Pub Date : 2025-02-27DOI: 10.1109/ACCESS.2025.3546301
Nemanja Filipović;Dragomir El Mezeni;Vladimir L. Petrović
5G new radio supports a variety of new technologies and services, demanding significant improvements in radio access network (RAN) latency, throughput, and flexibility. Disaggregation addresses these challenges by splitting the RAN into network units-central (CU), distributing (DU), and radio (RU), enabling data processing virtualization and implementation on off-the-shelf hardware. However, implementing upper physical layer (PHY) processing on off-the-shelf hardware alone might cause inefficient usage of the server processors. Therefore, acceleration is often needed to offload heavy processing, focusing on low-density parity-check (LDPC) codec as the most compute-intensive task in the PHY. Additionally, LDPC coding is tightly coupled with rate-matching. This paper presents novel hardware architectures of rate-matcher and rate-dematcher, targeting field programmable gate array (FPGA) RAN accelerators. The presented solution’s approach to memory organization allows highly parallel operation with efficient hardware resource usage. The architecture is flexible, enabling a selection of various parallelism levels for instant integration with other PHY components, and achieves a throughput of up to 150 Gbps for rate-matching, and up to 35 Gbps for rate-dematching. Both components have been integrated into a Peripheral Component Interconnect Express (PCIe) FPGA acceleration card with an LDPC encoder and decoder. The accelerator performance has been evaluated against OpenAirInterface PHY software. By measuring the acceleration impact on the processor load, it has been shown that with the proposed components, the acceleration efficiency can be increased by an order of magnitude compared to the LDPC-only solution.
{"title":"Scalable 5G NR Rate-Matcher and Rate-Dematcher for Efficient Use in FPGA Accelerators","authors":"Nemanja Filipović;Dragomir El Mezeni;Vladimir L. Petrović","doi":"10.1109/ACCESS.2025.3546301","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3546301","url":null,"abstract":"5G new radio supports a variety of new technologies and services, demanding significant improvements in radio access network (RAN) latency, throughput, and flexibility. Disaggregation addresses these challenges by splitting the RAN into network units-central (CU), distributing (DU), and radio (RU), enabling data processing virtualization and implementation on off-the-shelf hardware. However, implementing upper physical layer (PHY) processing on off-the-shelf hardware alone might cause inefficient usage of the server processors. Therefore, acceleration is often needed to offload heavy processing, focusing on low-density parity-check (LDPC) codec as the most compute-intensive task in the PHY. Additionally, LDPC coding is tightly coupled with rate-matching. This paper presents novel hardware architectures of rate-matcher and rate-dematcher, targeting field programmable gate array (FPGA) RAN accelerators. The presented solution’s approach to memory organization allows highly parallel operation with efficient hardware resource usage. The architecture is flexible, enabling a selection of various parallelism levels for instant integration with other PHY components, and achieves a throughput of up to 150 Gbps for rate-matching, and up to 35 Gbps for rate-dematching. Both components have been integrated into a Peripheral Component Interconnect Express (PCIe) FPGA acceleration card with an LDPC encoder and decoder. The accelerator performance has been evaluated against OpenAirInterface PHY software. By measuring the acceleration impact on the processor load, it has been shown that with the proposed components, the acceleration efficiency can be increased by an order of magnitude compared to the LDPC-only solution.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"38515-38532"},"PeriodicalIF":3.4,"publicationDate":"2025-02-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10906572","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143553277","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 : 2025-02-26DOI: 10.1109/ACCESS.2025.3546033
Anna Kalenkova;Lewis Mitchell;Matthew Roughan
Process mining is a well-established discipline of data analysis focused on the discovery of process models from information systems’ event logs. Recently, an emerging subarea of process mining, known as stochastic process discovery, has started to evolve. Stochastic process discovery considers frequencies of events in the event data and allows for a more comprehensive analysis. In particular, when the durations of activities are presented in the event log, performance characteristics of the discovered stochastic models can be analyzed, e.g., the overall process execution time can be estimated. Existing performance analysis techniques usually discover stochastic process models from event data, and then simulate these models to evaluate their execution times. These methods rely on empirical approaches. This paper proposes analytical techniques for performance analysis that allow for the derivation of statistical characteristics of the overall processes’ execution times in the presence of arbitrary time distributions of events modeled by semi-Markov processes. The proposed methods include express analysis, focused on the mean execution time estimation, and full analysis techniques that build probability density functions (PDFs) of process execution times in both continuous and discrete forms. These methods are implemented and tested on real-world event data, demonstrating their potential for what-if analysis by providing solutions without resorting to simulation. Specifically, we demonstrated that the discrete approach is more time-efficient for small duration support sizes compared to the simulation technique. Furthermore, we showed that the continuous approach, with PDFs represented as Mixtures of Gaussian Models (GMMs), facilitates the discovery of more compact and interpretable models.
{"title":"Performance Analysis: Discovering Semi-Markov Models From Event Logs","authors":"Anna Kalenkova;Lewis Mitchell;Matthew Roughan","doi":"10.1109/ACCESS.2025.3546033","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3546033","url":null,"abstract":"Process mining is a well-established discipline of data analysis focused on the discovery of process models from information systems’ event logs. Recently, an emerging subarea of process mining, known as stochastic process discovery, has started to evolve. Stochastic process discovery considers frequencies of events in the event data and allows for a more comprehensive analysis. In particular, when the durations of activities are presented in the event log, performance characteristics of the discovered stochastic models can be analyzed, e.g., the overall process execution time can be estimated. Existing performance analysis techniques usually discover stochastic process models from event data, and then simulate these models to evaluate their execution times. These methods rely on empirical approaches. This paper proposes analytical techniques for performance analysis that allow for the derivation of statistical characteristics of the overall processes’ execution times in the presence of arbitrary time distributions of events modeled by semi-Markov processes. The proposed methods include express analysis, focused on the mean execution time estimation, and full analysis techniques that build probability density functions (PDFs) of process execution times in both continuous and discrete forms. These methods are implemented and tested on real-world event data, demonstrating their potential for what-if analysis by providing solutions without resorting to simulation. Specifically, we demonstrated that the discrete approach is more time-efficient for small duration support sizes compared to the simulation technique. Furthermore, we showed that the continuous approach, with PDFs represented as Mixtures of Gaussian Models (GMMs), facilitates the discovery of more compact and interpretable models.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"38035-38053"},"PeriodicalIF":3.4,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10904251","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143553240","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 : 2025-02-26DOI: 10.1109/ACCESS.2025.3545587
Yihan Zhou;Xiaokang Yu;Zixi Huang;Feierdun Palati;Zeyu Zhao;Zihan He;Yuan Feng;Yuxi Luo
In recent years, substantial research has focused on automated systems for assessing depression levels using different types of data, such as audio and visual inputs. However, signals recorded from individuals with depression can be influenced by external factors, such as the recording equipment and environment, making it essential to create a system that is resilient to these interferences to maintain accuracy. This study introduces a fused-attention model for evaluating depression severity using enhanced multi-modal data inputs. Applying several pre-trained advanced models, this article incorporates audiovisual sequences with augmentation. The framework includes two novel components, which we term as the FIE and VIE blocks, for extracting detailed facial and vocal features. The FIE block utilizes ResNet-18 to enhance the feature representation of video frames and integrates two types of attention mechanisms to capture spatial-temporal patterns. Meanwhile, the VIE block processes the Mel spectrogram of the audio signal, followed by an optimized Swin transformer block to extract auditory features. The model demonstrates strong performance, accurately identifying depression severity in 3-second audiovisual sequences with an 81.4% accuracy rate on the AVEC2014 dataset, and achieves a Kappa score of 0.731 and an MF1 index of 0.798. Furthermore, it shows high resilience to noise, underscoring its ability to mitigate the effects of recording equipment and environmental conditions in depression level estimation.
{"title":"Multi-Modal Fused-Attention Network for Depression Level Recognition Based on Enhanced Audiovisual Cues","authors":"Yihan Zhou;Xiaokang Yu;Zixi Huang;Feierdun Palati;Zeyu Zhao;Zihan He;Yuan Feng;Yuxi Luo","doi":"10.1109/ACCESS.2025.3545587","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3545587","url":null,"abstract":"In recent years, substantial research has focused on automated systems for assessing depression levels using different types of data, such as audio and visual inputs. However, signals recorded from individuals with depression can be influenced by external factors, such as the recording equipment and environment, making it essential to create a system that is resilient to these interferences to maintain accuracy. This study introduces a fused-attention model for evaluating depression severity using enhanced multi-modal data inputs. Applying several pre-trained advanced models, this article incorporates audiovisual sequences with augmentation. The framework includes two novel components, which we term as the FIE and VIE blocks, for extracting detailed facial and vocal features. The FIE block utilizes ResNet-18 to enhance the feature representation of video frames and integrates two types of attention mechanisms to capture spatial-temporal patterns. Meanwhile, the VIE block processes the Mel spectrogram of the audio signal, followed by an optimized Swin transformer block to extract auditory features. The model demonstrates strong performance, accurately identifying depression severity in 3-second audiovisual sequences with an 81.4% accuracy rate on the AVEC2014 dataset, and achieves a Kappa score of 0.731 and an MF1 index of 0.798. Furthermore, it shows high resilience to noise, underscoring its ability to mitigate the effects of recording equipment and environmental conditions in depression level estimation.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"37913-37923"},"PeriodicalIF":3.4,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10904116","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143553062","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 : 2025-02-26DOI: 10.1109/ACCESS.2025.3546156
Victor Takashi Hayashi;Wilson Vicente Ruggiero
The globalization of the hardware supply chain reduces costs but increases security challenges with the potential insertion of hardware trojans by third parties. Traditional detection methods face scalability limitations by relying solely on simple examples (e.g., AES). Although open-source hardware promotes transparency, it does not guarantee security. In this research, Natural Language Processing (NLP) and Machine Learning (ML) techniques were applied to identify hardware trojans in complex open hardware designs (e.g., RISC-V, MIPS). Using data from existing benchmarks (ISCAS85-89, TrustHub) and synthetic data generated with Large Language Models (LLM), a dataset of 3,808 instances was used in this research. The approach using TF-IDF and Decision Tree (DT) achieved 97.26%, surpassing the state of the art. The use of LLMs with prompt optimization achieved a recall of 99%, minimizing false negatives. A novel framework integrating NLP, ML, and LLMs was developed to enhance the security of open-source hardware.
{"title":"Hardware Trojan Detection in Open-Source Hardware Designs Using Machine Learning","authors":"Victor Takashi Hayashi;Wilson Vicente Ruggiero","doi":"10.1109/ACCESS.2025.3546156","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3546156","url":null,"abstract":"The globalization of the hardware supply chain reduces costs but increases security challenges with the potential insertion of hardware trojans by third parties. Traditional detection methods face scalability limitations by relying solely on simple examples (e.g., AES). Although open-source hardware promotes transparency, it does not guarantee security. In this research, Natural Language Processing (NLP) and Machine Learning (ML) techniques were applied to identify hardware trojans in complex open hardware designs (e.g., RISC-V, MIPS). Using data from existing benchmarks (ISCAS85-89, TrustHub) and synthetic data generated with Large Language Models (LLM), a dataset of 3,808 instances was used in this research. The approach using TF-IDF and Decision Tree (DT) achieved 97.26%, surpassing the state of the art. The use of LLMs with prompt optimization achieved a recall of 99%, minimizing false negatives. A novel framework integrating NLP, ML, and LLMs was developed to enhance the security of open-source hardware.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"37771-37788"},"PeriodicalIF":3.4,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10904479","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143553244","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}
This study focused on security robots as a substitute for security guards. For patrol tasks, the operation of body scans using teleoperated manipulators equipped with metal detectors has been considered to determine whether individuals are carrying hazardous items. When a handheld metal detector is waved over a person to be inspected, it must follow the subject’s clothing from a certain distance. In this study, a method for acquiring an object’s shape in-situ and following its surface without contact is proposed. The resulting scheme captures the surface information of an object using Light Detection and Ranging (LiDAR) immediately prior to operation. The distance between the hand and the object is then controlled using a point cloud and the operator’s control input. To confirm the effectiveness of the proposed method, subject experiments were conducted with three operators. As a result, the percentage of distance errors significantly increases within ±10 mm in the proposed method. The percentage of the distance error within ±10 mm in the experiment without control was 78.7%, and the percentage in the experiment with control was 99.8%. In addition, the time in which the proposed method completes the surface-following task is significantly decreased: The median time in the experiment without control was 63.1 s, while that in the experiment with control was 24.6 s. These results indicate that the distance can be easily maintained when an input device is pressed forward and swung left and right, reducing the operation time.
{"title":"Contactless Surface Following via Teleoperated Manipulator Based on Point Cloud","authors":"Teppei Tsujita;Shunsuke Matsushima;Tadamasa Kitahara;Satoko Abiko","doi":"10.1109/ACCESS.2025.3546027","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3546027","url":null,"abstract":"This study focused on security robots as a substitute for security guards. For patrol tasks, the operation of body scans using teleoperated manipulators equipped with metal detectors has been considered to determine whether individuals are carrying hazardous items. When a handheld metal detector is waved over a person to be inspected, it must follow the subject’s clothing from a certain distance. In this study, a method for acquiring an object’s shape in-situ and following its surface without contact is proposed. The resulting scheme captures the surface information of an object using Light Detection and Ranging (LiDAR) immediately prior to operation. The distance between the hand and the object is then controlled using a point cloud and the operator’s control input. To confirm the effectiveness of the proposed method, subject experiments were conducted with three operators. As a result, the percentage of distance errors significantly increases within ±10 mm in the proposed method. The percentage of the distance error within ±10 mm in the experiment without control was 78.7%, and the percentage in the experiment with control was 99.8%. In addition, the time in which the proposed method completes the surface-following task is significantly decreased: The median time in the experiment without control was 63.1 s, while that in the experiment with control was 24.6 s. These results indicate that the distance can be easily maintained when an input device is pressed forward and swung left and right, reducing the operation time.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"38685-38697"},"PeriodicalIF":3.4,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10904143","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143553117","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 : 2025-02-26DOI: 10.1109/ACCESS.2025.3545807
Katarzyna Heryan;Janusz Skrzat
This study addresses the analysis of morphological and topological relationships within the human renal arterial tree, which have significant implications for improving clinical practices, particularly in minimally invasive renal surgeries, where detailed information on vascular supply is necessary to preserve healthy parenchyma. Recent advances in imaging technologies, coupled with increased computational power, provide unprecedented detail in anatomical assessment and enable a comprehensive analysis. The presented investigation employed $33~mu $ -CT scans of corrosion cast speciments of renal arterial trees, which underwent systematic processing, including binary reconstruction, artifact removal, cavity filling, skeletonization, and conversion into directed graph representations. This approach facilitated the computation of geometrical parameters and enabled morphological and topological analyses. The results confirm the feasibility of using $mu $ -CT imaging to investigate both global population-level trends and intra-renal tree variations. Key findings reveal no strong individual predisposition for specific branching degrees, with trifurcations and quadfurcations being notable exceptions. Additionally, a significant inverse correlation between vessel length and diameter across generations was observed. The analysis of subtrees indicated which vessels supply specific renal segments, offering valuable information for preoperative planning, particularly in tumor surgeries. Key insights into vessel branching patterns highlight their relevance for renal surgeries. These findings have direct applications in enhancing algorithms for the reconstruction, segmentation, and visualization of preoperative CT scans by leveraging insights from $mu $ -CT based vascular analysis. Despite limitations such as datasize constraints and artifacts inherent to historical corrosion cast specimens, this study provides a proof-of-concept framework. The proposed methodology lays the foundation for large-scale investigations and the integration of $mu $ -CT-derived vascular patterns into patient-specific preoperative models, ultimately improving surgical navigation and patients’ outcomes. The suite of algorithms developed by us is easily scalable as more data becomes available and can be adapted for other tree-like structures, further enhancing their applicability.
{"title":"Morphological and Topological Analysis of the Human Renal Arterial Tree Using μ-CT Scans of Corrosion Cast Specimens","authors":"Katarzyna Heryan;Janusz Skrzat","doi":"10.1109/ACCESS.2025.3545807","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3545807","url":null,"abstract":"This study addresses the analysis of morphological and topological relationships within the human renal arterial tree, which have significant implications for improving clinical practices, particularly in minimally invasive renal surgeries, where detailed information on vascular supply is necessary to preserve healthy parenchyma. Recent advances in imaging technologies, coupled with increased computational power, provide unprecedented detail in anatomical assessment and enable a comprehensive analysis. The presented investigation employed <inline-formula> <tex-math>$33~mu $ </tex-math></inline-formula>-CT scans of corrosion cast speciments of renal arterial trees, which underwent systematic processing, including binary reconstruction, artifact removal, cavity filling, skeletonization, and conversion into directed graph representations. This approach facilitated the computation of geometrical parameters and enabled morphological and topological analyses. The results confirm the feasibility of using <inline-formula> <tex-math>$mu $ </tex-math></inline-formula>-CT imaging to investigate both global population-level trends and intra-renal tree variations. Key findings reveal no strong individual predisposition for specific branching degrees, with trifurcations and quadfurcations being notable exceptions. Additionally, a significant inverse correlation between vessel length and diameter across generations was observed. The analysis of subtrees indicated which vessels supply specific renal segments, offering valuable information for preoperative planning, particularly in tumor surgeries. Key insights into vessel branching patterns highlight their relevance for renal surgeries. These findings have direct applications in enhancing algorithms for the reconstruction, segmentation, and visualization of preoperative CT scans by leveraging insights from <inline-formula> <tex-math>$mu $ </tex-math></inline-formula>-CT based vascular analysis. Despite limitations such as datasize constraints and artifacts inherent to historical corrosion cast specimens, this study provides a proof-of-concept framework. The proposed methodology lays the foundation for large-scale investigations and the integration of <inline-formula> <tex-math>$mu $ </tex-math></inline-formula>-CT-derived vascular patterns into patient-specific preoperative models, ultimately improving surgical navigation and patients’ outcomes. The suite of algorithms developed by us is easily scalable as more data becomes available and can be adapted for other tree-like structures, further enhancing their applicability.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"38014-38034"},"PeriodicalIF":3.4,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10904240","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143553022","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 : 2025-02-26DOI: 10.1109/ACCESS.2025.3545441
Jinzhi Liu;Tianhao Zuo
In this study, we propose a novel Deeply Optimized Adaptive chaotic synchronization algorithm system (DOA), which adopts the ideas of genetic algorithm, Deep Image Prior (DIP) network, Deep Convolutional Generative Adversarial Network (DCGAN) network, and slide mode control algorithm. Traditional Deep Learning (DL)-based methods perform well in complex multi-parameter operations, but training on large datasets is typically a complicated, time-consuming, and high-cost process. Such methods are also difficult to adapt to dynamic parameter changes. The algorithmic network model in the proposed DOA reduces reliance on large datasets by learning the deep mining methods of the data characteristics in DIP, and can adjust system parameters adaptively, accurately, and quickly, providing high synchronization efficiency and excellent stability over various chaotic signals. By applying Lyapunov stability theory, the robustness and global stability of the model in dynamic systems are proven. This paper also uses an advanced Recurrent Neural Network (RNN)-based chaotic synchronization system as a benchmark. The simulation results show that, when compared to the Recurrent Neural Network based synchronization system, the DOA architecture has significant advantages in robustness, convergence, and training over noisy channels. Experiments show that under strong noise (AWGN variance = 2) and parameter mismatch (±20 percent drift), the synchronization error of DOA (<0.3)>1.5), and the training data volume is reduced by more than 30%. Simulation results show that, the DOA architecture has significant advantages in robustness, convergence, and training over noisy channels. The proposed DOA scheme improves the effect of chaotic synchronization and paves the way for the development of a new class of modulator schemes that meet the robustness, convergence, and training requirements for encrypted communication.
{"title":"Deep Adaptive Chaos Synchronization Based on Optimization Algorithm","authors":"Jinzhi Liu;Tianhao Zuo","doi":"10.1109/ACCESS.2025.3545441","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3545441","url":null,"abstract":"In this study, we propose a novel Deeply Optimized Adaptive chaotic synchronization algorithm system (DOA), which adopts the ideas of genetic algorithm, Deep Image Prior (DIP) network, Deep Convolutional Generative Adversarial Network (DCGAN) network, and slide mode control algorithm. Traditional Deep Learning (DL)-based methods perform well in complex multi-parameter operations, but training on large datasets is typically a complicated, time-consuming, and high-cost process. Such methods are also difficult to adapt to dynamic parameter changes. The algorithmic network model in the proposed DOA reduces reliance on large datasets by learning the deep mining methods of the data characteristics in DIP, and can adjust system parameters adaptively, accurately, and quickly, providing high synchronization efficiency and excellent stability over various chaotic signals. By applying Lyapunov stability theory, the robustness and global stability of the model in dynamic systems are proven. This paper also uses an advanced Recurrent Neural Network (RNN)-based chaotic synchronization system as a benchmark. The simulation results show that, when compared to the Recurrent Neural Network based synchronization system, the DOA architecture has significant advantages in robustness, convergence, and training over noisy channels. Experiments show that under strong noise (AWGN variance = 2) and parameter mismatch (±20 percent drift), the synchronization error of DOA (<0.3)>1.5), and the training data volume is reduced by more than 30%. Simulation results show that, the DOA architecture has significant advantages in robustness, convergence, and training over noisy channels. The proposed DOA scheme improves the effect of chaotic synchronization and paves the way for the development of a new class of modulator schemes that meet the robustness, convergence, and training requirements for encrypted communication.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"38671-38684"},"PeriodicalIF":3.4,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10904221","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143553166","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 : 2025-02-26DOI: 10.1109/ACCESS.2025.3546108
Gizem Sungu Terci;Betül Boz
The graph coloring problem is a well-known optimization challenge, particularly relevant in dynamic environments where the graph undergoes continuous changes over time. Evolutionary algorithms, known for their adaptability and effectiveness in handling NP-hard problems, are well-suited for tackling the issues related to coloring dynamic graphs. In this paper, we present a novel Similarity and Pool-Based Evolutionary Algorithm designed to address the graph coloring problem on dynamic graphs. Our approach employs a partition-based representation that adapts to dynamic graph changes while preserving valuable historical information. The algorithm introduces an innovative similarity and conflict-based crossover operator aimed at minimizing the number of colors used, alongside a local search method to enhance solution diversity. We evaluated the performance of the proposed algorithm against a well-known heuristic for the graph coloring problem and a genetic algorithm with a dynamic population across a diverse set of dynamic graphs. Experimental results demonstrate that our algorithm consistently outperforms these alternatives by reducing the number of colors required in the majority of test cases.
{"title":"Coloring Dynamic Graphs With a Similarity and Pool-Based Evolutionary Algorithm","authors":"Gizem Sungu Terci;Betül Boz","doi":"10.1109/ACCESS.2025.3546108","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3546108","url":null,"abstract":"The graph coloring problem is a well-known optimization challenge, particularly relevant in dynamic environments where the graph undergoes continuous changes over time. Evolutionary algorithms, known for their adaptability and effectiveness in handling NP-hard problems, are well-suited for tackling the issues related to coloring dynamic graphs. In this paper, we present a novel Similarity and Pool-Based Evolutionary Algorithm designed to address the graph coloring problem on dynamic graphs. Our approach employs a partition-based representation that adapts to dynamic graph changes while preserving valuable historical information. The algorithm introduces an innovative similarity and conflict-based crossover operator aimed at minimizing the number of colors used, alongside a local search method to enhance solution diversity. We evaluated the performance of the proposed algorithm against a well-known heuristic for the graph coloring problem and a genetic algorithm with a dynamic population across a diverse set of dynamic graphs. Experimental results demonstrate that our algorithm consistently outperforms these alternatives by reducing the number of colors required in the majority of test cases.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"38054-38075"},"PeriodicalIF":3.4,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10904463","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143553484","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 : 2025-02-26DOI: 10.1109/ACCESS.2025.3546017
R. Karthik;R. Aswin;K. S. Geetha;K. Suganthi
Bean rust and angular leaf spot pose significant challenges to bean cultivation, impacting yields. Prompt disease identification maximizes productivity, but traditional methods need specialized expertise. This research presents an explainable deep learning model that combines the Pyramid Vision Transformer (PVT) and Group Context Aware Depthwise Shuffle Network (GCADSN). The PVT effectively models long-range dependencies, identifying disease patterns across larger leaf areas, while the GCADSN focuses on capturing nuanced, context-specific features. This combined approach leads to a richer representation of the input image, resulting in improved disease classification. Model explainability is provided through GradCAM visualizations, highlighting the image regions crucial for the model’s predictions and enabling transparent, class-specific insights. The model’s performance was rigorously tested using the IBean dataset, a collection of images depicting various bean leaf diseases, including common rust, angular leaf spot, and healthy leaves. Our proposed network achieved a high accuracy of 97.66%, outperforming many current state-of-the-art deep learning models. Furthermore, it demonstrated strong performance across other key metrics, with an F1 score of 97.67%, precision of 97.83%, and recall of 97.67%. Importantly, the model’s computational efficiency makes it well-suited for practical application in real-world agricultural scenarios.
{"title":"An Explainable Deep Learning Network With Transformer and Custom CNN for Bean Leaf Disease Classification","authors":"R. Karthik;R. Aswin;K. S. Geetha;K. Suganthi","doi":"10.1109/ACCESS.2025.3546017","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3546017","url":null,"abstract":"Bean rust and angular leaf spot pose significant challenges to bean cultivation, impacting yields. Prompt disease identification maximizes productivity, but traditional methods need specialized expertise. This research presents an explainable deep learning model that combines the Pyramid Vision Transformer (PVT) and Group Context Aware Depthwise Shuffle Network (GCADSN). The PVT effectively models long-range dependencies, identifying disease patterns across larger leaf areas, while the GCADSN focuses on capturing nuanced, context-specific features. This combined approach leads to a richer representation of the input image, resulting in improved disease classification. Model explainability is provided through GradCAM visualizations, highlighting the image regions crucial for the model’s predictions and enabling transparent, class-specific insights. The model’s performance was rigorously tested using the IBean dataset, a collection of images depicting various bean leaf diseases, including common rust, angular leaf spot, and healthy leaves. Our proposed network achieved a high accuracy of 97.66%, outperforming many current state-of-the-art deep learning models. Furthermore, it demonstrated strong performance across other key metrics, with an F1 score of 97.67%, precision of 97.83%, and recall of 97.67%. Importantly, the model’s computational efficiency makes it well-suited for practical application in real-world agricultural scenarios.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"38562-38573"},"PeriodicalIF":3.4,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10904208","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143553066","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 : 2025-02-26DOI: 10.1109/ACCESS.2025.3546158
Shaoyu Yang;Masamichi Shimosaka
Kinodynamic motion planning plays a vital role in robotics, particularly in autonomous driving, where planned trajectories must satisfy kinematic and dynamic constraints while ensuring both safety and efficiency. Although existing kinodynamic RRT* algorithms achieve asymptotic optimality, their high computational cost often limits their practicality in high-dimensional or complex environments, such as autonomous driving scenarios. To enhance efficiency in such scenarios, motion templates with predefined action sequences have been proposed as a guiding strategy for planners. However, traditional fixed templates lack the flexibility and adaptability required to handle dynamic and diverse driving conditions, reducing their effectiveness in real-world applications. To overcome these limitations, we propose Stochastic Template-Based RRT* (ST-RRT*), a novel approach that introduces stochasticity into the template generation process. By dynamically generating templates guided by probabilistic models, ST-RRT* achieves efficient exploration, improves adaptability to complex constraints, and retains the asymptotic optimality guarantees of RRT*. We demonstrate the effectiveness of ST-RRT* through experiments in automotive environments, showcasing its ability to generate high-quality trajectories under stringent motion constraints. Additionally, we validate its generalizability by applying it to other kinodynamic planning scenarios, highlighting its efficiency, robustness, and versatility compared to state-of-the-art methods.
{"title":"Efficient and Asymptotically Optimal Vehicle Motion Planning With Stochastic Template-Based RRT*","authors":"Shaoyu Yang;Masamichi Shimosaka","doi":"10.1109/ACCESS.2025.3546158","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3546158","url":null,"abstract":"Kinodynamic motion planning plays a vital role in robotics, particularly in autonomous driving, where planned trajectories must satisfy kinematic and dynamic constraints while ensuring both safety and efficiency. Although existing kinodynamic RRT* algorithms achieve asymptotic optimality, their high computational cost often limits their practicality in high-dimensional or complex environments, such as autonomous driving scenarios. To enhance efficiency in such scenarios, motion templates with predefined action sequences have been proposed as a guiding strategy for planners. However, traditional fixed templates lack the flexibility and adaptability required to handle dynamic and diverse driving conditions, reducing their effectiveness in real-world applications. To overcome these limitations, we propose Stochastic Template-Based RRT* (ST-RRT*), a novel approach that introduces stochasticity into the template generation process. By dynamically generating templates guided by probabilistic models, ST-RRT* achieves efficient exploration, improves adaptability to complex constraints, and retains the asymptotic optimality guarantees of RRT*. We demonstrate the effectiveness of ST-RRT* through experiments in automotive environments, showcasing its ability to generate high-quality trajectories under stringent motion constraints. Additionally, we validate its generalizability by applying it to other kinodynamic planning scenarios, highlighting its efficiency, robustness, and versatility compared to state-of-the-art methods.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"38533-38547"},"PeriodicalIF":3.4,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10904464","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143553242","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：