Grasping goals for visual coordination

The authors demonstrate that for problems in robot hand/eye coordination, topological analysis of the properties of the perceptual kinematic map (PKM), a mapping from the space of manipulator configurations to the camera space, facilitates a conceptually simple interpretation for grasping tasks. Within this framework, stable grasps correspond to regions on the hypersurface defined by the PKM. It is shown that existing work on the determination and analysis of stable grasps allows the determination of these regions. The computation of these regions is described in cases in which the visual input to the hand/eye system is in the form of a depth map supplied by a visual reconstruction module, or in the event that the target has been identified by an object recognition module.<<ETX>>