Journal Article Motion Segmentation and Recognition for Imitation Learning and Influence of Bias for Learning Walking Motion of Humanoid Robot Based on Human Demonstrated Motion

Takahashi, Yasutake  ,  Hatano, Hiroki  ,  Maida, Yosuke  ,  Usui, Kazuyuki  ,  Maeda, Yoichiro

Two main issues arise in practical imitation learningby humanoid robots observing human behavior –the first is segmenting and recognizing motion demonstratednaturally by a human beings and the second isutilizing the demonstrated motion for imitation learning.Specifically, the first involves motion segmentationand recognition based on the humanoid robotmotion repertoire for imitation learning and the secondintroduces learning bias based on demonstratedmotion in the humanoid robot’s imitation learning towalk. We show the validity of our motion segmentationand recognition in a practical way and report theresults of our investigation in the influence of learningbias in humanoid robot simulations.

Number of accesses :  

Other information