作者: Shane Griffith , Vladimir Sukhoy , Alexander Stoytchev
DOI: 10.1109/HUMANOIDS.2011.6100910
关键词:
摘要: This paper describes a new graph-based representation that captures the interaction possibilities between robot's hand and one or more objects in environment terms of dependencies their movements lack movements. The nodes graph correspond to tracked visual features, i.e., objects. edges pairwise movement features. As robot performs different behaviors with structure changes, are added deleted over time. tests hypothesis sequences such graphs can be used as signature essence some object categories. framework was tested container non-container tried insert small block into them. results show able distinguish these two categories based on corresponding dependency graphs.