Simple kinesthetic haptics for object recognition

Object recognition is an essential capability when performing various tasks. Humans naturally use either or both visual and tactile perception to extract object class properties. Typical approaches for robots, however, require complex systems multiple high-density sensors which can be highly expensive. In addition, they usually actual collection of a large dataset from real objects through direct interaction. this paper, we propose kinesthetic-based method that performed with any multi-fingered robotic hand in the kinematics known. The does not based on observing grasps objects. We utilize unique frame invariant parameterization learn instances shapes. To train classifier, training data generated rapidly solely computational process without interaction then compare between two iterative algorithms integrate trained classifier. classifiers are independent particular robot and, therefore, exerted ones. show experiments, few grasps, acquire accurate classification. Furthermore, approach scalable sizes. Similarly, global classifier identify general geometries (e.g., ellipsoid box) rather than ones demonstrated set Full scale experiments analysis provided performance method.