Files in this item

FilesDescriptionFormat

application/pdf

application/pdfMATTHEWS-DISSERTATION-2015.pdf (24MB)
(no description provided)PDF

Description

Title:Experiments in quasi-static manipulation of an elastic rod
Author(s):Matthews, Dennis
Director of Research:Bretl, Timothy W.
Doctoral Committee Chair(s):Bretl, Timothy W.
Doctoral Committee Member(s):Alleyne, Andrew G.; Hutchinson, Seth A.; Schutt-Aine, Jose E.
Department / Program:Electrical & Computer Eng
Discipline:Electrical & Computer Engr
Degree Granting Institution:University of Illinois at Urbana-Champaign
Degree:Ph.D.
Genre:Dissertation
Subject(s):robotics
elastic rod manipulation
motion planning
cost function
Abstract:The purpose of this dissertation is to experimentally validate a new approach to robotic manipulation of deformable objects. As a case study, it will focus on the manipulation of objects that can be modeled as Kirchhoff elastic rods, for example a metal wire that is held at each end by robotic grippers. Any curve traced by this wire when in static equilibrium can be described as the solution to an optimal control problem with boundary conditions that vary with the position and orientation of each gripper. Recent work has shown that the set of all local solutions to this problem over all possible boundary conditions is a smooth manifold of finite dimension that can be parameterized by a single chart, the coordinates for which have a direct interpretation as forces and torques. These coordinates-in principle-allow the problem of manipulation planning to be formulated as finding a path of the wire through its set of equilibrium configurations, something that was previously thought impossible and that has significant advantages. However, this approach has never before been applied to hardware experiments. We begin by considering a metal wire that is confined to a planar workspace. We derive global coordinates for this wire and characterize the extent to which they accurately describe its shape during robotic manipulation. In particular, we show that differences between predicted and observed manipulation (which can be quite large) derive primarily from small errors in the position and orientation of each robotic gripper. We reduce these differences in two ways. First, we give an algorithm for manipulation planning that locally minimizes sensitivity to errors in gripper placement. Second, we give a feedback control policy (based on force sensor data as well as on position and orientation estimates) that locally minimizes the sum-squared error between planned and observed paths in our global coordinate chart for the wire. We conclude by showing-again, with hardware experiments-that these results extend directly to enable robotic manipulation of a metal wire in a three-dimensional workspace.
Issue Date:2015-07-09
Type:Thesis
URI:http://hdl.handle.net/2142/87998
Rights Information:Copyright 2015 Dennis Matthews
Date Available in IDEALS:2015-09-29
Date Deposited:August 201


This item appears in the following Collection(s)

Item Statistics