检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Zeng, Fanhu Guo, Haiyang Zhu, Fei Shen, Li Tang, Hao State Key Laboratory of Multimodal Artificial Intelligence Systems Institute of Automation Chinese Academy of Sciences China Centre for Artificial Intelligence and Robotics HKISI-CAS Hong Kong School of Cyber Science and Technology Sun Yat-sen University China School of Computer Science Peking University China

Fine-tuning pre-trained models with custom data leads to numerous expert models on specific tasks. Merging models into one universal model to empower multi-task ability refraining from data leakage has gained popularity. With the expansion in data and model size, parameter efficient tuning becomes the common practice for obtaining task-specific models efficiently. However, we observe that existing methods designed for full fine-tuning merging fail under efficient tuning. To address the issues, we analyze from low-rank decomposition and reveal that maintaining direction and compensating for gap between singular values are crucial for efficient model merging. Consequently, we propose CoPA-Merging, a training-free parameter efficient merging method with complementary parameter adaptation. Specifically, we (1) prune parameters and construct scaling coefficients from inter-parameter relation to compensate for performance drop from task interference and (2) perform cross-task normalization to enhance unseen task generalization. We establish a benchmark consisting of diverse multimodal tasks, on which we conduct experiments to certificate the outstanding performance and generalizability of our method. Additional study and extensive analyses further showcase the effectiveness. Copyright © 2025, The Authors. All rights reserved.

关键词： Information leakage

来源：评论

学校读者我要写书评

暂无评论

Practical Continual Forgetting for Pre-trained Vision Models

arXiv

引用

arXiv 2025年

作者： Zhao, Hongbo Zhu, Fei Ni, Bolin Zhu, Feng Meng, Gaofeng Zhang, Zhaoxiang State Key Laboratory of Multimodal Artificial Intelligence Systems Institute of Automation Chinese Academy of Sciences Beijing100190 China University of Chinese Academy of Sciences Beijing100190 China Centre for Artificial Intelligence and Robotics Hong Kong Institute of Science & Innovation Chinese Academy of Sciences 999077 Hong Kong SenseTime Research Beijing100084 China

For privacy and security concerns, the need to erase unwanted information from pre-trained vision models is becoming evident nowadays. In real-world scenarios, erasure requests originate at any time from both users and model owners, and these requests usually form a sequence. Therefore, under such a setting, selective information is expected to be continuously removed from a pre-trained model while maintaining the rest. We define this problem as continual forgetting and identify three key challenges. (i) For unwanted knowledge, efficient and effective deleting is crucial. (ii) For remaining knowledge, the impact brought by the forgetting procedure should be minimal. (iii) In real-world scenarios, the training samples may be scarce or partially missing during the process of forgetting. To address them, we first propose Group Sparse LoRA (GS-LoRA). Specifically, towards (i), we introduce LoRA modules to fine-tune the FFN layers in Transformer blocks for each forgetting task independently, and towards (ii), a simple group sparse regularization is adopted, enabling automatic selection of specific LoRA groups and zeroing out the others. To further extend GS-LoRA to more practical scenarios, we incorporate prototype information as additional supervision and introduce a more practical approach, GS-LoRA++. For each forgotten class, we move the logits away from its original prototype. For the remaining classes, we pull the logits closer to their respective prototypes. We conduct extensive experiments on face recognition, object detection and image classification and demonstrate that our method manages to forget specific classes with minimal impact on other classes. Codes have been released on https://***/bjzhb666/GS-LoRA. Copyright © 2025, The Authors. All rights reserved.

关键词： Face recognition

来源：评论

学校读者我要写书评

暂无评论

ProtoGCD: Unified and Unbiased Prototype Learning for Generalized Category Discovery

arXiv

引用

arXiv 2025年

作者： Ma, Shijie Zhu, Fei Zhang, Xu-Yao Liu, Cheng-Lin State Key Laboratory of Multimodal Artificial Intelligence Systems Institute of Automation Chinese Academy of Sciences 95 Zhongguancun East Road Beijing100190 China School of Artificial Intelligence University of Chinese Academy of Sciences Beijing100049 China Centre for Artificial Intelligence and Robotics Hong Kong Institute of Science and Innovation Chinese Academy of Sciences 999077 Hong Kong

Generalized category discovery (GCD) is a pragmatic but underexplored problem, which requires models to automatically cluster and discover novel categories by leveraging the labeled samples from old classes. The challenge is that unlabeled data contain both old and new classes. Early works leveraging pseudo-labeling with parametric classifiers handle old and new classes separately, which brings about imbalanced accuracy between them. Recent methods employing contrastive learning neglect potential positives and are decoupled from the clustering objective, leading to biased representations and sub-optimal results. To address these issues, we introduce a unified and unbiased prototype learning framework, namely ProtoGCD, wherein old and new classes are modeled with joint prototypes and unified learning objectives, enabling unified modeling between old and new classes. Specifically, we propose a dual-level adaptive pseudo-labeling mechanism to mitigate confirmation bias, together with two regularization terms to collectively help learn more suitable representations for GCD. Moreover, for practical considerations, we devise a criterion to estimate the number of new classes. Furthermore, we extend ProtoGCD to detect unseen outliers, achieving task-level unification. Comprehensive experiments show that ProtoGCD achieves state-of-the-art performance on both generic and fine-grained datasets. Copyright © 2025, The Authors. All rights reserved.

关键词： Semi-supervised learning

来源：评论

学校读者我要写书评

暂无评论

PASS++: A Dual Bias Reduction Framework for Non-Exemplar Class-Incremental Learning

引用

IEEE Transactions on Pattern Analysis and Machine Intelligence 2025年 PP卷 PP页

作者： Zhu, Fei Zhang, Xu-Yao Cheng, Zhen Liu, Cheng-Lin Hong Kong Institute of Science and Innovation Chinese Academy of Sciences Centre for Artificial Intelligence and Robotics 999077 Hong Kong Institute of Automation of Chinese Academy of Sciences State Key Laboratory of Multimodal Artificial Intelligence Systems Beijing 100190 China University of Chinese Academy of Sciences School of Artificial Intelligence Beijing 100049 China

Class-incremental learning (CIL) aims to continually recognize new classes while preserving the discriminability of previously learned ones. Most existing CIL methods are exemplar-based, relying on the storage and replay of a subset of old data during training. Without access to such data, these methods typically suffer from catastrophic forgetting. In this paper, we identify two fundamental causes of forgetting in CIL: representation bias and classifier bias. To address these challenges, we propose a simple yet effective dual-bias reduction framework, which leverages self-supervised transformation (SST) in the input space and prototype augmentation (protoAug) in the feature space. On one hand, SST mitigates representation bias by encouraging the model to learn generic, diverse representations that generalize across tasks. On the other hand, protoAug tackles classifier bias by explicitly or implicitly augmenting the prototypes of old classes in the feature space, thereby imposing stronger constraints to preserve decision boundaries. We further enhance the framework with hardness-aware prototype augmentation and multi-view ensemble strategies, yielding significant performance gains. The proposed framework can be easily integrated with pre-trained models. Without storing any samples of old classes, our method performs comparably to state-of-the-art exemplar-based approaches that rely on extensive data storage. We hope to draw the attention of researchers back to non-exemplar CIL by rethinking the necessity of storing old samples. © 1979-2012 IEEE.

关键词： Catastrophic forgetting class-incremental learning continual learning prototype augmentation self-supervision

来源：评论

学校读者我要写书评

暂无评论

AI Enabled Automatic Mobile Robot Intelligent Navigation in Construction with Obstacle Awareness

引用

IEEE Transactions on automation Science and Engineering 2025年

作者： Zhang, Yinlong Liu, Yuanhao Cui, Yunge Zeng, Ziming Liang, Wei Chinese Academy of Sciences State Key Laboratory of Robotics Shenyang Institute of Automation Shenyang110016 China Chinese Academy of Sciences Key Laboratory of Networked Control Systems Shenyang110016 China University of Chinese Academy of Sciences Beijing101408 China Zeekr Group Zeekr Intelligent Driving Department Shanghai110078 China Shenzhen Polytechnic University School of Automotive and Transportation Engineering Shenzhen518055 China

In the construction industry, the integration of artificial intelligence (AI) and robotics has led to significant advancements in automating various tasks. One critical aspect is the intelligent navigation of automatic mobile robot (AMR) within construction sites, where dynamic environments pose challenges such as inaccurate robot state estimation, irregular and textureless obstacle detection. To solve these issues, this paper designs an AI enabled obstacle-aware AMR intelligent navigation approach while ensuring robot safety and efficiency. Specifically, the robot is equipped with the complementary RGB camera, inertial measurement unit (IMU) and wheel encoder to estimate the states (i.e., positions, orientations and velocities), which have been tightly fused in the optimization framework. Furthermore, the RGB-Depth images are jointly fed into the AI model. It combines of Mask-RCNN and multi-scale attention network, to detect and segment the obstacles, and estimate the corresponding relative depth. It should be noted that the system incorporates obstacle awareness mechanisms to dynamically switch the robot's velocities in response to obstacles in the environment, ensuring smooth and collision-free movement. Experimental results demonstrate the effectiveness and robustness of the proposed AI-enabled navigation system in real-world construction scenarios, showcasing its potential to enhance productivity and safety in construction automation. © 2004-2012 IEEE.

关键词： Construction industry

来源：评论

学校读者我要写书评

暂无评论

Safe and Reliable Distributed Fault-Tolerant Formation Control for Quadrotor Swarms

引用

IEEE Transactions on Aerospace and Electronic systems 2025年

作者： Hu, Yang Miao, Zhiqiang Wang, Yaonan Zhang, Hui Wang, Xiangke He, Wei Hunan University College of Electrical and Information Engineering Changsha410082 China Hunan University School of Robotics Changsha410082 China National University of Defense Technology College of Intelligence Science and Technology Changsha410073 China Beijing Information Science and Technology University School of Automation Institute of Artificial Intelligence China University of Science and Technology Beijing Key Laboratory of Intelligent Bionic Unmanned Systems Ministry of Education Beijing100083 China

This paper addresses formation control of quadrotor swarms subject to external time-varying disturbances and actuator failures, while ensuring collision avoidance. Firstly, a complete dynamics model incorporating translational movement, rotation, motor, and actuator failures was developed for each quadrotor. Then, nominal controllers for the position and attitude subsystems are designed separately to achieve formation control under ideal conditions. A higher-order sliding mode (HOSM) observer is implemented to effectively estimate and compensate for external disturbances and actuator bias faults, thereby improving the stability of the system. Next, a potential energy function-based collision avoidance mechanism is implemented to ensure inter-quadrotor collision avoidance. Furthermore, a fault detection and fault-tolerant control(FTC) method is designed to estimate the lift coefficient of actuator partial failures online in real time by employing Newtons correction method, which ensures the safe and reliable flight of the system. The stability of the closed-loop system is analyzed using Lyapunov stability theory. Finally, numerical simulations and hardware-in-the-loop (HIL) experiment results are performed to validate the approach's effectiveness and performance in quadrotor swarm formation control. © 2025 IEEE. All rights reserved.

关键词： Potential energy functions

来源：评论

学校读者我要写书评

暂无评论

Brain-inspired analogical mixture prototypes for few-shot class-incremental learning

arXiv

引用

arXiv 2025年

作者： Li, Wanyi Wei, Wei Luo, Yongkang Wang, Peng State Key Laboratory of Multimodal Artificial Intelligence Systems Institute of Automation Chinese Academy of Sciences Beijing100190 China School of Artificial Intelligence University of Chinese Academy of Sciences Beijing100049 China CAS Center for Excellence in Brain Science and Intelligence Technology Chinese Academy of Sciences Shanghai200031 China Centre for Artificial Intelligence and Robotics Hong Kong Institute of Science and Innovation Chinese Academy of Sciences 999077 Hong Kong

Few-shot class-incremental learning (FSCIL) poses significant challenges for artificial neural networks due to the need to efficiently learn from limited data while retaining knowledge of previously learned tasks. Inspired by the brain’s mechanisms for categorization and analogical learning, we propose a novel approach called Brain-inspired Analogical Mixture Prototypes (BAMP). BAMP has three components: mixed prototypical feature learning, statistical analogy, and soft voting. Starting from a pre-trained Vision Transformer (ViT), mixed prototypical feature learning represents each class using a mixture of prototypes and fine-tunes these representations during the base session. The statistical analogy calibrates the mean and covariance matrix of prototypes for new classes according to similarity to the base classes, and computes classification score with Mahalanobis distance. Soft voting combines both merits of statistical analogy and an off-shelf FSCIL method. Our experiments on benchmark datasets demonstrate that BAMP outperforms state-of-the-art on both traditional big start FSCIL setting and challenging small start FSCIL setting. The study suggests that brain-inspired analogical mixture prototypes can alleviate catastrophic forgetting and over-fitting problems in FSCIL. Copyright © 2025, The Authors. All rights reserved.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

ANALYSIS OF TIPOVER STABILITY FOR NOVEL SHAPE SHIFTING MODULAR ROBOT

引用

Chinese Journal of Mechanical Engineering 2006年第2期19卷 181-186页

作者： LIU Jinguo WANG Yuechao MA Shugen LI Bin Robotics Laboratory Shenyang Institute of Automation Chinese Academy of SciencesShenyang 110016 China Department of Systems Engineering Ibaraki UniversityIbaraki 316-8511 Japan Robotics Laboratory Shenyang Institute of AutomationChinese Academy of SciencesShenyang 110016 China

A novel three-module robot has been introduced. It can change its configuration to adapt to the uneven terrain and to improve its tipover stability. This three-module tracked robot has three kinds of symmetry configuration. They are line type, triangle type, and row type. After the factors and the countermeasures of mobile robot＇s tipover problem are analyzed, stability pyramid and tipover stabil-ity index are proposed to globally determinate the mobile robot＇s static stability and dynamic stability. The shape shifting robot is tested by this technique under the combined disturbance of pitch, roll and yaw in simulation. The simulation result shows that this technique is effective for the analysis of mobile robot＇s tipover stability, especially for the reconfigurable or shape shifting modular robot. Experiments on three symmetry configurations are made under unstructured environments. The environment experiment shows the same result as that of the simulation that the triangle type configuration has the best stability. Both simulation and experiment provide a valid reference for the reconfigurable robot＇s potential application.

关键词： Reconfigurable modular robot Shape shifting robot Stability pyramid Tipover stability index Unstructured environment

来源：评论

学校读者我要写书评

暂无评论

Gait Generation for the Multi-legged Robot with Injured Legs 41

Gait Generation for the Multi-legged Robot with Injured Legs

引用

第41届中国控制会议

作者： Ming Zeng Zhijing Wang Yu Ma Jiazhe Liu Baofeng Zhang Yunke Nie Institute of Robotics and Automation Systems School of Electrical and Information EngineeringTianjin University

ISBN: (数字)9789887581536

ISBN: (纸本)9781665482561

When multi-legged robots work in dangerous and complex environments,their legs are easily damaged,resulting in a sharp decline in their movement ability,or even completely out of *** this paper,the gait generation method combined deep reinforcement learning with bionic gait restrictions is presented for multi-legged robots with injured *** this study,we first built an ant gait video acquisition platform using the Camponotus Japonicus as the research object,and on this basis established a gait dataset of injured ***,the gait characteristics of ants in the injured state were ***,our proposed gait generation method combining bionic gait constraints with deep reinforcement learning was compared with the reinforcement-learning-based method in the MATLAB simulation *** test results show that the training reward values of our proposed algorithm are 1-2 times higher than those of the reinforcement learning method alone,and at the same time,the moving speed of the injured robot is also nearly twice as fast.

关键词： Injured multi-legged robot,Ant gait,Gait generation,Reinforcement learning Bionic constraints

来源：评论

学校读者我要写书评

暂无评论

Odor-Source Location Estimation and Search Using a Movable Sensor Array

Odor-Source Location Estimation and Search Using a Movable S...

引用

第三十二届中国控制会议

作者： ZHANG Su-Lin MENG Qing-Hao WU Yu-Xiu ZENG Ming JIA Hua LI Wei Institute of Robotics and Autonomous Systems School of Electrical Engineering and AutomationTianjin University

ISBN: (纸本)9789881563835

Using mobile robot with onboard gas and wind sensors to estimate and search hazardous gas/odor source has great application *** paper presents an odor-source localization(OSL) method using a mobile robot equipped with a gas sensor array(we call it movable sensor array for short) in both indoor and outdoor *** gas sensors and one wind sensor are placed at the circumference and the center of a circle,*** on the assumption that the wind is uniform in a small area,at each iteration cycle,the robot is controlled to move a step toward the source location estimated by fusing the sensors' *** source location is calculated by solving nonlinear least square equation with Levenberg-Marquardt algorithm. The odor source is declared by combing a spiral motion and heuristic *** proposed OSL method has been evaluated in both indoor and outdoor simulation plume *** results show that the movable sensor array is able to find the odor-source location with a small error,but the success rate is related to the arrangement of sensor array and motion step.

关键词： Odor source localization sensor array mobile robot estimation nonlinear least square

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：