Simulation software for the Utah/MIT dextrous hand

Abstract

The Utah/MIT dextrous hand is a 16-joint, four-finger manipulator that is being set-up in our laboratory for concept demonstrations of light assembly and repair tasks. In this paper, we describe a variety of simulation methods that graphically portray the grasps and tasks that the dextrous hand will subsequently carry out in practice. The first series of simulations have followed on from a taxonomy of human hand grasps. This human hand taxonomy shows how grasps can be divided into two basic types, namely power and precision. Since the Utah/MIT dextrous hand is architecturally different from the human hand, the first task has been to generate, in a computer graphic format, an equivalent taxonomy for the Utah/MIT hand. This has involved the development of a “modeler” that not only gives a graphic representation of the complex structure but also provides for realistic rendering, contact information and (with simpler rendering) real time animations of dextrous manipulations. Thus, it is emphasized that the modeler can show various simulations: the static taxonomy for the Utah/MIT hand; the dynamic joint movements; and interactions with objects of varying size. The spatial relationships between the hand and any arbitrary object can be simulated in order to check, for example, that the hand can geometrically bound and close upon the object. Programming tools for the dextrous hand are also discussed in the paper, in particular a sensor-based dataglove. When a human wears the instrumented glove, gesticulations can be stored for subsequent programming of the Utah/MIT hand. Since the architectures of human hand/glove and the Utah/MIT hand are different, this involves cross calibration and conversions between the two media.