检索结果-内蒙古大学图书馆

Deep fusion of discrete-time and continuous-time models for long-term prediction of chaotic dynamical systems

Nonlinear Dynamics 2025年第11期113卷 12545-12563页

作者： Cheng, Xinzhuang Liu, Zuolin Fang, Hongbin MOE Engineering Research Center of AI & Robotics Institute of AI and Robotics Fudan University Shanghai200433 China

Time series forecasting plays a crucial role in contemporary production and daily life by analyzing the historical data to predict future trends, patterns, and behaviors across various phenomena. However, forecasting chaotic time series remains a formidable challenge, primarily attributed to the high nonlinearity of chaotic dynamical systems and their extreme sensitivity to noise and initial conditions. In general, predicting through discrete-time models and continuous-time models each have their advantages, but there have been few explorations of their integration. In this paper, we delve into the fusion of these two models by leveraging deep neural networks to achieve a notable improvement in the length and accuracy of chaotic time series prediction. The discrete-time reservoir computing model is utilized to make initial predictions of the chaotic time series, while continuous-time differential equations and Physically Informed Neural Network (PINN) are then adopted to refine the results and mine hidden information from multiple perspectives. The effectiveness of the method is verified by applying it to the Lorenz system. Moreover, we discuss in detail the prediction performance of the method under extreme conditions such as noise, sparse sampling, etc., which reveals satisfactorily results even when a single model fails. By applying the method to an experimental double pendulum system, we further demonstrate the superiority of the method in long-term prediction of real chaotic dynamical systems. Overall, the proposed method represents a new paradigm of multi-method and multi-model fusion, providing innovative ideas for improving the accuracy of chaotic time series prediction. © The Author(s), under exclusive licence to Springer Nature B.V. 2025.

关键词： Continuous time systems

来源：评论

学校读者我要写书评

暂无评论

A Direct-Drive, Wearable Armband Device to Experiment Combined Continuous and Vibrotactile Haptic Feedback for Guidance in Motor Tasks 14th

A Direct-Drive, Wearable Armband Device to Experiment Combi...

引用

14th International Conference on Human Haptic Sensing and Touch Enabled Computer Applications, EuroHaptics 2024

作者： KhalilianMotamed Bonab, Ali Camardella, Cristian Frisoli, Antonio Leonardis, Daniele Institute of Mechanical Intelligence and the Department of Excellence in Robotics and AI Scuola Superiore Sant’Anna Pisa Italy

ISBN: (纸本)9783031700606

Wearable haptic devices have been proposed to convey guidance and feedback information in a variety of applications, ranging from navigation to virtual interaction and prosthetics. A design approach for armband devices using an actuated fabric belt and gearmotors actuators has been proposed and widely experimented in the field. In this work we experiment whether the use of direct-drive actuation in place of servomotors proves effective in rendering lower intensity, but cleaner linear feedback. Interestingly, the method allows for modulation of vibrotactile feedback as well, with the same actuation design. Here we present the design of a wearable haptic device for the upper limb, implementing a 2 degrees of freedom direct drive transmission and a soft belt interface. Experiments evaluate the capability of the haptic feedback to guide participants in two different motor tasks, exploiting the two haptic modalities the device can render: clenching and lateral stretch. Moreover, the addition of modulated vibrotactile feedback to the conventional linear motor activation is explored as a viable modality to enhance the perception and effectiveness of the perceived stimuli. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Linear motors

来源：评论

学校读者我要写书评

暂无评论

Learning Dynamic Representations in Large Language Models for Evolving Data Streams 27th

Learning Dynamic Representations in Large Language Models f...

引用

27th International Conference on Pattern Recognition, ICPR 2024

作者： Srivastava, Ashish Bhatnagar, Shalabh Narasimha Murty, M. Aravinda Raman, J. Indian Institute of Science Bangalore India Centre for AI & Robotics Bangalore India North Carolina State University RaleighNC United States

ISBN: (纸本)9783031781681

In the world of Large Language Modeling, incremental learning plays an important role in evolving data such as streaming text. We introduce an incremental learning approach for dynamic contextualized word embeddings in the setting of streaming data. We call the embeddings generated by our model as Incremental Dynamic Contextualized Word Embeddings (iDCWE). Our model introduces the incremental BERT (iBERT) (BERT stands for Bidirectional Encoder Representations from Transformers) to create a dynamic and incremental model to perform incremental training. Our model further captures the semantic drift of words using dynamic graphs. Our paper is the first in the line of research on (incremental) dynamic modeling of streaming text which we also refer to as Neural Dynamic Language Modeling. The performance of our model on the benchmark datasets is on par and even often out-performs the dynamic contextualized word embeddings which was the first paper to combine contextualization with dynamic word embeddings, to the best of our knowledge. Moreover, the compute time efficiency of our model is better than that of the aforementioned paper. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Modeling languages

来源：评论

学校读者我要写书评

暂无评论

Physics-Driven Data Generation for Contact-Rich Manipulation via Trajectory Optimization

arXiv

引用

arXiv 2025年

作者： Yang, Lujie Suh, H.J. Terry Zhao, Tong Græsdal, Bernhard Paus Kelestemur, Tarik Wang, Jiuguang Pang, Tao Tedrake, Russ Massachusetts Institute of Technology United States Robotics and AI Institute

We present a low-cost data generation pipeline that integrates physics-based simulation, human demonstrations, and model-based planning to efficiently generate large-scale, high-quality datasets for contact-rich robotic manipulation tasks. Starting with a small number of embodiment-flexible human demonstrations collected in a virtual reality simulation environment, the pipeline refines these demonstrations using optimization-based kinematic retargeting and trajectory optimization to adapt them across various robot embodiments and physical parameters. This process yields a diverse, physically consistent, contact-rich dataset that enables cross-embodiment data transfer, and offers the potential to reuse legacy datasets collected under different hardware configurations or physical parameters. We validate the pipeline’s effectiveness by training diffusion policies from the generated datasets for challenging long-horizon contact-rich manipulation tasks across multiple robot embodiments, including a floating Allegro hand and bimanual robot arms. The trained policies are deployed zero-shot on hardware for bimanual iiwa arms, achieving high success rates with minimal human input. Project website: https://***/physicsgen/. © 2025, CC BY.

关键词： Demonstrations

来源：评论

学校读者我要写书评

暂无评论

The Effect of User Learning for Online EEG Decoding of Upper-Limb Movement Intention

IEEE Transactions on Medical Robotics and Bionics

引用

IEEE Transactions on Medical robotics and Bionics 2025年第2期7卷 633-641页

作者： Ceradini, Matteo Tortora, Stefano Micera, Silvestro Tonin, Luca University of Padova Department of Information Engineering Padova Italy University of Padova Padua Neuroscience Center Padova Italy Scuola Superiore Sant'Anna BioRobotics Institute Department of Excellence in Robotics and AI Pisa Italy Translational NeuroEngineering Laboratory Neuro-X Institute Lausanne Switzerland Laboratory Milano Italy

Electroencephalography (EEG) based braincomputer interfaces (BCIs) offer a promising way for individuals with motor impairments to control prosthetic or rehabilitation devices. Accurately decoding movement intention (MI) is crucial for translating subjects' motor execution plans into action. Common challenges in EEG-based BCIs include performance discrepancies, often requiring frequent recalibration of decoding algorithms. The objective of this study was enhancing BCI decoding performance of upper-limb MI identification by exploiting both machine and subjects learning and maintaining stable decoding algorithms. Significant performance improvements were observed across most subjects from the first to the last session of the experiment. Some subjects also demonstrated stable performance without requiring any model recalibration between sessions. All subjects achieved high efficacy in online decoding of movement intention, as reflected in improvement of the F1 score from 0.5810.26 in the first session, to 0.84±0.13 in the final session. We emphasize the critical importance of allowing users sufficient time to improve their performance in BCIs for upper-limb MI decoding. Unlike existing studies, we specifically evaluate the effect of stable decoding strategies in online and longitudinal BCI sessions, which are key to achieving more reliable and effective BCIs. © 2018 IEEE.

关键词： Brain computer interface

来源：评论

学校读者我要写书评

暂无评论

Multi-hop Self-augmented Graph Contrastive Learning for Node Classification

Multi-hop Self-augmented Graph Contrastive Learning for Node...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Yutong Wang Xiaofeng Meng Minhao Zou Siyang Leng Institute of AI and Robotics Academy for Engineering and Technology Fudan University Shanghai China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Current Graph Contrastive Learning (GCL) methods primarily focus on adapting data augmentation techniques from Computer Vision (CV) or Natural Language Processing (NLP) domains. These techniques typically involve modifying input data via node sampling, edge perturbation, or graph structure perturbation. Alternatively, they may adjust the contrastive loss function by increasing or decreasing positive and negative samples based on graph properties. However, few GCL methods have explored designing and discussing the structure of Graph Neural Networks (GNNs) within GCL, despite the significant impact different GNN structures can have on self-supervised GCL performance. Motivated by this gap in research, our paper differs from the approach of most previous methods, designing a Multi-hop Self-augmented GCL method (MSGCL) based on the inherent structural characteristics of GNNs. This method leverages the intrinsic properties of the GNN model structure, utilizing multi-hop information for self-augmentation to generate enhanced views. The approach is simple yet effective, preserving the original graph structure information without resorting to complex and potentially unstable graph structure augmentation methods. We validate this method across five datasets under different labeling conditions. The experimental results indicate that our method surpasses the other advanced methods, even when only a limited number of labels are available. This suggests potential widespread applications in the field of graph signal processing.

关键词： Computer vision Filters Perturbation methods Contrastive learning Computer architecture Data augmentation Graph neural networks Natural language processing Labeling Speech processing

来源：评论

学校读者我要写书评

暂无评论

Low intensity trans-spinal focused ultrasound reduces mechanical sensitivity and suppresses spinal microglia activation in rats with chronic constriction injury

引用

Bioelectronic Medicine 2025年第1期11卷 1-10页

作者： Song, Weiguo Giannotti, Alice Bekiaridou, Alexandra Bloom, Ona Zanos, Stavros Feinstein Institutes for Medical Research 350 Community Dr ManhassetNY11030 United States The Biorobotics Institute and Department of Excellence in Robotics and AI Scuola Superiore Sant’Anna Pisa Italy Elmezzi Graduate School of Molecular Medicine 350 Community Dr ManhassetNY11030 United States

Low intensity, trans-spinal focused ultrasound (tsFUS) is a noninvasive neuromodulation approach that has been shown to modulate spinal circuit excitability in healthy rats. Here, we evaluated the potential of tsFUS for alleviating neuropathic pain by testing it in a chronic constriction injury (CCI) model. Male rats underwent CCI of the left sciatic nerve and then received tsFUS (2 kHz pulse repetition frequency;40% duty cycle) or sham stimulation, targeted at spinal segment level L5 for 3 min daily over three days. As expected, CCI causes significant reduction of von Frey Threshold (vFT), a measure of mechanical sensitivity. We found that tsFUS treatment is associated with increased vFT compared to sham;this increase persists beyond the duration of treatment, through days 4 to 23 post-CCI. In spinal cords of tsFUS-treated animals, counts of spinal microglia (Iba1 + cells) and of activated, pro-inflammatory microglia (Iba1 + /CD86 + cells), are reduced compared to sham-treated animals. This reduction in microglia counts is limited to the insonified side of the spinal cord, ipsilateral to CCI. These findings suggest that tsFUS may be a promising approach for treatment of neuropathic pain at early stages, possibly by attenuating the development of microglial-driven inflammation. © The Author(s) 2025.

关键词： Rats

来源：评论

学校读者我要写书评

暂无评论

Mechanics of origami/kirigami structures and metamaterials

引用

Theoretical and Applied Mechanics Letters 2025年第4期15卷

作者： Fang, Hongbin Wu, Haiping Institute of AI and Robotics Fudan University Shanghai200433 China

来源：评论

学校读者我要写书评

暂无评论

InstantPose: Zero-Shot Instance-Level 6D Pose Estimation from a Single View

引用

IEEE robotics and Automation Letters 2025年第6期10卷 6023-6030页

作者： Felice, Francesco Di Remus, Alberto Gasperini, Stefano Busam, Benjamin Ott, Lionel Thalhammer, Stefan Tombari, Federico Avizzano, Carlo Alberto Scuola Superiore sant'anna Department of Excellence in Robotics & AI Mechanical Intelligence Institute Pisa Italy Leonardo S.p.A. Robotics Laboratory Innovation Hub & Intellectual Property Genoa Italy Technical University of Munich TUM School of Computation Information and Technology Germany ETH Zurich Department of Mechanical and Process Engineering Autonomous Systems Lab Switzerland UAS Technikum Vienna Industrial Engineering Department Austria Google Zurich Switzerland

Object pose estimation using visual data is crucial for robotic interaction with the environment. Many existing instance-level methods are restricted by their requirements for 3D CAD models or multiple object views, which limits their flexibility and generalizability. Overcoming this limitation is critical to enhance the adaptability of pose estimation systems. In this work, a novel pipeline that leverages recent advances in reconstruction techniques is presented to address these challenges. To this end, Large Reconstruction Models (LRM) represent an advanced neural architecture capable of generating 3D object models from a limited set of views. Nevertheless, the resulting 3D models often lack relevant geometric and texture details due to insufficient input information. This research presents InstantPose, an innovative zero-shot instance-level pose estimation method that, building upon LRM, can determine the pose of unseen objects using as little as a single unposed RGB reference and RGB-D query images. Extensive experiments demonstrate that InstantPose achieves remarkable performance in object pose estimation on the YCB-V dataset, compared to methods conceived to rely on a geometrically perfect object's model. Furthermore, the 6D pose provided through the presented approach facilitates successful object grasping, highlighting its practical utility in robotic manipulation tasks. © 2016 IEEE.

关键词： 3D reconstruction

来源：评论

学校读者我要写书评

暂无评论

Diffuse-CLoC: Guided Diffusion for Physics-based Character Look-ahead Control

arXiv

引用

arXiv 2025年

作者： Huang, Xiaoyu Truong, Takara Zhang, Yunbo Yu, Fangzhou Sleiman, Jean Pierre Hodgins, Jessica Sreenath, Koushil Farshidian, Farbod UC Berkeley United States Stanford United States Robotics and AI Institute United States

We present Diffuse-CLoC, a guided diffusion framework for physics-based look-ahead control that enables intuitive, steerable, and physically realistic motion generation. While existing kinematics motion generation with diffusion models offer intuitive steering capabilities with inference-time conditioning, they often fail to produce physically viable motions. In contrast, recent diffusion-based control policies have shown promise in generating physically realizable motion sequences, but the lack of kinematics prediction limits their steerability. Diffuse-CLoC addresses these challenges through a key insight: modeling the joint distribution of states and actions within a single diffusion model makes action generation steerable by conditioning it on the predicted states. This approach allows us to leverage established conditioning techniques from kinematic motion generation while producing physically realistic motions. As a result, we achieve planning capabilities without the need for a high-level planner. Our method handles a diverse set of unseen long-horizon downstream tasks through a single pre-trained model, including static and dynamic obstacle avoidance, motion in-betweening, and task-space control. Experimental results show that our method significantly outperforms the traditional hierarchical framework of high-level motion diffusion and low-level tracking. Copyright © 2025, The Authors. All rights reserved.

关键词：

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：