This Small Business Innovation Research (SBIR) Phase I research project will create a generic robotic grasping system. Generic grasping is among the most difficult of robotic problems. The universe of objects is unlimited, and it is very challenging to create an algorithm that enables grasping of all of them. Many techniques exist, but they typically do not run in real time (on common computational platforms), and they are incomplete. There are no generic algorithms that give a fast (at runtime) method to grasping a very large number of objects with any type of gripping mechanism. This research will enable a new technique using an XML-configured object-oriented database of fast algorithms. Though it is not possible to create a database that includes every type of object, it is possible to include enough objects to make the database effectively comprehensive. To populate the database, this research proposes refinement algorithms that can modify and optimize nominal grasps for specific new objects. Algorithms for nominal grasps will be generated using an efficient human-supervised approach.
Generic robotic grasping, when solved, will change forever many human endeavors. Robotic grasping insufficiency has been identified as one of the two primary obstacles to wide robot adoption (the other is machine-vision insufficiency). Great potential exists both for further enabling the existing robotic base and for extending robotics to new applications. Applications include manufacturing, space-based applications, defense, medical, entertainment, energy, and agriculture. All these will benefit from improved grasping. Today, there are no persistent robotic manipulators in the home. No robot can open doors in a general household setting. Yet the need for home robotics, with an aging population in the developed world, is great. Low-cost grasping will enable solution of practical problems and be an enabling technology for home robotics.
