Degree Type

Thesis

Date of Award

2012

Degree Name

Master of Science

Department

Computer Science

First Advisor

Yan-Bin Jia

Abstract

This thesis introduces a strategy of grasping deformable objects using two fingers which specifies finger displacements rather than grasping forces. Grasping deformable objects must maintain its equilibrium before and after the induced deformation. The deformed shape and grasping force are computed using the finite element method (FEM). The equilibrium of the object is guaranteed automatically since the computed grasping force are collinear and sum up to 0. To achieve a grasp, the forces have to be tested for staying inside the pre- and post- deformation contact friction cones. This test could be as expensive as solving a large linear system, if the deformed shape is computed. We present an algorithm that performs a grasp test in O(n) time, where n is the number of discretization vertices under FEM, after obtaining the spectral decomposition of the object's stiffness matrix in O(n3) time. All grasps (up to discretization) can be found in additional O(n2) time. Robot grasping experiments have been conducted on thin 2.5D objects.

Copyright Owner

Feng Guo

Language

en

File Format

application/pdf

File Size

39 pages

Share

COinS