Grasp quality measures

Abstract

The correct grasp of objects is a key aspect for the right fulfillment of a given task. In robotics, the development of grippers more and more complex and versatile, such as mechanical hands, augments the necessity for algorithms to automatically determine grasps, and in a parallel way, arises the need to quantify the quality of the generated grasps in order to optimize them. This work reviews the quality measures proposed in the grasp literature to quantify the grasp quality. We classify the quality measures into two groups according to the main aspect evaluated by the measure: the location of the contact points on the object or the hand configuration. We also review the approaches that combine different quality measures from the two previous groups to obtain a global quality measure.

Grasp synthesis for 3D objects has been tackled with two different approaches: empirical or analytical. The empirical approach (knowledge-based) imitates human grasp using heuristics to choose a grasp shape from a set of basic hand postures. On the other hand, analytic approaches choose the finger positions and the hand configuration with kinematical and dynamical formulations, in general optimizing an objective function such as the grasp stability or the resistance to external perturbations. This technical report reviews the main analytical algorithms for grasp synthesis in polyhedral or arbitrarily shaped 3D objects. The relation between the fixture design for manufacturing tasks and the grasp synthesis is also discussed.