Virtual Articulation and Kinematic Abstraction in Robotics

292 Pages·2014·28.24 MB·English

Virtual Articulation and Kinematic Abstraction in
Robotics
by
Marsette Arthur Vona, III
B.A., Dartmouth College (1999)
S.M., Massachusetts Institute of Technology (2001)
Submitted to the Department of Electrical Engineering and Computer
Science
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in Electrical Engineering and Computer Science
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
September 2009
c Massachusetts Institute of Technology 2009. All rights reserved.
�
Author ..............................................................
Department of Electrical Engineering and Computer Science
September 1, 2009
Certified by..........................................................
Daniela Rus
Professor
Thesis Supervisor
Accepted by.........................................................
Terry P. Orlando
Chairman, Department Committee on Graduate Theses
2
Virtual Articulation and Kinematic Abstraction in Robotics
Marsette Arthur Vona, III
Submitted to the Department of Electrical Engineering and Computer Science
on September 1, 2009, in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in Electrical Engineering and Computer Science
Abstract
This thesis presents the theory, implementation, novel applications, and experimental
validation of a general-purpose framework for applying virtual modifications to an
articulated robot, or virtual articulations. These can homogenize various aspects of a
robot and its task environment into a single unified model which is both qualitatively
high-level and quantitatively functional.
This is the first framework designed specifically for the mixed real/virtual case.
It supports arbitrary topology spatial kinematics, a broad catalog of joints, on-line
structure changes, interactive kinostatic simulation, and novel kinematic abstractions,
where complex subsystems are simplified with virtual replacements in both space and
time. Decompositionalgorithms,includinganovelmethodofhierarchicalsubdivision,
enable scaling to large closed-chain mechanisms with 100s of joints.
Novel applications are presented in two areas of current interest: operating high-
DoF kinematic manipulation and inspection tasks, and analyzing reliable kinostatic
locomotion strategies based on compliance and proprioception. In both areas vir-
tual articulations homogeneously model the robot and its task environment, and ab-
stractions structure complex models. For high-DoF operations the operator attaches
virtual joints as a novel interface metaphor to define task motion and to constrain
coordinated motion (by virtually closing kinematic chains); virtual links can repre-
sent task frames or serve as intermediate connections for virtual joints. For compliant
locomotion, virtual articulations model relevant compliances and uncertainties, and
temporal abstractions model contact state evolution.
Results are presented for experiments with two separate robotic systems in each
area. For high-DoF operations, NASA/JPL’s 36 DoF ATHLETE performs previously
challenging coordinated manipulation/inspection moves, and a novel large-scale (100s
of joints) simulated modular robot is conveniently operated using spatial abstrac-
tions. For compliant locomotion, two experiments are analyzed that each achieve
high reliability in uncertain tasks using only compliance and proprioception: a novel
vertical structure climbing robot that is 99.8% reliable in over 1000 motions, and a
mini-humanoid that steps up an uncertain height with 90% reliability in 80 trials. In
bothcasesvirtualarticulationmodelscapturetheessenceofcompliant/proprioceptive
strategies at a higher level than basic physics, and enable quantitative analyses of the
limits of tolerable uncertainty that compare well to experiment.
Thesis Supervisor: Daniela Rus
Title: Professor
3
4
Acknowledgments
Theideaspresentedinthisthesistietogetherseverallong-standingthreadsinmyown
personal intellectual development. The oldest of these, an interest in the geometric
mechanisms of robots for their practical and aesthetic values, reaches back to my
childhood. Then as now, my parents have enthusiastically and unfailingly supported
each and every one of my projects, and deserve the first thanks. My mother often
tells the story of the first “machine” I ever built that “really worked”: an indoor
apparatus for melting snow. My father unexpectedly passed away just a few months
before the final submission of this dissertation. I still remember those early science
projects you helped me with, Dad: making a battery from a lemon and strips of
metal from the hardware store; stringing a working telegraph—complete with actual
Morse keys—across my first-grade classroom; the white-hot “light bulb” filament that
worked so well only with your special dental ligature wire. Thanks Dad. You will
always be the first “Dr. Vona.”
In terms of long-term support, my adviser Daniela Rus comes in an unusually
closesecond-placetomyparents. Ihavehadtheuncommonopportunitytoworkwith
Daniela both for my undergraduate thesis, which I completed at Dartmouth college
in 1999, as well as for this dissertation at MIT. At Dartmouth Daniela introduced
me to the field of self-reconfiguring robots, a still-productive line of research that
includes the large-scale modular tower example in Section 5.4. Since then, Daniela
has been instrumental in most of the major transitions in my career. Rather than
rigidly telling me what to do, she has always enabled and encouraged me to follow
the ideas that are the most productive and interesting.
This thesis entailed significant theoretical development, practical physical imple-
mentation of several robotic systems, and experiments using both those systems and
also the ATHLETE robot at NASA’s Jet Propulsion Laboratory. Many people have
helped in each of these three areas.
I thank the other members of my Ph.D. committee, Rodney Brooks and Erik
Demaine, for inspiration and guidance in the conceptual and theoretical directions of
5
the work. In this realm I also am grateful for key insights gained in discussions with
KeithKotay, withwhomIworkedbothatDartmouthandatMIT;IuliuVasilescuand
Eduardo Torres-Jara, also recent graduates of MIT; Russ Tedrake, associate professor
of EECS at MIT; and other members of past and current research groups including
in particular Nikolaus Correll, Mac Schwager, Peter Osagie (who designed and im-
plemented the path planner for the Shady climbing robot presented in Section J.3.2),
and Robert Fitch. Special thanks also go to Sanjay Sarma, professor of Mechanical
Engineering at MIT, with whom I studied during my first year back at MIT in 2003-4
(after working at JPL for two years between my M.S. and Ph.D.); and to Ulrich Ko-
rtenkamp, now a professor at the University of Education Karlsruhe, who introduced
me to the important property of continuity in interactive geometry software.
As a (primarily) computer science student, I have had much to learn about the
electromechanical design and fabrication that was needed for the interface device and
the climbing robot hardware I developed for this thesis (Appendices I and J). Keith
and Iuliu have helped me innumerable times here as well, as have other students in
our group including Carrick Detweiler (who designed the motor control boards used
on the climbing robot) and Marek Doniec. I learned about mechanical systems, sens-
ing, control, and actuation during my M.S. work with David Trumper, a professor
of Mechanical Engineering at MIT, and from my co-students at that time, includ-
ing Marten Byl, Katie Byl, and Michael Liebman. I also thank Mark Yim, now
professor of Mechanical Engineering at the University of Pennsylvania, and Satoshi
Murata, a professor at Tokyo Institute of Technology, both founders in the area of
self-reconfiguring robot hardware. Over the years I have picked up valuable knowl-
edge from each of them, both in person and through their writings. For example, the
singular-linkage grip mechanism in the climbing robot was motivated after a partic-
ular discussion with Professor Murata.
Some of the key experimental work in this thesis (Section 5.3) was carried out
using the excellent ATHLETE hardware at JPL, which was developed there by Brian
Wilcox, Todd Litwin, Je↵ Biesiadecki, Jaret Matthews, Matt Heverly, and others.
Je↵ Norris, a long-time collaborator, and now the supervisor of the Planning Soft-
6
ware Systems group at JPL, was instrumental both in developing and supporting
the concept and details of these experiments, as well as in arranging the logistics.
David Mittman was the main on-the-ground collaborator during my visit to JPL,
and deserves a major portion of the credit for the success of our experiments with
ATHLETE. The graphical simulation model of the robot pictured in this thesis is
provided courtesy of the RSVP Team at NASA/JPL.
Early experiments with the stair-stepping mini-humanoid in Section 6.7 were de-
veloped by Albert Wang and Jeremy Lai, then both undergraduates at MIT. They
developed the CAD model of the robot pictured in this thesis, based originally on a
public-domain model from Pedro Teodoro at the Technical University of Lisbon.
Melissa King has been an inspiring companion for almost exactly as many years
as I have been at work on this thesis. It’s hard to imagine how I could have done
it without her support and understanding. Other friends that have stuck with me
throughthislongprocessincludeBenGuaraldi, theHomerextendedfamily, JoeEdel-
man, and Brent Knopf. Cleopatra King, though she doesn’t talk and has no thumbs,
has suggested to me some of the amazing potential for low-impedance locomotion
mechanisms.
Work with ATHLETE was funded in part by a grant from the JPL/Caltech Di-
rectors Research and Development Fund Strategic University Research Partnership
program. OtheraspectsofthisthesiswerefundedbytheNationalScienceFoundation
under the Emerging Frontiers in Research and Innovation program.
7
8
Contents
1 Introduction 23
1.1 Virtual Articulations . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.2 Kinematic Abstractions . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.3 Organization of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . 27
1.4 System Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.5 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.5.1 Contributions of the Framework . . . . . . . . . . . . . . . . . 31
1.5.2 Contributions of the Operations Interface Applications . . . . 33
1.5.3 Contributions of the Compliant Motion Applications . . . . . 33
2 Related Work 35
2.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.2 Virtual Articulations in Specific Applications . . . . . . . . . . . . . . 36
2.3 Prior Work in Kinematic Abstraction . . . . . . . . . . . . . . . . . . 38
2.4 Prior Work in Kinematic Modeling . . . . . . . . . . . . . . . . . . . 39
2.5 Other Similar Systems . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.5.1 Physical Dynamics Simulation . . . . . . . . . . . . . . . . . . 40
2.5.2 Geometric Constraint Solving . . . . . . . . . . . . . . . . . . 41
2.6 Additional Categories of Related Work . . . . . . . . . . . . . . . . . 42
3 Model Structure and Topological Interaction Algorithms 43
3.1 Challenges of Mixed Real/Virtual Modeling . . . . . . . . . . . . . . 45
3.2 Linkages and The Kinematic Graph . . . . . . . . . . . . . . . . . . . 48
9
3.2.1 Spanning Tree Topology . . . . . . . . . . . . . . . . . . . . . 49
3.2.2 Names and Paths . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.2.3 Root Joints . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.2.4 Formal Linkage Definition . . . . . . . . . . . . . . . . . . . . 53
3.3 Joint Mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.3.1 Components of the Joint Space Representation . . . . . . . . . 55
3.3.2 Formal Mobility Definition . . . . . . . . . . . . . . . . . . . . 58
3.3.3 (t,✓) Parametrization of SE(3) . . . . . . . . . . . . . . . . . 59
3.3.4 The Catalog of Joint Types . . . . . . . . . . . . . . . . . . . 62
3.3.5 Joint Motion Limits . . . . . . . . . . . . . . . . . . . . . . . 70
3.4 Joint Space, Configuration Space, and DoF . . . . . . . . . . . . . . . 73
3.4.1 Properties of the Mobility of a Single Joint . . . . . . . . . . . 73
3.4.2 Properties of Whole-Linkage Mobility . . . . . . . . . . . . . . 74
3.4.3 Local vs. Global Properties of . . . . . . . . . . . . . . . . . 76
C
3.5 Hierarchical Linkages . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.5.1 Sub-linkage Disposition . . . . . . . . . . . . . . . . . . . . . . 80
3.5.2 Crossing Joints . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.5.3 Structure Abstraction . . . . . . . . . . . . . . . . . . . . . . 82
3.6 Interacting with Model Structure . . . . . . . . . . . . . . . . . . . . 84
3.6.1 Notation and Assumptions . . . . . . . . . . . . . . . . . . . . 84
3.6.2 Changing Joint Disposition and Reconnecting Joints . . . . . 87
3.6.3 Inverting Joints and Re-grounding . . . . . . . . . . . . . . . . 89
3.6.4 Operations on Sub-Linkages . . . . . . . . . . . . . . . . . . . 90
3.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4 Linkage Motion and Topological Decomposition Algorithms 95
4.1 Local Assembly by Linear Optimization . . . . . . . . . . . . . . . . . 98
4.1.1 Iterative Damped Least Squares . . . . . . . . . . . . . . . . . 101
4.2 Adding Di↵erential Control . . . . . . . . . . . . . . . . . . . . . . . 102
4.2.1 Residual Compensation and Nullspace Projection . . . . . . . 105
10