检索结果-内蒙古大学图书馆

16th European Conference on Computer Vision, ECCV 2020

作者： Xie, Haozhe Yao, Hongxun Zhou, Shangchen Mao, Jiageng Zhang, Shengping Sun, Wenxiu State Key Laboratory of Robotics and System Harbin Institute of Technology Harbin China Faculty of Computing Harbin Institute of Technology Harbin China SenseTime Research Shenzhen China Nanyang Technological University Singapore Singapore The Chinese University of Hong Kong Hong Kong SAR China Peng Cheng Laboratory Shenzhen China SenseTime Research Hong Kong SAR China

ISBN: (纸本)9783030585440

Estimating the complete 3D point cloud from an incomplete one is a key problem in many vision and robotics applications. Mainstream methods (e.g., PCN and TopNet) use Multi-layer Perceptrons (MLPs) to directly process point clouds, which may cause the loss of details because the structural and context of point clouds are not fully considered. To solve this problem, we introduce 3D grids as intermediate representations to regularize unordered point clouds and propose a novel Gridding Residual Network (GRNet) for point cloud completion. In particular, we devise two novel differentiable layers, named Gridding and Gridding Reverse, to convert between point clouds and 3D grids without losing structural information. We also present the differentiable Cubic Feature Sampling layer to extract features of neighboring points, which preserves context information. In addition, we design a new loss function, namely Gridding Loss, to calculate the L1 distance between the 3D grids of the predicted and ground truth point clouds, which is helpful to recover details. Experimental results indicate that the proposed GRNet performs favorably against state-of-the-art methods on the ShapeNet, Completion3D, and KITTI benchmarks. © 2020, Springer Nature Switzerland AG.

关键词： Computer vision

来源：评论

学校读者我要写书评

暂无评论

An adaptable robotic system for assembling irregular earphone parts on factory automation

An adaptable robotic system for assembling irregular earphon...

引用

2019 IEEE International Conference on robotics and Biomimetics, ROBIO 2019

作者： Zhang, Guiwei Li, Changle Zhang, Xuehe Liu, Gangfeng Liu, Yubin Zhao, Jie Xu, Shuo Harbin Institue of Technology State Key Laboratory of Robotics and System Harbin China

ISBN: (纸本)9781728163215

The assembly of earphone parts in electro acoustic factory automation is a challenging problem for industrial robots. The main difficulties come from the fact that conventional industrial robots perform poorly on rapid-updated earphone parts, which are characterized by small size, variety and irregular shapes. What's worse, the robot cannot adapt to the earphone parts with different poses effectively. In order to tackle above problems, an adaptable robotic system guided by the 3D laser triangulation sensor and six-axis F/T sensor is built in this paper. In addition, to reduce position errors caused by the 3D sensor accidently, an assembly strategy which is implemented by the admittance controller learned from human teaching, is proposed for assembling earphone parts. Specifically, AdaBoost-based GPR method is applied to derive the admittance gain, so that the controller perform effectively. A series of comparative experiments prove that the established system could demonstrate superior performance on the earphone parts with small size, variety and irregular shapes. Furthermore, it is also validated that the assembly strategy could be executed effectively on the earphone parts with different poses. © 2019 IEEE.

关键词： Controllers

来源：评论

学校读者我要写书评

暂无评论

A novel magnetoelastic torque sensor with planar spiral coil probes

引用

Biomimetic Intelligence and robotics 2025年

作者： Zijian Zhang Zitao Wang MingShao Yangyang Dong Fenglei Ni School of Astronautics Nanjing University of Aeronautics and Astronautics Nanjing 210016 China State Key Laboratory of Robotics and Systems Harbin Institute of Technology Harbin 150001 China

Humanoid robot joints require real-time torque detection to provide accurate force feedback information for the control system. To meet the measurement requirements and realize the miniaturization of the sensor, a torque sensor based on the magnetoelastic effect is developed, utilizing planar spiral coils as detection probes. In this work, a planar spiral coil mutual inductance calculation model is established to solve the mutual inductance coefficient, and the mechanical structure and circuit design of the sensor are completed. Finally, a torque loading platform is built to perform calibration experiments, and the hysteresis model is improved to compensate for the hysteresis phenomenon. The calibration results indicate that the sensor shows excellent loaded nonlinearity of 3.08%F.S., unloaded nonlinearity of 2.71%F.S., loaded repeatability of 2.48%F.S., unloaded repeatability of 1.89%F.S. and hysteresis of 1.9%F.S., at a compact probe size of 13.8 × 9 . 9 × 1.8 mm.

关键词： Magnetoelastic effect Torque sensor Planar spiral coil Mutual inductance calculation Hysteresis model

来源：评论

学校读者我要写书评

暂无评论

A Path Planning Algorithm of Spray Robot based on 3D Point Cloud

A Path Planning Algorithm of Spray Robot based on 3D Point C...

引用

International Conference on Communications, Information system and Computer Engineering (CISCE)

作者： Jubo Yang Minxiu Kong Rundong Wang School of Mechatronics Engineering Harbin Institute of Technology State Key Laboratory of Robotics and System Harbin China

The following topics are dealt with: learning (artificial intelligence); feature extraction; neural nets; diseases; convolutional neural nets; computer aided instruction; power engineering computing; teaching; educati... 详细信息

ISBN: (数字)9781728197616

ISBN: (纸本)9781728197623

关键词： Three-dimensional displays Spraying Robot sensing systems Robots Robot kinematics Covariance matrices Gravity

来源：评论

学校读者我要写书评

暂无评论

Design and simulation of an accelerometer allocation scheme for six-dimensional acceleration sensor

Design and simulation of an accelerometer allocation scheme ...

引用

2019 IEEE International Conference on robotics and Biomimetics, ROBIO 2019

作者： Wu, Zhiwei Sun, Yongjun Ni, Fenglei Liu, Yiwei Liu, Hong Harbin Institute of Technology State Key Laboratory of Robotics and System Harbin Heilongjiang Province China China

ISBN: (纸本)9781728163215

In this paper, a novel spatial geometry of sixdimensional acceleration sensor(SDAS) is designed. And the sensor equation is calculated. The reliability of SDAS is improved by adding redundant line accelerometer. The GDOP(Geometric Dilution of Precision) index used in satellite navigation is introduced into SDAS for configuration evaluation. Compared with the classic cube configuration, the three indexes have improved significantly. The relationship between the output error of the SDAS and the installation error of the linear accelerometer is analyzed theoretically. Compared with the proposed layout structure using 12 accelerometers, it has obvious advantages. © 2019 IEEE.

关键词： Accelerometers

来源：评论

学校读者我要写书评

暂无评论

Compliance control of the hydraulic-driven single legged robot

Compliance control of the hydraulic-driven single legged rob...

引用

2019 IEEE International Conference on robotics and Biomimetics, ROBIO 2019

作者： Fan, Yixiao Li, Xu Feng, Haibo Zhang, Songyuan Zhang, Chengbo Fu, Yili Harbin Institute of Technology State Key Laboratory of Robotics and System Harbin Heilongjiang Province150001 China

ISBN: (纸本)9781728163215

This paper presents the force control strategies and a kind of compliance control strategies of the hydraulic- driven unit (HDU) in the legged robots. The model of HDU is established and the transfer function of the HDU force control system is given. The analysis of the force control strategies is introduced according to the characteristic of the HDU system. Based on the force control strategy called velocity feedback control strategy which performs the best, the virtual model control strategy which is a kind of compliance control strategy is presented to enhance the adaptability of the legged robot. Finally, the experiments are taken on the HDU experiment platform to test the performance of the force control strategies and the compliance control strategy. The landing buffer experiments are taken on the single leg robot to verify the effectiveness of the compliance control strategy. In the landing buffer experiments, the impact force of the single leg robot's knee HDU is about 0.33 time of what it is without the compliance control strategy while landing. © 2019 IEEE.

关键词： Compliance control

来源：评论

学校读者我要写书评

暂无评论

Using permanent magnetic array device to detumble and despin defunct spacecraft 5

Using permanent magnetic array device to detumble and despin...

引用

2019 5th International Conference on Applied Materials and Manufacturing technology, ICAMMT 2019

作者： Xie, Zongwu Gong, Pixu Liang, Mengde Zhang, Qi Liu, Hong State Key Laboratory of Robotics and System Harbin Institute of Technology Heilongjiang Province Harbin150001 China

For the mission of detumbling and despinning the space debris, like defunct satellites, it is highly reliable and safe to use the permanent magnetic array device (PMAD). Based on the electromagnetic principle, the target's attitude dynamics is analysed, the electromagnetic torque is studied, the formula of electromagnetic force is simplified, and the stability of the nutation Angle is obtained. The rotation speed and attitude of PMAD are controlled by a dual-manipulator space robot to ensure that the component of the torque on the spin axis is zero. So the target will be only affected by the transverse component of the torque, and then the nutation angle decreases gradually to zero. When mission completed, the dual-manipulator space robot can deorbit or maintain on-orbit servicing, which is a significant exploration. © Published under licence by IOP Publishing Ltd.

关键词： Space debris

来源：评论

学校读者我要写书评

暂无评论

On Modeling Drilling Load in Lunar Regolith Simulant

引用

Chinese Journal of Mechanical Engineering 2018年第1期31卷 41-52页

作者： Qi-Quan Quan Chong-Bin Chen Zong-Quan Deng Jun-Yue Tang De-Wei Tang State Key Laboratory of Robotics and System Harbin Institute of Technology

Drilling and coring, as effective ways to obtain lunar regolith along the longitudinal direction, are widely applied in the lunar sampling field. Conventionally, modeling of drill-soil interaction was divided into soil cutting and screw conveyance processes, ignoring the differences in soil mechanical properties between them. To improve the modeling accuracy, a hypothesis that divides the drill-soil interaction into four parts: cuttings screw conveyance, cuttings extruding, cuttings bulldozing, and in situ simulant cutting, is proposed to establish a novel model based on the passive earth pressure theory. An iterative numerical calculation method is developed to predict the drilling loads. A drilling and coring testbed is developed to conduct experimental tests. Drilling experiments indicate that the drilling loads calculated by the proposed model match well the experimental results. The proposed research provides the instructions to adopt a suitable drilling strategy to match the rotary and penetrating motions, to increase the safety and reliability of drilling control in lunar sampling missions.

关键词： Lunar regolith drilling Lunar regolith simulant Drill-simulant interaction Screw conveyance

来源：评论

学校读者我要写书评

暂无评论

Extrinsic calibration between a stereo system and a 3D LIDAR

Extrinsic calibration between a stereo system and a 3D LIDAR

引用

2019 IEEE International Conference on robotics and Biomimetics, ROBIO 2019

作者： Zhai, Yanwu Feng, Haibo He, Jia Li, Enbo Zhang, Songyuan Fu, Yili Harbin Institute of Technology State Key Laboratory of Robotics and System Harbin Heilongjiang Province150001 China

ISBN: (纸本)9781728163215

As the robot gradually develops from indoor to outdoor, multi-sensor fusion is increasingly being applied to the field of robot perception. Among them, LiDAR and binocular sensors are most widely used. Robots use the complementary and redundant data they provide to perform environmental perception and implement certain functions of the robot, such as SLAM, autonomous obstacle avoidance and so on. However, the data obtained by each sensor are relative to its own coordinate system. Data must be converted to a unified coordinate system before data fusion. Therefore, the calibration of the transformation among sensors is the first problem to be solved in data fusion. Moreover, the accuracy of calibration is directly related to the effect of data fusion. Therefore the calibration between sensors is very important and necessary. In this paper, we propose a novel method to find accurate rigid-body transformation for the extrinsic calibration of a 3D LiDAR and a stereo cameras, using two Aruco calibration boards. Also the validity of the algorithm is demonstrated by some experiments. © 2019 IEEE.

关键词： Calibration

来源：评论

学校读者我要写书评

暂无评论

Model learning for two-wheeled robot self-balance control

Model learning for two-wheeled robot self-balance control

引用

2019 IEEE International Conference on robotics and Biomimetics, ROBIO 2019

作者： Li, Enbo Feng, Haibo Zhou, Haitao Li, Xu Zhai, Yanwu Zhang, Songyuan Fu, Yili Harbin Institute of Technology State Key Laboratory of Robotics and System Harbin Heilongjiang Province150001 China

ISBN: (纸本)9781728163215

Two-wheeled robots have many advantages over other mobile robots, but they are difficult to self-balance compared with other wheeled robots. Reinforcement learning (RL) is a general framework for sequential decision-making problems. So far, there are many applications of reinforcement learning to solve robot control problems, but most of them are used in simulators because of the large amount of data required. In addition, due to the reality gap, the policy learned in a simulation environment cannot be transferred directly to a real robot. Real robot data often expensive due to the potential damage to the robot. Model-based methods require far fewer robot data than model-free methods, but these methods have the problem of model bias. In this paper, we use a model-based reinforcement learning method to achieve self-balance of a two-wheeled robot. We present a model learning method that can reduce the problem of model bias. Our method combines the simulator and a few real robot data to learn a probabilistic dynamics model of the robot through an iterative way, which requires no expertise and can learn from scratch. Then the control policy is optimized based on the learned model. © 2019 IEEE.

关键词： Reinforcement learning

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：