Grasp Selection for In-Hand Robotic Manipulation of Non-Rigid Objects with Shape Control

Abstract

The ability to intelligently manipulate non-rigid objects with a robotic hand is a requirement for many tasks which remain to be automated in domains such as agriculture, food processing, or medicine. This ability includes the nontrivial skill of controlling the shape of the object during the manipulation, which normally requires advanced knowledge of the properties of the object under manipulation as well as simulation capabilities. This work proposes an efficient algorithm and control framework that integrates RGB-D computer vision with a three-finger robotic gripper in a systemic approach for selecting the initial grasp points for a 2D shape control task through visual inspection of the object and desired target shape contours.