A control paradigm for general purpose manipulation systems

Mechanical end effectors capable of dextrous manipulation are now a reality. Solutions to the high level control issues, however, have so far proved difficult to formulate. We propose a methodology for control which produces the functionality required for a general purpose manipulation system. It is clear that the state of a hand/object system is a complex interaction between the geometry of the object, the character of the contact interaction, and the conditioning of the manipulator. The objective of this work is the creation of a framework within which constraints involving the manipulator, the object, and the hand/object interaction can be exploited to direct a goal oriented manipulation strategy. The set of contacts that are applied to a task can be partitioned into subsets with independent objectives. The individual contacts may then be driven over the interaction surface to improve the state of the grasp while the configuration of the hand addresses the application of required forces. A system of this sort is flexible enough to manage large numbers of contacts and to address manipulation tasks which require the removal and replacement of fingers in the grasp. A simulator has been constructed and results of its application to position synthesis for initial grasps is presented. A discussion of the manipulation testbed under construction at the University of Utah employing the Utah/MIT Dextrous hand is presented.