Abstract
This paper discusses the retinal model and 3D surface detection model to grasp unknown objects for partner robot using 3D range camera. A partner robot for coexisting with human should perceive specific information and take stable action in general environment. In our research, we use the partner robot for table clearing that perceives a fallen down paper cup or a paper ball by 3D range camera and takes table clearing action by robot arm. A 3D range camera measures a huge volumes of distance data. On the other hand, it is difficult to recognize relevant information from a huge volumes of distance data including noisy information. Therefore, we propose the retinal model that pays attention only relevant information area based on human visual perception. Moreover, we propose the 3D surface detection model to recognize a position and posture of unknown objects. Through experiments, we discuss effectivity of our proposal.
