检索结果-内蒙古大学图书馆

33rd ieee International conference on Computer Communications and Networks (ICCCN)

作者： Al Lail, Mustafa Pinto, Daniela Almanza, Luis Alvarez Salazar, Francisco Rizzi, Carolina Texas A&M Int Univ Sch Engn Laredo TX 78041 USA

ISBN: (纸本)9798350348439;9798350384611

Access control systems in large organizations often struggle with managing complex policies and workloads. However, there is potential for deep learning models to address these challenges. This study delves into the suitability of various deep learning architectures for making real-time access control decisions. Six prominent Convolutional Neural Network (CNN) models (ResNet, DenseNet, Xception, Inception, AlexNet, VGG-16) are evaluated, and the impact of data augmentation using SMOTE on their performance is analyzed. The findings demonstrate that most deep learning models consistently deliver results in access control. ResNet outperforms other models, showing high accuracy across original and SMOTE-augmented datasets. Moreover, SMOTE generally enhances performance for most models, highlighting its potential for addressing data imbalance. These results indicate that deep learning shows promise for improving access control tasks.

关键词： Access control Deep learning CNN Security

来源：评论

学校读者我要写书评

暂无评论

Efficient Real2Sim2Real of Continuum Robots Using Deep Reinforcement learning With Koopman Operator

引用

ieee TRANSACTIONS ON INDUSTRIAL ELECTRONICS 2025年

作者： Ji, Guanglin Gao, Qian Xiao, Yin Sun, Zhenglong Chinese Univ Hong Kong Sch Sci & Engn Shenzhen 518172 Peoples R China Chinese Univ Hong Kong Shenzhen Inst Artificial Intelligence & Robot Soc Shenzhen 518129 Peoples R China

Accurate control of continuum robots is challenging, especially in the presence of external disturbances. To address this issue, reinforcement learning (RL) has been increasingly investigated in continuum robot control due to its online policy updating capability. However, the gap in Sim2Real transfer caused by inaccurate modeling results in RL implementation a dilemma. In this article, we propose a safety-critical Real2Sim2Real online RL framework, where the Real2Sim transfer is first achieved by the identification of a continuum robot using the Koopman operator. We further improve the training efficiency by introducing an imperfect demonstration into the RL framework. The offline policy is trained in simulations and then tested on a real continuum robot platform. During tests, the tracking performance is influenced by the hysteresis effect that cannot be captured by the Koopman operator. This results in a millimeter-level tracking root mean square error (RMSE). To address this issue, we online update the policy as well as the model, and the RMSE of the online controller outperforms the offline controller by 89.16% in free space and 85.70% under external payload, respectively.

关键词： Robots Continuum robots Training Real-time systems Kinematics Aerospace electronics Accuracy Sun Service robots Manipulators Continuum robot data-driven control learning from demonstration reinforcement learning (RL)

来源：评论

学校读者我要写书评

暂无评论

Performance Evaluation and Order Selection for Pattern Modeling Methods Through Granger Causality Analysis 13

Performance Evaluation and Order Selection for Pattern Model...

引用

13th ieee data driven control and learning systems conference, DDCLS 2024

作者： Zhu, Shaotong Zheng, Niannian Luan, Xiaoli Liu, Fei Jiangnan University Wuxi214028 China Wuxi214122 China

ISBN: (纸本)9798350361674

In the realm of processes modeling and control, the monitoring of processes is crucial for ensuring safety, and a key task in multivariate statistical process monitoring is to extract the operating patterns of the process. Dynamic strategies, exemplified by dynamic-inner principal components analysis (DIPCA), dynamic-inner partial least squares (DIPLS), and dynamic-inner canonical correlation analysis (DICCA), are employed to tackle temporal correlation and the ever-changing nature of processes using AR structures. Dynamic controlled principal components analysis (DCPCA) additionally considers dynamic controlled characteristics, showcasing remarkable accuracy in both the monitoring of processes and dynamic prediction. Notwithstanding the precision demonstrated by DCPCA, the task of ascertaining optimal control input orders remains intricate, often necessitating reliance on empirical methodologies or trial-and-error approaches. This research introduces an innovative methodology, drawing inspiration from Granger causality analysis, to establish causal relationships between control inputs and patterns. Causal influence (CI) was employed to assess the impact of introducing control inputs and the effects of control inputs at different orders, aiding the model in selecting the optimal control input order. Theoretically supported performance evaluation and order selection, providing assurance for practical applications. © 2024 ieee.

关键词： Statistical process control

来源：评论

学校读者我要写书评

暂无评论

data Representation with Robust Hypergraph Coding 6

Data Representation with Robust Hypergraph Coding

引用

6th International conference on data-driven Optimization of Complex systems (DOCS)

作者： Yan, Xueming Lei, Mingqi Jin, Yaochu Zhu, Xiuyun Guangdong Univ Foreign Studies Sch Informat Sci & Technol Guangzhou 510006 Peoples R China Westlake Univ Sch Engn Hangzhou 310030 Peoples R China Southern Univ Sci & Technol Affiliated Hosp 2 Sch Med Shenzhen 518000 Peoples R China

ISBN: (纸本)9798350377859;9798350377842

This paper introduces a novel matrix factorization framework, data Representation with Robust Hypergraph Coding (MFRHL), designed to enhance unsupervised learning techniques in data mining and machine learning. Traditional matrix factorization methods often fail to capture complex semantic information in datasets. We leverage hypergraph-based Laplacian Eigenmaps to construct a discriminative basis that captures advanced semantic relationships, followed by sparse coding to optimize this basis. The MFRHL framework significantly improves performance in data clustering tasks by effectively handling high-dimensional data and preserving intricate data relationships. Comprehensive experimental results demonstrate that this method consistently outperforms other cutting-edge matrix factorization methods in data clustering.

关键词： Matrix factorization Hypergraph learning Dimensionality reduction data clustering

来源：评论

学校读者我要写书评

暂无评论

IS2N: Intent-driven Security Software-Defined Network With Blockchain

引用

ieee NETWORK 2024年第3期38卷 118-127页

作者： Song, Yanbo Feng, Tao Yang, Chungang Mi, Xinru Jiang, Shanqing Guizani, Mohsen Xidian Univ State Key Lab Integrated Serv Networks Xian 710071 Peoples R China Acad Mil Sci PLA Inst Syst Engn Beijing 100039 Peoples R China Southeast Univ Sch Cyber Sci & Engn Nanjing 210096 Peoples R China Mohamed Bin Zayed Univ Artificial Intelligence MBZ Machine Learning Dept Abu Dhabi U Arab Emirates

Software-defined network (SDN) is characterized by its programmability, flexibility, and the separation of control and data planes. However, SDN still have many challenges, particularly concerning the security of network information synchronization and network element registration. Blockchain and intent-driven networks are recent technologies to establish secure and intelligent SDN. This article investigates the blockchain-based architecture and intent-driven mechanisms to implement intent-driven security software-defined networks (IS2N). Specifically, we propose a novel four-layer architecture of the IS2N with security capabilities. We integrate an intent-driven security management mechanism in the IS2N to achieve automate network security management. Finally, we develop an IS2N platform with blockchain middle-layer to achieve security capabilities and security store network-level snapshots, such as device registration and OpenFlow messages. Our simulations show that IS2N is more flexible than conventional strategies at resolving problems during network operations and has a minimal effect on the SDN.

关键词： Blockchains Security Monitoring Network security Middleware control systems Security management Software defined networking Internet

来源：评论

学校读者我要写书评

暂无评论

DNN Assisted Projection Based Deep Reinforcement learning for Safe control of Distribution Grids

引用

ieee TRANSACTIONS ON POWER systems 2024年第4期39卷 5687-5698页

作者： Zhang, Mengfan Guo, Guodong Zhao, Tianyang Xu, Qianwen KTH Royal Inst Technol Elect Power & Energy Syst Div S-11428 Stockholm Sweden State Grid Econ & Technol Res Inst Co Ltd Beijing 102209 Peoples R China

Deep reinforcement learning (DRL) is a promising solution for voltage control of distribution grids with high penetration of inverter-based renewable energy sources (RESs). Yet, when adopting the DRL-based control method, the safe and optimal operation of the system cannot be guaranteed at the same time, as the conventional DRL agent is not designed to solve the hard constraint problem. To address this challenge, this paper proposes a deep neural network (DNN) assisted projection based DRL method for safe control of distribution grids. First, a finite iteration projection algorithm is proposed to guarantee hard constraints by converting a non-convex optimization problem into a finite iteration problem. Next, a DNN assisted projection method is proposed to accelerate the calculation of projection and achieve the practical implementation of hard constraints in DRL problem. Finally, a DNN Projection embedded twin-delayed deep deterministic policy gradient (DPe-TD3) method is proposed to achieve optimal operation of distribution grids with guaranteed 100% safety of the distribution grid. The safety of the DRL training is guaranteed via the embedded Projection DNN in TD3 with participation in gradient return process, which could smoothly and effectively project the DRL agent actions into the feasible area, thus guaranteeing the safety of data driven control and the optimal operation at the same time. The case studies and comparisons are conducted in the ieee 33 bus system to show the effectiveness of the proposed method.

关键词： Deep reinforcement learning safety projection deep neural network inverter interfaced RESs distribution grid

来源：评论

学校读者我要写书评

暂无评论

Generation of Synthetic data for Deep learning in Manufacturing Quality control systems 22

Generation of Synthetic Data for Deep Learning in Manufactur...

引用

ieee 22nd Mediterranean Electrotechnical conference (MELECON)

作者： Werda, Mohamed Slim Taibi, Hamza Kouiss, Khalid Chebak, Ahmed SIGMA Clermont Univ FORVIA Nanterre France Univ Mohammed VI Polytech Green Tech Inst Benguerir Morocco

ISBN: (纸本)9798350387032;9798350387025

The automotive industry has improved its quality control systems for detecting manufacturing defects through the implementation of deep learning. However, collecting and preparing datasets to train AI models is considered a challenge, given the high diversity of product types and the wide range of defects that can occur. This paper demonstrates the possibility of using synthetic data to train inspection models, and how it reduces the time and the cost involved in data collection, preparation, and cleaning. We describe how we create a virtual environment to generate synthetic images and how we prepare the 3D model of products to render auto-labelled training images with high realism and variety. We demonstrate in two industry use cases that a deep learning model trained with synthetic data can match the accuracy of one trained with real world data.

关键词： Computer vision Artificial intelligence Deep learning Defect detection Synthetic data Quality inspections Industry 4.0

来源：评论

学校读者我要写书评

暂无评论

Adaptive Path-Tracking controller Embedded With Reinforcement learning and Preview Model for Autonomous Driving

引用

ieee TRANSACTIONS ON VEHICULAR TECHNOLOGY 2025年第3期74卷 3736-3750页

作者： Xia, Qi Chen, Peng Xu, Guoyan Sun, Haodong Li, Liang Yu, Guizhen Beihang Univ Sch Transportat Sci & Engn Key Lab Autonomous Transportat Technol Special Veh Minist Ind & Informat Technol Beijing 100191 Peoples R China Beijing Municipal Inst City Planning & Design Beijing 100045 Peoples R China Beijing Municipal Commiss Planning & Nat Resources Beijing 100045 Peoples R China

Path tracking control is a crucial function in an autonomous vehicle. Previous studies have presented conventional model-based controllers, which lack generalization ability, especially in dealing with high-curvature cases. In contrast to those model-based controllers, data-driven controllers, represented by reinforcement learning (RL), are promising to improve tracking control accuracy and generalization by online learning. However, the inherent inexplicability of RL degrades the stability of an RL-based controller, further limiting its applications in real-world autonomous driving cases. To resolve this issue, this study proposes an adaptive tracking controller based on RL, allowing to enhance network stability through a preview model and adaptive correction. It is composed of three modules, namely a lateral controller, a lateral adaptive corrector, and a longitudinal speed planner. Specifically, the preview control theory is combined with the twin delayed deep deterministic policy gradient (TD3) method to improve the tracking convergence of the lateral controller;the adaptive corrector derives correction compensations based on predicted waypoints along a path, thus enhancing control accuracy;the longitudinal planner outputs a speed profile via a reward increment soft actor-critic algorithm, through which large lateral tracking errors are reduced. Using an intelligent vehicle with sufficient steering capability, simulations and real-world tests show that curvy paths with a curvature larger than 2.0 m(-1) can be closely tracked within an error bound of 0.1 m.

关键词： Adaptation models Autonomous vehicles Accuracy Training Vehicle dynamics control theory Neural networks Adaptive systems Tracking Planning Tracking control reinforcement learning preview control autonomous driving

来源：评论

学校读者我要写书评

暂无评论

From Raw data to Safety: Reducing Conservatism by Set Expansion 6

From Raw Data to Safety: Reducing Conservatism by Set Expans...

引用

6th Annual learning for Dynamics and control conference

作者： Bajelani, Mohammad van Heusden, Klaske Univ British Columbia Sch Engn 3333 Univ Way Kelowna BC V1V 1V7 Canada

Safety filters have been proposed as a modular technique in response to safety concerns associated with learning-based algorithms. Generally, these filters heavily rely on the system's model, which is contradictory if they are intended to enhance data-driven or end-to-end learning solutions. This paper extends our previous work, a purely data-driven Safety Filter (DDSF) based on Willems' lemma, to extremely short-sighted and non-conservative settings. Specifically, we propose online and offline sample-based methods to expand the safe set of DDSF and reduce its conservatism. Since this method is defined in an input-output framework, it can systematically handle unknown time-delay LTI systems using only one batch of data. The proposed method is applied to a time-delay system under various settings to evaluate its performance. The simulation results validate the effectiveness of the set expansion algorithm in generating a notably large input-output safe set, resulting in safety filters that are not conservative, even with an extremely short prediction horizon.

关键词： data-driven Safety Filter learning-based control Behavioural System Theory Time-Delay systems Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Identification of Unstable Linear systems using data-driven Koopman Analysis 21

Identification of Unstable Linear Systems using Data-driven ...

引用

21st International conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology (ECTI-CON)

作者： Ketthong, Patinya Samkunta, Jirayu Nghia Thi Mai Hashikura, Kotaro Kamal, Md Abdus Samad Murakami, Iwanori Yamada, Kou Gunma Univ Grad Sch Sci & Technol 1-5-1 Tenjincho Kiryu Gumma 3768515 Japan Thai Nichi Inst Technol Fac Engn Bangkok Thailand Posts & Telecommun Inst Technol Dept Elect & Elect Km10 Hanoi Vietnam Gunma Univ Div Mech Sci & Technol 1-5-1 Tenjincho Kiryu Gumma 3768515 Japan

ISBN: (纸本)9798350381566;9798350381559

System identification plays a crucial role in modern control techniques, enabling the data-driven learning of input-output maps or mathematical models. However, practical applications face challenges as the actual number of states is often unknown, and observed variables may be limited. Additionally, unstable systems present further difficulties, as their outputs rapidly diverge or saturate, hindering long-term measurement. This paper addresses these challenges by proposing a novel input-aware modeling method for unstable linear systems using data-driven Koopman analysis. Unlike traditional Koopman analysis which focuses solely on state dynamics, our method explicitly incorporates the influence of the input function u(t). This enables us to accurately capture the complete behavior of the system, even under the influence of external control signals. By leveraging Koopman operator theory on augmented state-input data, we capture both the intrinsic dynamics and the sensitivity to external control, crucial for accurate prediction and control of unstable systems. This input-aware approach extends the capabilities of data-driven Koopman analysis to improve modeling and control of complex unstable systems in various applications.

关键词： Sparse modeling HAVOK algorithm System identification

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：