|
Monday, November 17, 2003 at 2:00
PM in
IST's Classrooms.
A 45 minute presentation will be followed by 15 minutes of discussion.
NAUTILUS (Navy AUTomated Intelligent Language
Understanding System)
Dr. Dennis Perzanowski
"Research dealing with intelligent interfaces between humans and machines has
frequently dealt with issues and problems involving a single user interacting
with a single machine, either or both of which are stationary.
Interactions between these entities is
constrained; namely, the human is most often seated at a computer terminal and
monitor. In the real world, however, people are free to point and free to move
about while they freely communicate with each other. Many of the solutions
obtained in more limited dialogs are applicable to multi-agent discourses, but
some have yet to be resolved. One such problem, inherent in all
discourses, concerns identifying unique points of reference in the real world.
In the first stage of our work, we have been
investigating a gestural solution to this problem. While other discourse
problems, not present in more limited dialogs, can only be resolved when more
complex discourses are considered, one problem we are investigating is
identifying a verbally ambiguous directional command. We propose a combined
linguistic and non-linguistic, specifically a gestural, solution to this
problem. In the first stage of this research effort, we investigate how gesture
is used to clarify and disambiguate elements of a dialog between two mobile
agents."
{See
http://www.aic.nrl.navy.mil/~wauchope/nautilus.html}
_____________________
Dr. Perzanowski performs research at the Naval research Laboratory (NRL)
pertaining to automated Natural Language understanding. He has investigated
the use of gesture incorporated vocalized speech as a means of enhancing
man-machine communications. He was also instrumental in developing the
NAUTILUS Natural Language understanding software. Furhter information
regarding his work is given below. His home page can be found at
http://www.aic.nrl.navy.mil/~dennisp/ .
Mon. July 21
11 a.m.
IST Classrooms.
A 45-minute presentation will be followed by 15 minutes of discussion
Understanding Video
Mubarak Shah
Recently, computer vision has gradually been making the transition away from
understanding single images to analyzing image sequences, or video
understanding. Video understanding deals with understanding video sequences,
e.g., recognition of gestures, activities, and facial expressions. In order to
make significant progress in video understanding, the original high-level
vision problem, which requires more qualitative than quantitative information,
and employs knowledge and context needs to be solved. Besides being able to
recognize, for instance, a set of predefined motions (gestures, expressions,
etc.), the video understanding system should have a "learning" capability.
The difference between a single image and a video is motion. The video contains
motion; the motion can be of objects present in the scene, the camera, or both.
Therefore, the emphasis in video understanding is on the use of motion to solve
some important problems. The motion occurs in 3D but is projected on 2D in
video images. The challenge is to solve these problems using 2D image motion.
In this talk, Dr. Shah will present his work, where only 2D information is used
to solve important research problems, and briefly discuss three example
projects: visual tracking of multiple objects in multiple uncalibrated cameras.
See project description at
www.cs.ucf.edu/~vision/projects/multipleCameras/multiple-cameras.html,
view-invariant representation and recognition of human action
www.cs.ucf.edu/~vision/projects/ViewInvariance/ViewInvariance.html,
and automatic synthesis of photo realistic video sequences using only two or
three images and without any 3D model or camera calibration
www.cs.ucf.edu/~vision/projects/viewMorphing/index.html.
Dr. Mubarak Shah, a professor of Computer Science, and the founding
Director of the Computer Visions Lab at University of Central Florida,
Orlando, is a researcher in computer vision, video computing and video
surveillance and monitoring [<http://www.cs.ucf.edu/~vision/projects/projects.html>].He
has supervised several Ph.D., MS, and BS students to completion, and is
currently directing fifteen Ph.D. and several B.S. students. He has published
close to one hundred articles in leading journals and conferences on topics
including visual motion, tracking, video registration, edge and contour
detection, shape from shading and stereo, activity and gesture recognition and
multi-sensor fusion
www.cs.ucf.edu/~vision/resume.pdf.
Dr. Shah is a fellow of IEEE, he was an IEEE Distinguished Visitor speaker for
1997-2000, and is often invited to present seminars, tutorials and invited
talks all over the world. He received Harris Corporation Engineering
Achievement Award in 1999, the TOKTEN awards from UNDP in 1995, 1997, and 2000;
Teaching Incentive Program awards in 1995 and 2003, Research Incentive award in
2003, and IEEE Outstanding Engineering Educator Award in 1997. He is co-author
of two books ("Motion-Based Recognition", Kluwer Academic Publishers, 1997; and
"Video Registration", Kluwer Academic Publishers, 2003), an editor of
international book series on ``Video Computing" by Kluwer
Academic Publishers, an editor in chief of Machine Vision and Applications
Journal, and an associate editor of the journal Pattern Recognition. In
addition he was an associate editor the IEEE Transactions on PAMI (1998-2002),
and a guest editor of the special issue of International Journal of Computer
Vision on Video Computing.
Wed. July 16
10 a.m.
IST Classrooms
A 45-minute presentation followed by 15 minutes of discussion
A Mathematical Model for an Aircraft Dynamics
Brian Goldiez
Mr. Goldiez will discuss the generation of the
mathematical model for an aircraft dynamics,
principally looking at where assumptions are made in what are viewed as
rigorous models. He will touch on the use of Z-Transforms in taking
these continuous models and making them discrete for ultimate conversion
into software.
This is a top level discussion that demonstrates why a broad view
of composability needs to be taken.
This area of research and discussion involves
many open issues. A few of
these issues include:
* methods for extracting models from data
* methods for making models have a tractable solution space
* methods for making models have tractable computer solutions
* model reusability
* model abstraction and compositing
Brian Goldiez is deputy director of IST.
Wed. July 9
11 a.m.
IST Classrooms.
A 45 minute presentation will be followed by 15 minutes of discussion.
Modeling the Decision Process of a Joint Task Force
Commander
John A. Sokolowski
Dr. Sokolowski will discuss research on modeling
the human decision process used by senior military commanders at the
operational level of warfare. The U.S. military uses modeling and simulation as
a tool to help meet its warfighting needs. A key element within military
simulations is the ability to accurately represent human behavior. This is
especially true in a simulation's ability to emulate realistic military
decisions. However, current decision models fail to provide the variability and
flexibility that human decision makers exhibit. Further, most decision models
are focused on tactical decisions and ignore the decision process of senior
military commanders at the operational level of warfare. In an effort to
develop a better decision model that would mimic the decision process of a
senior military commander, this research sought to identify an underlying
cognitive process and computational techniques that could adequately implement
it. Recognition-Primed Decision making (RPD) was identified as one such model
that characterized this process. Multiagent system simulation was identified as
a computational system that could mimic the cognitive process identified by RPD.
The result was a model of RPD called RPDAgent. Using an operational military
decision scenario to test model validity, decisions produced by RPDAgent were
compared against decisions made by military officers. It was found that
RPDAgent produced decisions that were equivalent to its human counterparts.
RPDAgent's decisions were not optimum decisions, but decisions that reflected
the variability inherent in those made by humans in an operational military
environment.
_______________________
John A. Sokolowski is a Senior Research Scientist at Old Dominion University's
Virginia Modeling, Analysis and Simulation Center. He holds a Bachelor of
Science in Computer Science from Purdue University, and both a Master of
Engineering Management and a Ph.D. in Engineering (Modeling & Simulation) from
Old Dominion University. Dr. Sokolowski served for 27 years in the U. S. Navy
as a submarine officer; his final Navy assignment was as Head, Modeling &
Simulation Division, Joint Warfighting Center, U. S. Joint Forces Command. His
research interests include human behavior modeling and multiagent system
simulation.
Wed. May 21
11 a.m.
IST Classrooms.
A 45 minute presentation will be followed by 15 minutes of discussion.
Intelligent Planning for and Cooperative Control
of Mobile Robots
Zhihua Qu
Dr. Qu will discuss intelligent planning for and cooperative
control of mobile robots. Particular emphasis will be given to research
funding opportunities such as the NSF.
Autonomous robots are ideal tools in many civilian and
military applications, and how to make robots intelligent and adaptive to
dynamically changing environment has been the focus of recent research in many
disciplines.
The aim of this talk is to present a steering control framework that enables
a mobile robot to operate in the presence of dynamically moving obstacles. The
main features of the proposed framework are twofold: analytical solution for
real-time reconfiguration, and explicit consideration of dynamics and
environment change to ensure performance and robustness. In addition,
cooperative control and other features can be incorporated into the framework.
Technically, the proposed framework is to determine a collision-free path
and corresponding steering control for motion in a dynamic and uncertain
environment by carrying out the following steps. First, by explicitly
considering a canonical model of nonholonomic systems, the family of feasible
trajectories and their corresponding steering controls are derived in a closed
form and also expressed in terms of one adjustable parameter for the purpose of
collision avoidance. Then, a new collision avoidance condition is developed for
the dynamically changing environment, it consists of a time criterion and a
geometrical criterion, and it has explicit physical meanings in both the
transformed space and the original working space. By imposing the avoidance
condition, one can determine one (or a class of) collision-free path(s) in a
closed form. Such a path meets all boundary conditions, is twice
differentiable, and can be updated in real time once a change in the
environment is detected. Solvability condition of the problem is explicitly
found. Simulation results are presented to illustrate the proposed framework.
Dr. Qu is Professor & Director of the Electrical
Engineering Program
in the School of Electrical Engineering & Computer Science, College of
Engineering & Computer Science, University of Central Florida. He received
his doctorate from the Georgia Institute of Technology in 1990 and his
research interests include controls, system theory, robotics and automation,
power systems.
Thurs. March 27
11:00 a.m.
IST Classrooms.
A 45 minute presentation followed by 15 minutes of discussion.
Simulation: Its Meaning For The Future
Gordon Vaeth
Mr. Vaeth will recall the early days of simulation and
look ahead to its use in achieving tomorrow's goals—including sending human
beings to Mars.
Formerly technical member for man-in-space a the Defense
Department's Advanced Research Projects Agency, and, later, Director of Weather
Satellite Systems Engineering and Operations for the National Oceanic and
Atmospheric Administration (NOAA), Gordon Vaeth began his career in simulation
and training.
As a member of the Navy Special Devices and Training
Device Center from 1947 through 1958, he worked the areas of aircraft,
missiles, underwater vehicles and ordnance. Before the creation of NASA, Mr.
Vaeth was projecting the role of simulation in preparing the way for space
flight.
Mr. Vaeth has authored six books on aerospace subjects
and served as Associate Fellow of the American Institute of Aeronautics and
Astronautics.
Tuesday, Feb 18
11:00 a.m.
IST classrooms
New Directions in Experimental Economics
Dr. Glenn W. Harrison
Dr. Harrison is the Dewey H. Johnson Professor of Economics
at the Moore School of Business of the University of South Carolina. His
research interests include Experimental Economics, Environmental and
Resource Economics, Law and Economics, Health Economics, and International
Trade Policy. Dr. Harrison obtained his Ph.D. in Economics in 1982 from
UCLA. Tuesday, Feb. 18
1:30 p.m.
IST classrooms
Beliefs in Strategic Decision Making
Dr. Elisabet Rutstrom
Dr. Rutstrom is an Associate Professor in the Economics Department at the Moore
School of Business of the University of South Carolina. Her main research
interest is Experimental Economics, where she studies peoples' behavior in a
variety of economic situations. In addition her interests include Applied Game
Theory, Applied General Equilibrium Analysis, International Trade and
Development, and Environmental Economics. Dr. Rutstrom obtained her Ph.D. in
Economics from the Stockholm School of Economics in 1990. |
