ScholarWorks@동국대학교

초록

Robots currently being developed for ubiquitous applications contain diverse kinds of system components to smoothly control themselves and the mutual synchronization among one another to achieve their predefined goals. Given that motor primitives determine a robot's capabilities, research on how to define and generate such primitives is crucial to understanding the core techniques of robotic control. One method that generates motor primitives of greater variety than other methods is based on learning by demonstration. However, the generated motor primitives are similar when compared by a Euclidean-distance algorithm. This necessitates a mechanism that can accurately compare two motor primitives. In this paper, a NAO framework that generates motor primitives and compares them accurately is proposed. In an experiment in which the framework was applied to a NAO, 1487 candidate motor primitives were generated and 40 were utilized. © Springer-Verlag Berlin Heidelberg 2014.

키워드