检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Ægidius, Sebastian Hadjivelichkov, Dennis Jiao, Jianhao Embley-Riches, Jonathan Kanoulas, Dimitrios Robot Perception and Learning Lab Department of Computer Science University College London Gower Street LondonWC1E 6BT United Kingdom Archimedes RC Athens Greece

Understanding the traversability of terrain is essential for autonomous robot navigation, particularly in unstructured environments such as natural landscapes. Although traditional methods, such as occupancy mapping, provide a basic framework, they often fail to account for the complex mobility capabilities of some platforms such as legged robots. In this work, we propose a method for estimating terrain traversability by learning from demonstrations of human walking. Our approach leverages dense, pixel-wise feature embeddings generated using the DINOv2 vision Transformer model, which are processed through an encoder-decoder MLP architecture to analyze terrain segments. The averaged feature vectors, extracted from the masked regions of interest, are used to train the model in a reconstruction-based framework. By minimizing reconstruction loss, the network distinguishes between familiar terrain with a low reconstruction error and unfamiliar or hazardous terrain with a higher reconstruction error. This approach facilitates the detection of anomalies, allowing a legged robot to navigate more effectively through challenging terrain. We run real-world experiments on the ANYmal legged robot both indoor and outdoor to prove our proposed method. The code is open-source, while video demonstrations can be found on our website: https://***/STEPP/ © 2025, CC BY.

关键词： Open source software

来源：评论

学校读者我要写书评

暂无评论

FrontierNet: learning Visual Cues to Explore

引用

IEEE robotics and Automation Letters 2025年第7期10卷 6576-6583页

作者： Sun, Boyang Chen, Hanzhi Leutenegger, Stefan Cadena, Cesar Pollefeys, Marc Blum, Hermann ETH Zurich Computer Vision and Geometry Group Zurich 8092 Switzerland Technical University of Munich Mobile Robotics Lab München 80333 Germany ETH Zurich Mobile Robotics Lab Zurich 8092 Switzerland ETH Zurich Robotic Systems Lab Zurich 8092 Switzerland AI Lab Microsoft Mixed Reality Zurich 8038 Switzerland University of Bonn Lamarr Institute for ML and AI Robot Perception and Learning Lab Bonn 53115 Germany

Exploration of unknown environments is crucial for autonomous robots;it allows them to actively reason and decide on what new data to acquire for different tasks, such as mapping, object discovery, and environmental assessment. Existing solutions, such as frontier-based exploration approaches, rely heavily on 3D map operations, which are limited by map quality and, more critically, often overlook valuable context from visual cues. This work aims at leveraging 2D visual cues for efficient autonomous exploration, addressing the limitations of extracting goal poses from a 3D map. We propose a visual-only frontier-based exploration system, with FrontierNet as its core component. FrontierNet is a learning-based model that (i) proposes frontiers, and (ii) predicts their information gain, from posed RGB images enhanced by monocular depth priors. Our approach provides an alternative to existing 3D-dependent goal-extraction approaches, achieving a 15% improvement in early-stage exploration efficiency, as validated through extensive simulations and real-world experiments. © 2016 IEEE.

关键词： Deep learning Motion and Path Planning Perception and Autonomy

来源：评论

学校读者我要写书评

暂无评论

O-PRESS: Boosting OCT axial resolution with Prior guidance, Recurrence, and Equivariant Self-Supervision

引用

MEDICAL IMAGE ANALYSIS 2025年 99卷 103319页

作者： Li, Kaiyan Yang, Jingyuan Liang, Wenxuan Li, Xingde Zhang, Chenxi Chen, Lulu Wu, Chan Zhang, Xiao Xu, Zhiyan Wang, Yueling Meng, Lihui Zhang, Yue Chen, Youxin Zhou, S. Kevin Univ Sci & Technol China USTC Sch Biomed Engn Div Life Sci & Med Hefei 230026 Anhui Peoples R China USTC Ctr Med Imaging Robot Analyt Comp & Learning MIRAC Suzhou Inst Adv Res Suzhou 215123 Jiangsu Peoples R China Chinese Acad Med Sci Dept Ophthalmol Peking Union Med Coll Hosp Beijing 100730 Peoples R China Chinese Acad Med Sci Key Lab Ocular Fundus Dis Beijing 100730 Peoples R China Johns Hopkins Univ Dept Biomed Engn Baltimore MD 21287 USA USTC Key Lab Precis & Intelligent Chem Hefei 230026 Anhui Peoples R China Chinese Acad Sci Inst Comp Technol Key Lab Intelligent Informat Proc Beijing 100190 Peoples R China Univ Sci & Technol China Sch Phys Sci Hefei 230026 Anhui Peoples R China

Optical coherence tomography (OCT) is a noninvasive technology that enables real-time imaging of tissue microanatomies. The axial resolution of OCT is intrinsically constrained by the spectral bandwidth of the employed light source while maintaining a fixed center wavelength for a specific application. Physically extending this bandwidth faces strong limitations and requires a substantial cost. We present a novel computational approach, called as O-PRESS, , for boosting the axial resolution of O CT with P rior guidance, a R ecurrent mechanism, and E quivariant S elf-Supervision. S upervision. Diverging from conventional deconvolution methods that rely on physical models or data-driven techniques, our method seamlessly integrates OCT modeling and deep learning, enabling us to achieve real-time axial-resolution enhancement exclusively from measurements without a need for paired images. Our approach solves two primary tasks of resolution enhancement and noise reduction with one treatment. Both tasks are executed in a self-supervised manner, with equivariance imaging and free space priors guiding their respective processes. Experimental evaluations, encompassing both quantitative metrics and visual assessments, consistently verify the efficacy and superiority of our approach, which exhibits performance on par with fully supervised methods. Importantly, the robustness of our model is affirmed, showcasing its dual capability to enhance axial resolution while concurrently improving the signal-to-noise ratio.

关键词： Retinal imaging OCT image reconstruction Resolution Self-supervised learning Prior guidance

来源：评论

学校读者我要写书评

暂无评论

Multimodal Emotion Fusion Mechanism and Empathetic Responses in Companion robots

引用

IEEE TRANSACTIONS ON COGNITIVE AND DEVELOPMENTAL SYSTEMS 2025年第2期17卷 271-286页

作者： Liu, Xiaofeng Lv, Qincheng Li, Jie Song, Siyang Cangelosi, Angelo Hohai Univ Coll Artificial Intelligence & Automat Changzhou 213022 Peoples R China Hohai Univ Coll Informat Sci & Engn Changzhou 213022 Peoples R China Nanjing Med Univ Engn Res Ctr Intelligent Theranost Technol & Instr Minist Educ Nanjing 211166 Peoples R China Nanjing Med Univ Sch Biomed Engn & Informat Nanjing 211166 Peoples R China Univ Cambridge Dept Comp Sci & Technol Affect Intelligence & Robot Lab Cambridge CB3 0FT Cambs England Univ Manchester Machine Learning & Optimisat Manchester M13 9RL Lancs England

The ability of humanoid robots to exhibit empathetic facial expressions and provide corresponding responses is essential for natural human-robot interaction. To enhance this, we integrate the GPT3.5 model with a facial expression recognition model, creating a multimodal emotion recognition system. Additionally, we address the challenge of realistically mimicking human facial expressions by designing the physical structure of a humanoid robot. Initially, we develop a humanoid robot capable of adjusting the positions of its facial organs and neck through servo displacement to achieve more natural facial expressions. Subsequently, to overcome the current limitation where emotional interaction robots struggle to accurately recognize user emotions, we introduce a coupled generative pretrained transformer (GPT)-based multimodal emotion recognition method that utilizes both text and images, thereby enhancing the robot's emotion recognition accuracy. Finally, we integrate the GPT-3.5 model to generate empathetic responses based on recognized user emotional states and language text, which are then mapped onto the robot to enable empathetic expressions that can achieve a more comfortable human-machine interaction experience. Experimental results on benchmark databases demonstrate that the performance of the coupled GPT-based multimodal emotion recognition method using text and images outperforms other approaches, and it possesses unique empathetic response capabilities relative to alternative methods.

关键词： Emotion recognition Face recognition Analytical models Brain modeling Accuracy Electronic mail Empathetic capabilities GPT-35 humanoid robot multimodal emotion recognition multimodal emotion recognition

来源：评论

学校读者我要写书评

暂无评论

learning manipulation skills from a single demonstration

引用

INTERNATIONAL JOURNAL OF robotICS RESEARCH 2018年第1期37卷 137-154页

作者： Englert, Peter Toussaint, Marc Univ Stuttgart Machine Learning & Robot Lab Stuttgart Germany

We consider the scenario where a robot is demonstrated a manipulation skill once and should then use only a few trials on its own to learn to reproduce, optimize, and generalize that same skill. A manipulation skill is generally a high-dimensional policy. To achieve the desired sample efficiency, we need to exploit the inherent structure in this problem. With our approach, we propose to decompose the problem into analytically known objectives, such as motion smoothness, and black-box objectives, such as trial success or reward, depending on the interaction with the environment. The decomposition allows us to leverage and combine (i) constrained optimization methods to address analytic objectives, (ii) constrained Bayesian optimization to explore black-box objectives, and (iii) inverse optimal control methods to eventually extract a generalizable skill representation. The algorithm is evaluated on a synthetic benchmark experiment and compared with state-of-the-art learning methods. We also demonstrate the performance on real-robot experiments with a PR2.

关键词： Combined optimization and learning reinforcement learning imitation learning manipulation skills

来源：评论

学校读者我要写书评

暂无评论

learning to Search in Task and Motion Planning With Streams

引用

IEEE robotICS AND AUTOMATION LETTERS 2023年第4期8卷 1983-1990页

作者： Khodeir, Mohamed Agro, Ben Shkurti, Florian Univ Toronto Robot Vis & Learning Lab Robot Inst Toronto ON M5S Canada

Task and motion planning problems in robotics combine symbolic planning over discrete task variables with motion optimization over continuous state and action variables. Recent works such as PDDLStream [Garrett et al. 2020)] have focused on optimistic planning with an incrementally growing set of objects until a feasible trajectory is found. However, this set is exhaustively expanded in a breadth-first manner, regardless of the logical and geometric structure of the problem at hand, which makes long-horizon reasoning with large numbers of objects prohibitively time-consuming. To address this issue, we propose a geometrically informed symbolic planner that expands the set of objects and facts in a best-first manner, prioritized by a Graph Neural Network that is learned from prior search computations. We evaluate our approach on a diverse set of problems and demonstrate an improved ability to plan in difficult scenarios. We also apply our algorithm on a 7DOF robotic arm in block-stacking manipulation tasks.

关键词： Planning Task analysis Optimized production technology Generators Cognition Stacking Search problems Task and motion planning integrated planning and learning manipulation planning

来源：评论

学校读者我要写书评

暂无评论

Fast motion planning from experience: trajectory prediction for speeding up movement generation

引用

AUTONOMOUS robotS 2013年第1-2期34卷 111-127页

作者： Jetchev, Nikolay Toussaint, Marc FU Berlin Machine Learning & Robot Lab D-14195 Berlin Germany

Trajectory planning and optimization is a fundamental problem in articulated robotics. Algorithms used typically for this problem compute optimal trajectories from scratch in a new situation. In effect, extensive data is accumulated containing situations together with the respective optimized trajectories-but this data is in practice hardly exploited. This article describes a novel method to learn from such data and speed up motion generation, a method we denote tajectory pediction. The main idea is to use demonstrated optimal motions to quickly predict appropriate trajectories for novel situations. These can be used to initialize and thereby drastically speed-up subsequent optimization of robotic movements. Our approach has two essential ingredients. First, to generalize from previous situations to new ones we need a situation descriptor-we construct features for such descriptors and use a sparse regularized feature selection approach to improve generalization. Second, the transfer of previously optimized trajectories to a new situation should not be made in joint angle space-we propose a more efficient task space transfer. We present extensive results in simulation to illustrate the benefits of the new method, and demonstrate it also with real robot hardware. Our experiments in diverse tasks show that we can predict good motion trajectories in new situations for which the refinement is much faster than an optimization from scratch.

关键词： Motion planning Machine learning Motion representation Articulated robotics

来源：评论

学校读者我要写书评

暂无评论

An intelligent Q-learning-based tree routing method in underwater acoustic sensor networks

引用

ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE 2025年 152卷

作者： Khoshvaght, Parisa Haider, Amir Rahmani, Amir Masoud Altulyan, May Zaidi, Monji Mohamed Yousefpoor, Mohammad Sadegh Yousefpoor, Efat Hosseinzadeh, Mehdi Duy Tan Univ Inst Res & Dev Da Nang Vietnam Duy Tan Univ Sch Engn & Technol Da Nang Vietnam Sejong Univ Dept Artificial Intelligence & Robot Seoul South Korea Natl Yunlin Univ Sci & Technol Future Technol Res Ctr Yunlin Taiwan Prince Sattam bin Abdulaziz Univ Coll Comp Engn & Sci Dept Comp Engn Al Kharj Saudi Arabia King Khalid Univ Coll Engn Dept Elect Engn Abha 61421 Saudi Arabia King Khalid Univ Ctr Engn & Technol Innovat Abha 61421 Saudi Arabia Lebanese French Univ Ctr Res & Strateg Studies Erbil Kurdistan Reg Iraq Gachon Univ Sch Comp Pattern Recognit & Machine Learning Lab Seongnam 13120 South Korea

Nowadays, underwater acoustic sensor networks (UASNs) have emerged as an advanced and promising technology for developing various underwater applications. However, several routing protocols have been suggested for these networks in recent years. This subject is still facing many challenges such as low propagation speed, low bandwidth, and energy restrictions. To solve the above-mentioned challenges, this paper proposes an intelligent Q-learning-based tree routing method called QTRU for underwater acoustic sensor networks. The proposed scheme includes a network bootstrapping process to be aware of local network topology and calculate the neighboring table related to each node. QTRU also contains a Q-learning-based tree construction process to transmit data from sensor nodes to the sink node. In the routing tree construction process, the reward function in the Q-learning algorithm consists of four parameters, including the depth of the node, remaining energy, successful transmission probability, and the size of the candidate set. In addition, to calculate the state set in the Q-learning algorithm, each node carries out two screening operations on its neighboring nodes on the network. The first screening operation ensure that each node in the routing tree has the least number of hops to the sink node. The second screening operation is to avoid the formation of routing loops between sensor nodes and ensure a tree-based network topology. QTRU also designs a recovery mechanism and allows the sensor nodes present in the void area to select their best parent node in the routing tree. Finally, QTRU is implemented in network simulator version 2 (NS2), and its results are compared with three routing methods, namely reinforcement learning-based opportunistic routing protocol (RLOR), Q-learning-based multi-level routing protocol (MURAO), and energy-efficient depth-based routing protocol (EE-DBR). These results show that QTRU improves the packet delivery rate (about 8.94%), data

关键词： Underwater acoustic sensor networks Routing Artificial intelligence Q-learning Internet of Underwater Things Machine learning

来源：评论

学校读者我要写书评

暂无评论

Bathymetric Reconstruction From Sidescan Sonar With Deep Neural Networks

引用

IEEE JOURNAL OF OCEANIC ENGINEERING 2023年第2期48卷 372-383页

作者： Xie, Yiping Bore, Nils Folkesson, John Royal Inst Technol Robot Percept & Learning Lab SE-10044 Stockholm Sweden

In this article, we propose a novel data-driven approach for high-resolution bathymetric reconstruction from sidescan. Sidescan sonar intensities as a function of range do contain some information about the slope of the seabed. However, that information must be inferred. In addition, the navigation system provides the estimated trajectory, and normally, the altitude along this trajectory is also available. From these, we obtain a very coarse seabed bathymetry as an input. This is then combined with the indirect but high-resolution seabed slope information from the sidescan to estimate the full bathymetry. This sparse depth could be acquired by single-beam echo sounder, Doppler velocity log, and other bottom tracking sensors or bottom tracking algorithm from sidescan itself. In our work, a fully convolutional network is used to estimate the depth contour and its aleatoric uncertainty from the sidescan images and sparse depth in an end-to-end fashion. The estimated depth is then used together with the range to calculate the point's three-dimensional location on the seafloor. A high-quality bathymetric map can be reconstructed after fusing the depth predictions and the corresponding confidence measures from the neural networks. We show the improvement of the bathymetric map gained by using sparse depths with sidescan over estimates with sidescan alone. We also show the benefit of confidence weighting when fusing multiple bathymetric estimates into a single map.

关键词： Bathymetric mapping data-driven neural network sidescan sonar (SSS)

来源：评论

学校读者我要写书评

暂无评论

Modeling and Simulation of Sidescan Using Conditional Generative Adversarial Network

引用

IEEE JOURNAL OF OCEANIC ENGINEERING 2021年第1期46卷 195-205页

作者： Bore, Nils Folkesson, John Royal Inst Technol Robot Percept & Learning Lab SE-10044 Stockholm Sweden

Sidescan sonar has been used for maritime surveys since the mid-20th century. Due to its wide swath coverage and the sharpness of the produced images, it is an invaluable tool still to this day. When simulating sidescan data, there is a tradeoff between the quality of the produced images, the fidelity of the environment simulation, and the complexity of the sidescan model. In this article, we propose data-driven models as a way of removing some of this tradeoff. Using recently proposed conditional generative adversarial nets, we create a generative model that takes the environment as input, and produces realistic sidescan measurements. We show that end-to-end learning of flexible models allows simulating more complex sidescan data than would otherwise be possible given only geometric bathymetry.

关键词： cGAN data-driven sidescan sonar

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：