Parallel Algorithms for Synthesis of Whole-Hand Grasps

Abstract

This paper describes a parallel, dynamic programming algorithm for computing the space of high quality, whole-hand grasps of a target object that match a given grasp prototype, or example grasp. The grasp prototype is used to define a quality measure that allows the contacts of a grasp to be optimized independently. Link configurations that achieve high quality contacts are identified and chained together to form good hand configurations. This algorithm results in a six-dimensional projection of the global solution space. Grasps synthesized using this technique were tested using the Salisbury hand. An example grasp is shown.