Evidence-directed, topological reasoning about robot regrasping tasks

An intelligent robot prehension paradigm is introduced, focusing attention on evidential reasoning that would be used in combination with topological reasoning for regrasping tasks. The prehension paradigm is based on a topological model of multifingered hands capable of tip prehension, palm prehension or a combination of tip and palm prehension. The topological model constitutes a much richer set of available grasps than those provided by current grasp models. Associated with this topological model is a set of reasoning algorithms for grasp selection and for regrasping. The grasp selection scheme, using mainly topological reasoning, is characterized by a set of continuous transformations from a given, symbolic task description into a set of numeric parameters specifying the hand posture and functionality of a selected grasp. The regrasping scheme using mainly evidential reasoning to guide a change of grasps according to perceptual data. From a control point of view, both reasoning schemes will operate concurrently for grasp synthesis and regrasping