Deformable objects are ubiquitous in our daily life. The ability to manipulate them is an important measure of the robot's intelligence and dexterity. This project investigates several fundamental issues related to robot grasping of deformable objects (which has remained an underdeveloped research area): (i) efficient and interactive modeling of grasp formation; (ii) analysis and synthesis of two-finger squeeze grasps with understanding of the role of elasticity; (iii) resistance to and absorption of external forces; and (iv) sensing and force control for grasp achievement. A graphical interface called GraspDeform is under development not only for simulation purpose, but also to assist the design, evaluation, and implementation of grasping strategies over a robot platform. The PI hopes to acquire in-depth understanding about the geometry and mechanics of grasping in the presence of deformation, by examining issues like contact, strain energy, elasticity constants, and friction. The project intends to demonstrate that a robot hand can reliably grasp various deformable objects. Results are expected to considerably widen up research on manipulation of deformable objects, influence deformable modeling in computer graphics, and have potential applications in medical and home robotics. They will be disseminated to the research community via the development of an interactive website on deformable grasping.
Deformable objects are ubiquitous in our daily life. The skill to manipulate them constitutes an important measure of the robot's intelligence and dexterity. Unfortunately, such skill is not possessed by most of today's robots despite their ever-increasing capabilities, due to lagging research efforts on the subject. The objective of this project was to program a robot hand to grasp soft objects at ease just like the human hand. We have designed a method that controls two robotic fingers to achieve a grip of a deformable object simply by squeezing it, taking full acount of factors including friction and evolving contacts. The method is backed up via applying the so-called ``finite element analysis'', which is based on the elasticity theory from the mechanics of materials. We have also derived several ``optimal'' ways of squeezing that bear close resemblance to those used by the human hand, as well as a way of using ``minimum'' efforts to resist external disturbances once the object is grasped. Equipped with our method, a robotic hand from Barrett Technologies, Inc. has successfully grasped a range of soft solid and hollow objects of different shapes and materials. Our results have potential impacts on medical robotics, where stable handling of soft organs has ultimate importance, and on the development of home robots to relieve us from housekeeping chores. This research has also yielded computational solutions to several problems in related scientific disciplines.
