A sensory gripper using tactile sensors for object recognition, orientation control and stable manipulation

Abstract

A sensory gripper is presented which incorporates two tactile sensing matrices. The tactile sensors acquire 3D information about object contours of interest. An algorithm is presented that is used to determine the orientation of various objects lying randomly, regardless of their identity. The robot gripper is then aligned appropriately, and stable hold sites are found. Object identification is carried out if required for subsequent actions that may be desired. The sensory gripper is therefore capable of the manipulation of objects without a need for prior knowledge of their identity and from confined surroundings. An absolute moment invariant of a 3D volumetric image of an object is defined that enables its discrimination in an object-dominated environment. A database or a lookup table containing all objects within an environment is then used to obtain physical information relating to an object of interest after the process of discrimination is completed. The stability of the holding arrangement as well as the optimal grip force is determined by 3D rigid body dynamics method.