检索结果-内蒙古大学图书馆

第42届中国控制会议

作者： Ganghui Cao Jinzhi Wang State Key Laboratory for Turbulence and Complex Systems Department of Mechanics and Engineering ScienceCollege of Engineering Peking University

In this paper, a distributed disturbance observer is revisited, and here attention is paid to the completely distributed design. To monitor a linear dynamical system, assume N observations are collected by a group of sensor nodes severally. N observers(collectively referred to as the distributed disturbance observer) are constructed to cooperatively estimate the states as well as the disturbance signals. However, the coupling gains of the N observers need an overall integration design. Given that some global information is not readily available, an adaptive strategy is applied to avoid using it, and therefore completely distributed.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Bipartite consensus tracking for antagonistic topologies with leader's unknown input

Bipartite consensus tracking for antagonistic topologies wit...

引用

作者： Nan, Xiaoya Lv, Yuezu Duan, Zhisheng State Key Laboratory for Turbulence and Complex Systems Department of Mechanics and Engineering Science College of Engineering Peking University Beijing China School of Mathematics Southeast University Nanjing China

In this paper, the bipartite consensus tracking problem of linear multiagent systems on leader-follower signed directed graph is studied, where the leader's unknown input is taken into consideration. Based on only local relative information of neighboring agents, discontinuous distributed adaptive protocols are proposed, where nonlinear function is introduced for each agent to tackle leader's unknown input. To avoid tremors phenomenon, continuous distributed protocols are further presented to achieve practical bipartite consensus. Both the state feedback and output feedback cases are investigated. Different from existing literature showing bipartite consensus achievement with gauge transformation, this paper reveals a novel property of the signed Laplacian matrix, which plays a key role in simplifying the bipartite consensus demonstration. Two practical examples on synchronization of Chua's circuits are presented to illustrate the effectiveness of the designed protocols. © 2021 Chinese Automatic Control Society and John Wiley & Sons Australia, Ltd.

关键词： Multi agent systems

来源：评论

学校读者我要写书评

暂无评论

A Canonical Correlation Analysis-Based Approach for Battery State of Charge Estimation

A Canonical Correlation Analysis-Based Approach for Battery ...

引用

第35届中国控制与决策会议

作者： Zichuan Ni Ying Yang State Key Laboratory for Turbulence and Complex Systems Department of Mechanics and Engineering ScienceCollege of EngineeringPeking University

This brief presents a novel method based on canonical correlation analysis（CCA） and particle filter（PF） for battery state of charge（SOC） *** specifically,CCA is adopted to provide a universal way for battery SOC estimation with PF for error ***,the input data are first mapped to polynomial terms before training to find the intrinsic ***,l-norm regularization is further added to the loss function in order to prevent ***,the SOC estimation result from CCA training is combined with the Coulomb counting model and updated by PF for error *** results using battery testing data at different testing profiles and temperatures show the high accuracy of estimation and robustness to wrong initial *** particular,compared with linear regression,support vector regression,and neural network,the error of the proposed method is reduced to 1.47%.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Dual Intention Escape: Penetrating and Toxic Jailbreak Attack against Large Language Models 25

Dual Intention Escape: Penetrating and Toxic Jailbreak Attac...

引用

34th ACM Web Conference, WWW 2025

作者： Xue, Yanni Wang, Jiakai Yin, Zixin Ma, Yuqing Qin, Haotong Tao, Renshuai Liu, Xianglong State Key Laboratory of Complex & Critical Software Environment Beihang University Beijing China Zhongguancun Laboratory Beijing Beijing China Beihang University Beijing China Institute of Artificial Intelligence Beihang University Beijing China ETH Zurich Zurich Switzerland Institute of Information Science Beijing Jiaotong University Beijing China Zhongguancun Laboratory Beijing China Institute of Data Space Hefei Comprehensive National Science Center Hefei China

ISBN: (纸本)9798400712746

Recently, the jailbreak attack, which generates adversarial prompts to bypass safety measures and mislead large language models (LLMs) to output harmful answers, has attracted extensive interest due to its potential to reveal the vulnerabilities of LLMs. However, ignoring the exploitation of the characteristics in intention understanding, existing studies could only generate prompts with weak attacking ability, failing to evade defenses (e.g., sensitive word detect) and causing malice(e.g., harmful outputs). Motivated by the mechanism in the psychology of human misjudgment, we propose a dual intention escape (DIE) jailbreak attack framework to generate more stealthy and toxic prompts to deceive LLMs to output harmful content. For stealthiness, inspired by the anchoring effect, we designed the Intention-anchored Malicious Concealment(IMC) module that hides the harmful intention behind a generated anchor intention by the recursive decomposition block and contrary intention nesting block. Since the anchor intention will be received first, the LLMs might pay less attention to the harmful intention and enter response status. For toxicity, we propose the Intention-reinforced Malicious Inducement (IMI) module based on the availability bias mechanism in a progressive malicious prompting approach. Due to the ongoing emergence of statements correlated to harmful intentions, the output content of LLMs will be closer to these more accessible intentions, i.e., more toxic. We conducted extensive experiments under black-box settings, supporting that DIE could achieve 100% ASR-R and 92.9% ASR-G against GPT3.5-turbo. © 2025 Copyright held by the owner/author(s).

关键词： dual intention escape

来源：评论

学校读者我要写书评

暂无评论

Analyzing Multiple Phenotypes Based on Principal Component Analysis

引用

Acta Mathematicae Applicatae Sinica 2022年第4期38卷 843-860页

作者： De-liang BU San-guo ZHANG Na LI School of Mathematical Sciences University of Chinese Academy of SciencesBeijing 100049China Key Laboratory of Big Data Mining and Knowledge Management Chinese Academy of SciencesBeijing 100049China LSC NCMISAcademy of Mathematics and Systems ScienceChinese Academy of SciencesBeijing 100190China

Joint analysis of multiple phenotypes can have better interpretation of complex diseases and increase statistical power to detect more significant single nucleotide polymorphisms(SNPs)compare to traditional single phenotype analysis in genome-wide association *** component analysis(PCA),as a popular dimension reduction method,has been broadly used in the analysis of multiple *** PCA transforms the original phenotypes into principal components(PCs),it is natural to think that by analyzing these PCs,we can combine information across *** PCA-based methods can be divided into two categories,either selecting one particular PC manually or combining information from all *** this paper,we propose an adaptive principle component test(APCT)which selects and combines the PCs adaptively by using Cauchy combination *** proposed method can be seen as a generalization of traditional PCA based method since it contains two existing methods as special *** simulation shows that our method is robust and can generate powerful result in various *** real data analysis of stock mice data also demonstrate that our proposed APCT can identify significant SNPs that are missed by traditional methods.

关键词： multiple phenotypes principal component analysis cauchy combination method

来源：评论

学校读者我要写书评

暂无评论

UniAdapter: All-in-One Control for Flexible Video Generation

引用

IEEE Transactions on Circuits and systems for Video Technology 2025年第6期35卷 6059-6073页

作者： Wang, Cong Hu, Panwen Zhao, Haoyu Guo, Yuanfan Gu, Jiaxi Dong, Xiao Han, Jianhua Xu, Hang Liang, Xiaodan Chinese University of Hong Kong Hong Kong School of Science and Engineering Shenzhen China Fudan University School of Computer Science and Technology Shanghai China Huawei Noah'ark Lab Shanghai201206 China Sun Yat-Sen University School of Artificial Intelligence Zhuhai Campus Zhuhai519082 China Shenzhen Campus of Sun Yat-sen University School of Intelligent Systems Engineering Guangdong No. 66 Gongchang Road Guangming District Shenzhen518107 China Peng Cheng Laboratory and Guangdong Key Laboratory of Big Data Analysis and Processing Guangzhou510006 China

Condition-based video generation aims to create video content based on given information that describes specific subjects. However, most existing works can only utilize a single condition to guide the denoising process, thereby limiting their applicability to specific scenarios. Although some works attempt to accommodate multiple conditions within one framework, they often require multiple encoders, leading to inefficiencies in integrating multi-condition features. In this work, we present a framework that, with the support of the proposed Unified Adapter (UniAdapter), enables simultaneous multi-condition control of video generation within a single model. To effectively merge these conditions, we propose a novel Probabilistic Multi-condition Concatenator (PMC) module, which employs a unified encoder to accommodate multiple conditions and concatenate condition features at the pixel level to achieve fine-grained control. Following the PMC module, we employ 2D down-sampling blocks to refine features for injection into the Video Diffusion Model (VDM). Moreover, our UniAdapter is designed to be model-agnostic and compatible with any U-Net-based VDM, offering a versatile solution for improving video generation quality. Experimental results on public benchmarks UCF-101 and MSR-VTT show that our method achieves superior results in both quantitative and qualitative evaluations. © 2025 IEEE.

关键词： Signal encoding

来源：评论

学校读者我要写书评

暂无评论

Full-State Planning of Nonholonomic Mobile Robots via Dynamic Velocity Vector Fields

Full-State Planning of Nonholonomic Mobile Robots via Dynami...

引用

第35届中国控制与决策会议

作者： Xiaodong He Zhongkui Li State Key Laboratory for Turbulence and Complex Systems Department of Mechanics and Engineering ScienceCollege of EngineeringPeking University

As a rigid body,the nonholonomic mobile robot contains both states of position and *** order to plan these states simultaneously,this paper investigates the fullstate planning problem of nonholonomic mobile robots,in the sense that the robots should reach the specified positions and meanwhile point to the desired orientations at the terminal *** this end,we propose a velocity vector field which guides the mobile robots to the goal ***,the dynamics of the robot orientation is brought into the vector field,so that the attitude angle of the robot can converge to the specified value following the orientation ***,we study the obstacle avoidance and mutual-robot-collision avoidance by proposing another velocity vector field,which guides the robots moving along the tangential direction of the dangerous ***,several numerical simulation examples are provided to support the theoretical results.

关键词： Motion planning velocity vector field terminal orientation constraint nonholonomic mobile robot

来源：评论

学校读者我要写书评

暂无评论

Deep Learning-based Moving Object Segmentation:Recent Progress and Research Prospects

引用

Machine Intelligence Research 2023年第3期20卷 335-369页

作者： Rui Jiang Ruixiang Zhu Hu Su Yinlin Li Yuan Xie Wei Zou College of Information Engineering Shanghai Maritime UniversityShanghai201306China School of Computer Science and Technology East China Normal UniversityShanghai200062China Research Center of Precision Sensing and Control Institute of AutomationChinese Academy of SciencesBeijing100190China State Key Laboratory for Management and Control of Complex Systems Institute of AutomationChinese Academy of SciencesBeijing100190China

Moving object segmentation(MOS),aiming at segmenting moving objects from video frames,is an important and challenging task in computer vision and with various *** the development of deep learning(DL),MOS has also entered the era of deep models toward spatiotemporal feature *** paper aims to provide the latest review of recent DL-based MOS methods proposed during the past three ***,we present a more up-to-date categorization based on model characteristics,then compare and discuss each category from feature learning(FL),and model training and evaluation *** FL,the methods reviewed are divided into three types:spatial FL,temporal FL,and spatiotemporal FL,then analyzed from input and model architectures aspects,three input types,and four typical preprocessing subnetworks are *** terms of training,we discuss ideas for enhancing model *** terms of evaluation,based on a previous categorization of scene dependent evaluation and scene independent evaluation,and combined with whether used videos are recorded with static or moving cameras,we further provide four subdivided evaluation setups and analyze that of reviewed *** also show performance comparisons of some reviewed MOS methods and analyze the advantages and disadvantages of reviewed MOS methods in terms of ***,based on the above comparisons and discussions,we present research prospects and future directions.

关键词： Moving object segmentation(MOS) change detection background subtraction deep learning(DL) video understanding

来源：评论

学校读者我要写书评

暂无评论

A Physics-Informed Gaussian Process Regression Algorithm for the Dynamics of the Planar Pendulum

A Physics-Informed Gaussian Process Regression Algorithm for...

引用

第42届中国控制会议

作者： Tianzhi Li Jinzhi Wang State Key Laboratory for Turbulence and Complex Systems Department of Mechanics and Engineering Science College of EngineeringPeking University

Gaussian process regression has received considerable attention due to its performance in solving the problem of learning and predicting the dynamics of certain systems in the machine learning area. However, this data-driven method ignores the prior physical information. A feasible method to tackle this problem is to embed prior dynamics into the Gaussian process regression. This naturally relies on numerical discretizations of continuous-time differential equations that describe the ***, conventional discretization schemes do not respect the intrinsic geometric structure of the system, which plays an important role when analyzing the properties of the mechanical system. In this work, we develop a physic-informed Gaussian process regression algorithm based on Hamel's formalism and its variational integrator. Computational properties are illustrated by the numerical experiment of learning and predicting the dynamics of a planar pendulum.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Towards Autonomous Robotic Ultrasound Scanning Using the Reinforcement Learning-Based Volumetric data Navigation Method 5

Towards Autonomous Robotic Ultrasound Scanning Using the Rei...

引用

5th World Robot Conference Symposium on Advanced Robotics and Automation, WRC SARA 2023

作者： Shen, Cuifeng Deng, Zhaokun Wang, Jinzhi Wang, Shuangyi Chen, Chen Peking University State Key Laboratory for Turbulence and Complex Systems Department of Mechanics and Engineering Science College of Engineering Beijing100871 China University of Chinese Academy of Sciences School of Artificial Intelligence Beijing China Institute of Automation Chinese Academy of Sciences State Key Laboratory of Multimodal Artificial Intelligence Systems Beijing China Hong Kong Institute of Science & Innovation Chinese Academy of Sciences Centre for Artificial Intelligence and Robotics Hong Kong

ISBN: (纸本)9798350307320

Ultrasound scanning is an indispensable diagnostic tool in modern medicine, but obtaining an accurate standard plane relies heavily on the sonographer's skill and expertise. This paper presents a novel approach toward autonomous robotic ultrasound scanning using a reinforcement learning-based navigation method with volumetric ultrasound data. The proposed method enables an ultrasound robot to autonomously acquire high-quality 2D standard planes from patients, thereby reducing the need for human intervention. The core of the approach is a deep reinforcement learning algorithm that detects 2D standard planes within the 3D ultrasound volume and guides the ultrasound probe to extract high-quality standard planes. To assess the effectiveness of the 2D standard plane detection method, an evaluation was conducted using a realworld ultrasound dataset, and a comparison was made between its performance and that of other state-of-the-art autonomous approaches. The simulation experiment was also conducted to validate the autonomous robotic ultrasound acquisition workflow. Results demonstrate that the reinforcement learning-based volumetric data navigation method outperforms existing techniques in terms of standard plane accuracy and is very effective for volumetric data-based robotic navigation. © 2023 IEEE.

关键词： Diagnosis

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：