检索结果-内蒙古大学图书馆

Guiding reinforcement learning with shaping rewards provided by the vision–language model

Engineering Applications of Artificial Intelligence 2025年 155卷

作者： Wang, Kaiyi Zhao, Yongjia He, Yichen Dai, Shuling Zhang, Ning Yang, Minghao China's State Key Laboratory of Virtual Reality Technology and Systems Beihang University Beijing100191 China Jiangxi Research Institute Beihang University Nanchang330096 China Institute of Intelligent Information Processing Beihang University Beijing100191 China Institute of Automation Chinese Academy of Sciences Beijing100190 China

Enabling robots to learn manipulation tasks is a practical engineering application, and reinforcement learning is one of the key artificial intelligence methods to achieve it, but reinforcement learning always faces the trade-off between the ease of designing reward functions and the ease of learning from rewards. Reward shaping provides a solution and recent works have designed rewards using images and language descriptions, which is a simple and convenient shaping method for non-expert users. However, some of them only adopt the pretrained model without fine-tuning, while others train the reward model with absolute score labels, which all have difficulties in capturing the spatial relationships within the images, so the performance of reward shaping models is limited. In this work, we propose a novel reward shaping method to generate additional rewards from task descriptions and scene images. We utilize the pretrained vision–language model as backbone for efficient cross-modal information fusion and design a downstream task trained with pair-wise comparison for reward shaping. Extensive experiments are conducted to demonstrate the effectiveness of each component in the method. The approach is validated in the Meta-World environment and the results demonstrate that it outperforms standard reinforcement learning and the existing work in terms of the policy learning efficiency. © 2025 Elsevier Ltd

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Structural properties and generative model of non-giant connected components in social networks

引用

Science China Earth Sciences 2016年第12期59卷 229-234页

作者： Jianwei NIU Lei WANG State Key Laboratory of Virtual Reality Technology and Systems School of Computer Science and EngineeringBeihang University

Most previous studies have mainly focused on the analyses of one entire network(graph) or the giant connected components of networks. In this paper, we investigate the disconnected components(non-giant connected component) of some real social networks, and report some interesting discoveries about structural properties of disconnected components. We study three diverse, real networks and compute the significance profile of each component. We discover some similarities in the local structure between the giant connected component and disconnected components in diverse social networks. Then we discuss how to detect network attacks based on the local structure properties of networks. Furthermore, we propose an empirical generative model called i Friends to generate networks that follow our observed patterns.

关键词： disconnected components giant connected component structural properties significance profile generative model

来源：评论

学校读者我要写书评

暂无评论

A guidance law for UAV autonomous aerial refueling based on the iterative computation method

引用

Chinese Journal of Aeronautics 2014年第4期27卷 875-883页

作者： Luo Delin Xie Rongzeng Duan Haibin Department of Automation Xiamen University State Key Laboratory of Virtual Reality Technology and Systems Beihang University

The rendezvous and formation problem is a significant part for the unmanned aerial vehicle（UAV） autonomous aerial refueling（AAR） technique. It can be divided into two major phases： the long-range guidance phase and the formation phase. In this paper, an iterative computation guidance law（ICGL） is proposed to compute a series of state variables to get the solution of a control variable for a UAV conducting rendezvous with a tanker in AAR. The proposed method can make the control variable converge to zero when the tanker and the UAV receiver come to a formation flight eventually. For the long-range guidance phase, the ICGL divides it into two sub-phases： the correction sub-phase and the guidance sub-phase. The two sub-phases share the same iterative process. As for the formation phase, a velocity coordinate system is created by which control accelerations are designed to make the speed of the UAV consistent with that of the *** simulation results demonstrate that the proposed ICGL is effective and robust against wind disturbance.

关键词： Autonomous aerial refueling Aerial rendezvous Formation control Guidance law Unmanned aerial vehicle

来源：评论

学校读者我要写书评

暂无评论

Hierarchical Visual Attention Model for Saliency Detection Inspired by Avian Visual Pathways

引用

IEEE/CAA Journal of Automatica Sinica 2019年第2期6卷 540-552页

作者： Xiaohua Wang Haibin Duan the State Key Laboratory of Virtual Reality Technology and Systems School of Automation Science and Electrical Engineering Beihang University the State Key Laboratory of Virtual Reality Technology and Systems School of Automation Science and Electrical EngineeringBeihang University the Peng Cheng Laboratory

Visual attention is a mechanism that enables the visual system to detect potentially important objects in complex environment. Most computational visual attention models are designed with inspirations from mammalian visual ***, electrophysiological and behavioral evidences indicate that avian species are animals with high visual capability that can process complex information accurately in real time. Therefore,the visual system of the avian species, especially the nuclei related to the visual attention mechanism, are investigated in this paper. Afterwards, a hierarchical visual attention model is proposed for saliency detection. The optic tectum neuron responses are computed and the self-information is used to compute primary saliency maps in the first hierarchy. The "winner-takeall" network in the tecto-isthmal projection is simulated and final saliency maps are estimated with the regularized random walks ranking in the second hierarchy. Comparison results verify that the proposed model, which can define the focus of attention accurately, outperforms several state-of-the-art *** study provides insights into the relationship between the visual attention mechanism and the avian visual pathways. The computational visual attention model may reveal the underlying neural mechanism of the nuclei for biological visual attention.

关键词： Avian visual pathways bio-inspired saliency detection visual attention

来源：评论

学校读者我要写书评

暂无评论

Anomaly detection by exploiting the tracking trajectory in surveillance videos

引用

Science China(Information Sciences) 2020年第5期63卷 197-199页

作者： Zixuan XUE Wei WU Faculty of Information Technology Beijing University of Technology State Key Laboratory of Virtual Reality Technology and Systems Beihang University

Surveillance systems have become increasingly ubiquitous, which has led to a requirement to detect anomalies for efficiently preventing terrorism and reducing crimes. The increasing number of surveillance networks has imposed major technical challenges on intelligent anomaly detection because of the inconsistent appearance of pedestrians owing to posture deformation and clutter in video

关键词： Anomaly detection by exploiting the tracking trajectory in surveillance videos frame

来源：评论

学校读者我要写书评

暂无评论

Semantic part segmentation of single-view point cloud

引用

Science China(Information Sciences) 2020年第12期63卷 251-253页

作者： Haotian PENG Bin ZHOU Liyuan YIN Kan GUO Qinping ZHAO State Key Laboratory of Virtual Reality Technology and Systems School of Computer Science and EngineeringBeihang University

As a classic topic in computer graphics, the semantic part segmentation of 3D data is helpful for3D part-level editing and modeling. Single-views point cloud is the raw format of 3D data. Giving each point a semantic annotation in single-view point cloud, i.e., single-view point cloud semantic part segmentation, is meaningful and challenging.

关键词： computers and graphics single-view point cloud annotation transfer semantic part segmentation category independent matching point clouds registration

来源：评论

学校读者我要写书评

暂无评论

Localizing object parts in 3D from a single image

引用

Science China(Information Sciences) 2019年第7期62卷 217-219页

作者： Shen YIN Bin ZHOU Mingjia YANG Yu ZHANG State Key Laboratory of Virtual Reality Technology and Systems School of Computer Science and EngineeringBeihang University

Object localization in 3D from 2D images is an important computer vision problem that enables modern robotic vision systems to interact properly with the objects present in the real world. Owing to its significance, techniques for recovering the 6-DOF pose and dimensions of objects from images has received increasing attention in recent years [1–3]. However, to the best of our knowledge,the existing approaches all focus on object-level inference instead of analyzing object parts. Most of

关键词： pro Localizing object parts in 3D from a single image

来源：评论

学校读者我要写书评

暂无评论

Multi-Grained Guided Diffusion for Quantity-Controlled Remote Sensing Object Generation

引用

IEEE Geoscience and Remote Sensing Letters 2025年 22卷

作者： Yu, Zhiping Liu, Chenyang Zhong, Chuyu Zou, Zhengxia Shi, Zhenwei Beihang University School of Astronautics Department of Aerospace Intelligent Science and Technology Beijing100191 China Beihang University State Key Laboratory of Virtual Reality Technology and Systems Beijing100191 China Shanghai Artificial Intelligence Laboratory Shanghai200232 China Beihang University Shen Yuan Honors College Beijing100191 China

Accurate object counts represent essential semantical information in remote sensing imagery, significantly impacting applications such as traffic monitoring and urban planning. Despite the recent advances in text-to-image (T2I) generation in remote sensing, existing methods still face challenges in precisely controlling the number of object instances in generated images. To address this challenge, we propose a novel method, multi-grained guided diffusion (MGDiff). During training, unlike previous methods that relied solely on latent-space noise constraints, MGDiff imposes constraints at three distinct granularities: latent pixel, global counting, and spatial distribution. The multi-grained guidance mechanism matches the quantity prompts with object spatial layouts in the feature space, enabling our model to achieve precise control over object quantities. To benchmark this new task, we present Levir-QCG, a dataset comprising 10504 remote sensing images across five object categories, annotated with precise object counts and segmentation masks. We conducted extensive experiments to benchmark our method against previous methods on the Levir-QCG dataset. Compared to previous models, the MGDiff achieves an approximately +40% improvement in counting accuracy while maintaining higher visual fidelity and strong zero-shot generalization. To the best of our knowledge, this is the first work to research accurate object quantity control in remote sensing T2I generation. © 2004-2012 IEEE.

关键词： Semantic Segmentation

来源：评论

学校读者我要写书评

暂无评论

Haptics-equiped interactive PCI simulation for patient-specific surgery training and rehearsing

引用

Science China(Information Sciences) 2016年第10期59卷 233-238页

作者： Shuai LI Qing XIA Aimin HAO Hong QIN Qinping ZHAO State Key Laboratory of Virtual Reality Technology and Systems Beihang University Department of Computer Science Stony Brook University

Despite the long history of medical simulations, suffering from the patient-specific heterogeneous heart physiological structure and complex intravascular procedures, it is still challenging for patient-specific percutaneous coronary intervention(PCI) surgery simulation. In this paper, we advocate a haptics-equiped interactive prototype system towards PCI surgeons training and patient-specific surgery rehearsing, which can afford trainees the opportunity to approximately experience the entire PCI procedures and customized emergency cases that might occur in common clinical settings. The full simulation covers tissue deformation, catheter and wire simulation, X-ray simulation, haptic feedback, and 3D realistic rendering, which in all give rise to the integrated physical, visual, haptic, and procedural realism. Our system can accommodate various comprehensive operations involved in PCI-related procedures, including feeding wires, releasing stents, injecting contrast medium, simulating X-ray, bleeding, etc. Moreover, our system framework is fully built upon CUDA, and thus can achieve real-time performance even on a common desktop. The high-fidelity, real-time efficiency and stableness of our system show great potentials for its practical applications in clinical training fields.

关键词： percutaneous coronary intervention(PCI) PCI surgery simulation patient-specific surgery rehearsing physically-based modeling haptics

来源：评论

学校读者我要写书评

暂无评论

Lighting virtual objects in a single image via coarse scene understanding

引用

Science China(Information Sciences) 2014年第9期57卷 58-71页

作者： CHEN XiaoWu JIN Xin WANG Ke State Key Laboratory of Virtual Reality Technology and Systems School of Computer Science and EngineeringBeihang University

Achieving convincing visual consistency between virtual objects and a real scene mainly relies on the lighting effects of virtual-real composition *** problem becomes more challenging in lighting virtual objects in a single real ***,scene understanding from a single image has made great *** estimated geometry,semantic labels and intrinsic components provide mostly coarse information,and are not accurate enough to re-render the whole ***,carefully integrating the estimated coarse information can lead to an estimate of the illumination parameters of the real *** present a novel method that uses the coarse information estimated by current scene understanding technology to estimate the parameters of a ray-based illumination model to light virtual objects in a real *** key idea is to estimate the illumination via a sparse set of small 3D surfaces using normal and semantic *** coarse shading image obtained by intrinsic image decomposition is considered as the irradiance of the selected small *** virtual objects are illuminated by the estimated illumination *** results show that our method can convincingly light virtual objects in a single real image,without any pre-recorded 3D geometry,reflectance,illumination acquisition equipment or imaging information of the image.

关键词： illumination estimation lighting virtual object single image scene understanding photo editing

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：