Grasping at Part Edges

Abstract:

We propose a new model for grasping parts with an industrial parallel-jaw gripper where two define candidate grasps that are resistant to slipping and torque about the part’s center of mass.  These grasps are optimized to require minimum friction and torque at the contacts and must be accessible and robust to perturbations in part position.  An O(n³) algorithm is given for computing and ranking O(n²) such grasps on an n-sided polygonal slice through the part.  The algorithm will be part of a design and simulation system that can rapidly provide feedback to designers; thus it must run quickly and reliably.  We have also implemented the algorithm in a Java applet with a graphical user interface that allows Internet users to define a part; the applet computes, ranks, and displays the set of computed grasps.  To try the applet, please visit:

Introduction:

Parallel-jaw grasping is an important aspect of automated assembly.  The aim of this project is to develop an efficient algorithm that directly computes grasps for holding a polyhedral part with a parallel-jaw gripper.  In particular we consider the common case of a gripper attached to a SCARA-type robot arm with 4 degrees of freedom.  In contrast previous grasp synthesis methods our algorithm ensures that grasp points are physically accessible to the gripper, considers the location of the center of mass (com), and determines grasps that are robust to small perturbations in position.  The grasps are also optimized by minimizing dependence on friction or necessary torque, or when possible both.

A grasp is defined by two points on the boundary of the part, and the line connecting those points is referred to as the grasp axis.  SCARA robot kinematics requires that the grasp axis lies in a horizontal plane.  This reduces to grasping a 2D, polygonal part formed by the intersection of the 3D polyhedral part with a horizontal plane, the grasp plane, as shown in the following figure.

The algorithm considers all pairs of the polygon’s edges and finds the optimum, accessible grasp, if one exists, for each pair.  The algorithm runs in O(n³) time for an n-sided polygonal part.  Implementation in a Java applet available on the WWW allows users to design parts and provides a graphical user interface to adjust parameters such as center of mass, friction coefficient, vertex uncertainty, and grasp ranking criteria.  The algorithm has also been implemented in an industrial simulation package.

Publications:

• Gordon Smith, Eric Lee, Ken Goldberg, Karl Böhringer, and John Craig. "Computing Parallel-Jaw Grips", in IEEE International Conference on Robotics and Automation, Detroit, MI, 1999.[PDF]

Acknowledge:

This work was supported by the National Science Foundation under Presidential Faculty Fellow Award IRI-9553197, and Postdoctoral Research Associate Award CDA-9705022. Research fund was also provided by Adept Technology, Inc..