Saturday, October 09, 2004

Robotic grasp theory

Robotic grasp theory is a magnificent marriage of science and engineering. The human hand is a magnificent piece of design -- flexible and adaptable beyond belief. It's no surprise that many robotic efforts have attempted to duplicate its capabilities.

Sadly, these efforts often go too far in either extreme. Some attempt to duplicate the human hand too closely. The result? Anthropomorphic robot hands that require hopelessly complex control algorithms, and which lack the fine sensing capabilities of a human hand. These devices are cumbersome, and often impractical to use. Others use primitive parallel jaw grippers, which are decidedly more restricted in their capabilities. Since most parts require some custom tooling in order to be held securely, this greatly limits the range of workpieces that a parallel jaw gripper can hold.

(Other variants exist, such as three-fingered end-effectors. The end results is similar, though -- limited grasp capabilities.)

Velasco et al. offered a compromise in which rapid prototyping technology was used to develop gripper tooling based on CAD models of the parts to be grasped. The CAD capabilities allowed the tooling to be developed rapidly, and with due consideration to part clearances and manufacturing variability. This gave parallel jaw grippers a great deal more versatility than could be achieved using rudimentary tooling, without requiring the complex mechanisms or ocntrol algorithms required of anthropomorphic robot hands.