[1]王建强,黄开启,苏建华.基于前馈径向基网络的动态抓取参数估计方法[J].深圳大学学报理工版,2022,39(3):334-342.[doi:10.3724/SP.J.1249.2022.03334]
 WANG Jianqiang,HUANG Kaiqi,and SU Jianhua.Dynamic grabbing parameter estimation based on feedforward radial basis function network[J].Journal of Shenzhen University Science and Engineering,2022,39(3):334-342.[doi:10.3724/SP.J.1249.2022.03334]
点击复制

基于前馈径向基网络的动态抓取参数估计方法()
分享到:

《深圳大学学报理工版》[ISSN:1000-2618/CN:44-1401/N]

卷:
第39卷
期数:
2022年第3期
页码:
334-342
栏目:
电子与信息科学
出版日期:
2022-05-16

文章信息/Info

Title:
Dynamic grabbing parameter estimation based on feedforward radial basis function network
文章编号:
202203012
作者:
王建强黄开启苏建华
1)江西理工大学电气工程与自动化学院,江西赣州 341000;2)中国科学院自动化研究所复杂系统管理与控制国家重点实验室,北京 100190
Author(s):
WANG Jianqiang HUANG Kaiqi and SU Jianhua
1) School of Electrical Engineering and Automation, Jiangxi University of Science and Technology, Ganzhou 341000, Jiangxi Province, P. R. China 2) State Key Laboratory of Complex System Management and Control, Institute of automation, Chinese Academy of Sciences, Beijing 100190, P. R. China
关键词:
机器人技术运动估计核相关滤波卡尔曼滤波径向基函数非合作目标动态抓取
Keywords:
robot technology motion estimation kernel correlation filter Kalman filter radial basis function non-cooperative goals dynamic crawling
分类号:
TP242;TP181
DOI:
10.3724/SP.J.1249.2022.03334
文献标志码:
A
摘要:
针对机器人抓取非匀速、随机方向运动的机动物体的过程中手眼信息不同步问题,提出一种基于径向基函数神经网络预测的物体运动参数估计方法.该方法采用核相关滤波(kernel correlation filter, KCF)算法实时跟踪目标位置并建立目标的运动模型,再通过前馈感知机和径向基神经网络分别预测目标的运动参数,并动态调整卡尔曼滤波运动估计方程的采样时间,大幅降低了计算耗时,提高了机器人抓取目标运动中参数估计的准确性.在UR5机器人抓捕平面自由移动的物体上进行实验,结果表明,所提出方法与前馈感知机网络相比,既将手爪与目标在追捕过程中最大落后距离减少了一半,又将抓取所需的时间减少了20%.该方法能较好地解决机器人抓捕过程中目标速度或方向突变引发的动态自适应问题.
Abstract:
Aiming at the asynchronous problem of hand-eye information on the robot grab of the non-uniform and random direction moving objects, we propose an estimation method of object motion parameters based on radial basis function neural network prediction. The method uses the kernel correlation filter (KCF) algorithm to track the target position in real time and establish the target motion model. Then the motion parameters of the target are predicted by feedforward perceptron and radial basis function neural network, respectively. The sampling time of Kalman filter motion estimation equation is dynamically adjusted, which greatly reduces the calculation time and improves the accuracy of parameter estimation in the process of robot grasping in the target motion. Experiments are carried out on the UR5 robot catching objects that move freely on the plane. The experimental results show that compared with the feedforward perceptron network, the proposed method not only reduces the maximum lagging distance between the paw and the target in the pursuit process by half, but also reduces the time required for fetching by 20%.Our methoid can better solve the dynamic adaptive problem caused by the sudden change of target speed or direction in the process of robot capture.

相似文献/References:

[1]储颖,唐超伦,纪震,等.基于拉伸空间变换算法在DSA配准中的应用[J].深圳大学学报理工版,2005,22(2):127.
 CHU Ying,TANG Chaolun,JI Zhen,et al.Application of a spatial transformation method based on stretching to digital subtraction angiography registration[J].Journal of Shenzhen University Science and Engineering,2005,22(3):127.
[2]黄海明,吴林源,林俊豪,等.基于内嵌型光学弯曲传感器的软体手感知[J].深圳大学学报理工版,2019,36(3):237.[doi:10.3724/SP.J.1249.2019.03229]
 HUANG Haiming,WU Linyuan,LIN Junhao,et al.Soft hand perception based on embedded optical fiber bending sensor[J].Journal of Shenzhen University Science and Engineering,2019,36(3):237.[doi:10.3724/SP.J.1249.2019.03229]

更新日期/Last Update: 2022-05-30