter reduction exceeding 10% compared to the original baseline models on ExDARK dataset.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"1423-1438"},"PeriodicalIF":3.6,"publicationDate":"2025-12-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11317963","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145898199","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：