Grasp Stability from Vision and Touch

We study the exploitation of tactile and visual feedback to predict the stability of a grasp before attempting to lift an object and thus to prevent failures. Our robot learns an empirical representation of stable and unstable grasps by exploring given grasping configurations.