Learning deep movement primitives using convolutional neural networks

Learning deep movement primitives using convolutional neural networks
复制标题

使用卷积神经网络学习深度运动原语

DOI:
--
复制
发表时间:
2017
期刊:
IEEE-RAS International Conference on Humanoid Robots
影响因子:
--
通讯作者:
Dongheui Lee
Dongheui Lee
中科院分区:
--
文献类型:
--
作者:
Affan Pervez;Yuecheng Mao;Dongheui Lee

文献摘要

参考文献

被引文献

相似文献

动态运动原语(dmp)被广泛用于运动数据的编码。任务参数化DMP (TP-DMP)可以使学习到的技能适应不同的情况。通常使用定制的视觉系统来提取任务特定的变量。这限制了此类系统在现实世界场景中的使用。本文提出了一种将DMP与卷积神经网络(CNN)相结合的方法。我们的方法保留了与DMP相关的泛化属性,而CNN从相机图像中学习任务特定的特征。这消除了提取任务参数的需要,通过在运动再现期间直接利用相机图像。通过一个真正的机器人执行的垃圾清理任务,演示了所开发方法的性能。通过数据增强,我们还证明了学习到的扫描技巧可以推广到任意对象。实验表明,我们的方法对几种不同的设置具有鲁棒性。
Dynamic Movement Primitives (DMPs) are widely used for encoding motion data. Task parameterized DMP (TP-DMP) can adapt a learned skill to different situations. Mostly a customized vision system is used to extract task specific variables. This limits the use of such systems to real world scenarios. This paper proposes a method for combining the DMP with a Convolutional Neural Network (CNN). Our approach preserves the generalization properties associated with a DMP, while the CNN learns the task specific features from the camera images. This eliminates the need to extract the task parameters, by directly utilizing the camera image during the motion reproduction. The performance of the developed approach is demonstrated through a trash cleaning task, executed with a real robot. We also show that by using the data augmentation, the learned sweeping skill can be generalized for arbitrary objects. The experiments show the robustness of our approach for several different settings.
DOI: 10.1109/whc.2017.7989877
发表时间: 2017-06
期刊: 2017 IEEE World Haptics Conference (WHC)
影响因子: --
作者:
Affan Pervez;Arslan Ali;J. Ryu;Dongheui Lee
通讯作者: Affan Pervez;Arslan Ali;J. Ryu;Dongheui Lee
末端执行器和零空间运动基元的增量动觉教学
DOI: 10.1109/icra.2015.7139694
发表时间: 2015
期刊: 2015 IEEE International Conference on Robotics and Automation (ICRA)
影响因子: --
作者:
Saveriano
通讯作者: Saveriano