Abstract :
The purpose of this paper is to analyze, in some depth, the processes involved in the manipulation of solid objects using robotic hands. Such an analysis has implications for design, control, and motion planning. We make a case for the use of finger surfaces which have a distributed compliance and show how such surfaces may be realized by fluid filled pods. The arguments given are based on an analysis of the complexity of the kinematic programming problem, an enumeration of the number of independent feedback control channels necessary for achieving a firm grasp and an examination of frictional forces. We emphasize the necessity of dealing with models which lead to well-posed problems throughout the grasping process and point out some of the ways that earlier models lead to ambiguous situations.