基于改进Soft-DTW的人类示教轨迹模板生成方法  

Trajectory Template Generation Method for Human Demonstration Based on Improved Soft-DTW

作　　者：薛俊楠 李志海[1,2] 于洪鹏 XUE Junnan;LI Zhihai;YU Hongpeng(Shenyang Institute of Automation,Chinese Academy of Sciences,Shenyang 110016,China;Institutes for Robotics and Intelligent Manufacturing,Chinese Academy of Sciences,Shenyang 110169,China;University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院沈阳自动化研究所,沈阳110016 [2]中国科学院机器人与智能制造创新研究院,沈阳110169 [3]中国科学院大学,北京100049

出　　处：《小型微型计算机系统》2025年第3期528-534,共7页Journal of Chinese Computer Systems

基　　金：广东省重点领域研发计划项目(2020B090925001)资助.

摘　　要：针对人类示教轨迹样本存在的时间和空间不对齐导致难以提取运动特征的问题,首先提出了基于典型时间规整(Canonical Time Warping,CTW)算法用于多条轨迹对齐的方法,并将其引入到软-动态时间规整(soft-dynamic time warping,soft-DTW)算法中以提取轨迹模板,其次在CTW算法中引入了一个新的变量,以提升CTW算法在对齐多条轨迹方面的能力;最后,在实验中利用多种轨迹验证了所提出的轨迹模板提取方法,实验结果表明所提出的方法可以从人类示教轨迹中快速地提取共有的运动特征,并且对示教轨迹在时间和空间上的差异具有较好的鲁棒性.To solve the problem that the time and space misalignment of human demonstration trajectories makes it difficult to extract motion features.Firstly,a method based on canonical time warping(CTW)algorithm for multiple trajectories alignments is proposed,and introduced into soft-dynamic time warping(soft-DTW)algorithm to extract trajectory templates.Secondly,a new variable is introduced into the CTW algorithm to improve the CTW algorithm′s ability to align multiple trajectories.Finally,using multiple trajectories verifies the proposed trajectory template extraction method in the experiment.The results show that the method can quickly extract common motion features from demonstration trajectories,and has good robustness to differences in time and space of demonstration trajectories.

关 键 词：示教学习 人类示教轨迹 轨迹模板 CTW soft-DTW 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]