Fuzzy Logic rule based systems have proven quite effective in solving problems that are nearly intractable analytically. We have applied fuzzy logic to grasping by developing rules which accept input from a vision system and use those rules to determine precisely how to manipulate objects.
The Barrett Hand is a 3-finger, 4 degree of freedom hand. It is equipped with clutches on each of the fingers that do not allow the hand to continue closing once an object is encountered. By adjusting the degree of spread of the fingers, the hand can accommodate a wide variety of object shapes and sizes. We use a vision system to determine the shape and size of an object. A fuzzy logic rule base then configures the hand to achieve a stable grasp. For example, if the vision algorithm identifies a rectangle 2 inches in height, the hand is configured with two fingers adjacent to each other, and the third approximately 180° from the two fingers. If it identifies a 5 inch diameter sphere, the three fingers spread approximately 120° from each other. Generally, these two grasp configurations manipulate objects quite well, however, this algorithm is more than adequate. Given a suite of generic shapes for the rule base, e.g., box or sphere, the system computes how the unknown object maps into the generic shapes. The combination of these data determines the finger spread that will effect a stable grasp. The approach is simple, but robust. Two objects are shown below. The first is a common shape that is part of the database (ellipsoid); the second an irregular shape that becomes a combination of the database shapes.
For additional information, please contact
Ron Lumia, Professor
Department of Mechanical Engineering
University of New Mexico
Albuquerque, NM 87131
505-272-7155
lumia@unm.edu