道具機能表現を分類するための道具身体化モデルの転移学習

抄録

Our purpose is to observe the superiority of transfer learning method by using large numbers of datasets and comparisons of variance ratio. We researched tool-body assimilation model using neuro-dynamical system for robot tool-use where robot recognizes the tools as part of its own body. This model uses transfer learning, where first the robot learns its own body model before learning tool-use experiences. Comparisons between simultaneous and transfer learning is performed. By performing object manipulation task, we statistically confirmed the clustering of tool function.