http:^^www.cs.utexas.edu^users^qr^abstracts.html

This data set contains WWW-pages collected from computer science departments of various universities

The semantics of QPC's modeling language are grounded in the
mathematics of ordinary differential equations and their solutions.
This formalization enables the statement and proof of QPC's
correctness.  If the domain theory is adequate and the initial
description of the system is correct, then the actual behavior of the
system must be in the set of possible behaviors QPC predicts.  <p>

QPC has been successfully applied to problems in Botany and complex
examples drawn from Chemical Engineering, as well as numerous smaller
problems.  Experience has shown that the modeling language is
expressive enough to describe complex domains and that the inference
mechanism is powerful enough to predict the behavior of substantial
systems.  <p>


<hr>

<H2><a name="Franke">A Theory of Teleology</a></H2>


David Wayne Franke.  1993.
<a href="file://ftp.cs.utexas.edu/pub/qsim/papers/Franke-PhD-92.ps.Z">
<b>A Theory of Teleology</b></a>
Doctoral dissertation, Department of Computer Sciences,
The University of Texas at Austin. <p>

<H3>Abstract</H3>

 A representation language for teleological descriptions, or
descriptions of purpose, is defined.  The teleological language, TeD,
expresses the descriptions of purpose in terms of design modifications
that guarantee the satisfaction of design specifications.  These
specifications express potential behaviors the designed artifact
should or should not exhibit.  We define an abstraction relation on
behavior and implement model checking and classification algorithms
that computethis abstraction relation. The model checking algorithm
determines whether or not a behavior satisfies a specification.  The
classification algorithm provides effective indexing of behaviors and
teleological descriptions.  We implement an acquistion technique for
teleological descriptions and demonstrate how teleological
descriptions can subsequently be used in diagnosis, explanation,
case-based reasoning, design by analogy, and design reuse.  <p>

We demonstrate the behavior language, teleology language, acquisition
of teleological descriptions, and application of teleological
descriptions in explanation, diagnosis, and design reuse via examples
in the thermal, hydraulic, electrical, and mechanical domains.  We
define additional teleological operators that express purposes like
prevent, order, synchronize, maintain, and regulate, demonstrating the
ability to represent common human-generated descriptions of purpose in
TeD.  Expressing the purpose of preventing an undesirable behavior is
unique to TeD, and is an example of TeD's ability to express purposes
regarding missing behaviors and components removed from a design.  <p>

The teleology language developed in this work represents a significant
advance over previous work by providing a formal language that 1) is
independent of any particular domain of mechanisms or behavior
language, 2) can be effectively acquired during the design process,
and 3) provides an effective means of classifying and indexing
teleological descriptions.  <p>


<hr>

<H2><a name="Froom">High-Speed Navigation with Approximate Maps</a></H2>


Richard Allan Froom.  1995.  
<a href="file://ftp.cs.utexas.edu/pub/qsim/papers/Froom-PhD-95.ps.Z">
<B>High-Speed Navigation with Approximate Maps.</B></a>
Ph.D. Dissertation, The University of Texas at Austin.  <P> 


<H3>Abstract</H3>

A global map of a mobile robot's environment is essential for high-performance
navigation in large-scale space.  When portions of the environment are not
visible, a map is needed for route planning and enables high performance by
allowing the robot to anticipate regions that are occluded or beyond sensor
range.  Yet, autonomously acquired global map information is inevitably
uncertain due to the low positioning accuracy of mobile robots and the
possibility of changes to the environment. <P>

Previous work in high-speed navigation falls into two categories.  Global
optimization approaches assume that an accurate model of environment geometry
and robot dynamics are available, and address the problem of efficiently
approximating the minimum-time control between a start and goal state.  Reactive
navigation methods use only immediately sensed environment geometry to avoid
obstacles while moving to a specified goal position.  The global optimization
approach has the theoretical advantage of high performance, but it does not
address the significant uncertainty typical of mobile robots.  The reactive
navigation approach can respond to unanticipated geometry, but its overall
performance is limited. <P>

This dissertation describes a method for high-speed map-guided navigation in
realistic conditions of uncertainty.  A previously-developed method is used to
acquire a topologically correct, metrically approximate map of the environment
despite positioning errors.  Information in the approximate map guides the
operation of a novel, high-performance reactive navigator.  Performance does not
critically depend on the availability of expensive, accurate metrical
information.  Nonetheless, the map may be elaborated with more detailed
information, and, as its level of detail and accuracy is improved, performance
smoothly improves. <P>


<hr>

<H2><a name="Hartman">Automatic Control Understanding for Natural Programs</a></H2>


John Hartman.  1991.
<a href="file://ftp.cs.utexas.edu/pub/qsim/papers/Hartman-PhD-91.ps.Z">
<b>Automatic Control Understanding for Natural Programs</b></a>
Doctoral dissertation, Department of Computer Sciences,
The University of Texas at Austin.  <p>

<H3>Abstract</H3>

 Program understanding involves recognizing abstract concepts like
"read-process loop" in existing programs.  Programmers spend much of
their time understanding programs, so studying and automating the
process has many benefits.  <p>

Programming plans are units of programming knowledge connecting
abstract concepts and their implementations.  Existing research
assumes that plan instances can be recognized to recover the
programmer's abstract concepts and intentions, but this approach has
not been confirmed empirically.  <p>

We present a practical method for bottom-up control concept
recognition in large, unstructured imperative programs.  Control
concepts are abstract notions about interactions between control flow,
data flow and computation, such as "do loop", "read process loop", and
"bounded linear search".  They are recognized by comparing an abstract
program representation against a library of standard implementation
plans.  The program representation is a hierarchical control flow/data
flow graph decomposed into a tree of sub-models using propers (single
entry/exit control flow sub-graphs).  Plans are represented by similar
graphs with added qualifications.  Recognition is based on simple
matching between sub-models and plans.  The method was implemented in
the UNPROG program understander and tested with Cobol and Lisp source
programs.  <p>

This method is robust, efficient, and scalable.  The program
representation can be formed for all language constructs which permit
static determination of control and data flow.  Comparing sub-models
and comparisons increases linearly with program size.  <p>

UNPROG has been applied to automatic Cobol restructuring.  Knowledge
associated with plans and concepts permits more specific and
insightful transformation, code generation, and documentation than is
possible with syntactic methods.  Control understanding can similarly
raise the level of other reverse engineering and re-engineering tools
for applications like analysis, documentation, and translation.  <p>

We also showed how our method and UNPROG can be used for empirical
study of programs at the conceptual level.  Results can be used to
improve recognizer performance, acquire plans, catalog natural plans
and concepts, test the hypothesis that programs are planful, and
characterize program populations.  <p>

<hr>

<H2><a name="Akira">Geometrical Motion Planning for Highly Redundant Manipulators Using a Continuous Model</a></H2>


Akira Hayashi.  1991.
<a href="file://ftp.cs.utexas.edu/pub/qsim/papers/Hayashi-PhD-91.ps.Z">
<b>Geometrical Motion Planning for Highly Redundant Manipulators Using a Continuous Model</b></a>
Doctoral dissertation, Department of Computer Sciences,
The University of Texas at Austin.  <p>

<H3>Abstract</H3>

 There is a need for highly redundant manipulators to work in complex,
cluttered environments.  Our goal is to plan paths for such
manipulators efficiently.  <p>

The path planning problem has been shown to be PSPACE-complete in
terms of the number of degrees of freedom (DOF) of the manipulator.
We present a method which overcomes the complexity with a strong
heuristic: utilizing redundancy by means of a continuous manipulator
model.  The continuous model allows us to change the complexity of the
problem from a function of both the DOF of the manipulator (believed
to be exponential) and the complexity of the environment (polynomial),
to a polynomial function of the complexity of the environment only.  <p>

The power of the continuous model comes from the ability to decompose
the manipulator into segments, with the number, size, and boundaries
of the segments, varying smoothly and dynamically.  First, we develop
motion schemas for the individual segments to achieve a basic set of
goals in open and cluttered space.  Second, we plan a smooth
trajectory through free space for a point robot with a maximum
curvature constraint.  Third, the path generates a set of position
subgoals for the continuous manipulator which are achieved by the
basic motion schemas.  Fourth, the mapping from the continuous model
to an available jointed arm provides the curvature bound and obstacle
envelopes required (in step 2) to guarantee a collision-free path.  <p>

The validity of the continuous model approach is also supported by an
extensive simulation which we performed.  While the simulation has
been performed in 2-D, we show a natural extension to 3-D for each
technique we have implemented for the 2-D simulation.  <p>

<hr>

<H2><a name="Kay">Refining Imprecise Models and Their Behaviors</a></H2>


Herbert Kay.  1996.  
<a href="file://ftp.cs.utexas.edu/pub/qsim/papers/Kay-PhD-96.ps.Z">
<B>Refining Imprecise Models and Their Behaviors</B></a>.
Doctoral dissertation, Department of Computer Sciences, The University of Texas
at Austin, December 1996.

<H3>Abstract</H3>

This dissertation describes methods for simulating and refining
imprecisely-defined Ordinary Differential Equation (ODE) systems.
When constructing a model of a physical process, a modeler must cope
with uncertainty due to incomplete knowledge of the process.  For
tasks such as design and diagnosis, the effects of this uncertainty
must be considered.  However, predicting the behavior of an
imprecisely-defined model is not easy since the model covers a space
of many precise instances, each of which behaves differently.  <p>

While model uncertainty cannot be completely eliminated, it is
possible to reduce it.  Model refinement uses observations
of a physical process to rule out portions of the model space that
could not have produced the observations.  As more experience with the
physical process is gained, the imprecision in the model is further
reduced.  <p>

This dissertation describes three methods for reasoning with imprecise
ODE models.  SQSim is a simulator
that produces a guaranteed bound on the behavior of an imprecise ODE
model.  By using a multiple-level representation and inference methods
that span the qualitative-to-quantitative spectrum, SQSim produces
predictions whose uncertainty is consistent with model
imprecision.  We demonstrate SQSim on a complex, nonlinear chemical
process and compare it to other methods for simulating imprecise
ODE models.  <p>

MSQUID is a function estimator for fitting (and bounding) noisy data
that is known to be monotonic.  It uses a neural-network inspired
model and nonlinear constrained optimization to search a
space of monotonic functions.  We prove that MSQUID can estimate any
monotonic function and show that it produces better estimates than
does unconstrained optimization.  <p>