Estimation of easily visible position for unknown object grasping

This paper describes the easily visible position for deciding action of unknown object grasping without all of predefined knowledge in a real situation. We have proposed a perception system which is composed of an online processable object detection method based on plane detection, and an invariant perceptual information for a grasping action. The object detection method can detect the point data on the object without the predefined knowledge. The invariant detection method for a grasping action is explained by inertia tensor and fuzzy inference. As an experiment, we compare the object extraction result and the invariant perceptual information value through experimentation of the various sensing position and the object pose. We discuss the relationship between the sensing position and the object detection result. Furthermore, we discuss the invariant perceptual information value which can be estimated the easily visible position of the object.