|
Past Medin Lectures 2004, 2005, 2006, 2007 Other years
Designing Scalable and Adaptive Heterogeneous Multiagent Teams
Dr. R. Paul Wiegand, ITT Industries Advanced Engineering and Sciences
Friday, Feb. 16, 2007
11:00 a.m.
Room 208, Partnership II bldg, 3100 Technology Pky
Orlando, FL (Central Florida Research Park)
Dr. Wiegand will present an extension to an existing artificial physics based representation for control behaviors of multiple interacting agents. The extension allows for natural incorporation of heterogeneous member behaviors in the team, and it is particularly well-suited for situations where team behaviors are based on
geometric formations (e.g., rings and lattices). This representation allows for a great deal of flexibility and adaptability in terms of the makeup of the team. In addition, he will discuss a principled method for constructing modular and scalable heterogeneous team behaviors that allows an engineer to incorporate domain knowledge during design.
Solutions to variants of a resource protection problem will be demonstrated. Finally, he will discuss how these ideas were applied to this problem in the context of physical robotic platforms, paying particular attention to the robotic team's ability to gracefully adapt to changes in team capabilities.
Dr. R. Paul Wiegand works for ITT Industries Advanced Engineering & Sciences and conducts research at the Naval Center for Applied Research in Artificial Intelligence, part of the Naval Research Laboratory in Washington, DC. His work at NRL began with a postdoctoral fellowship provided by the American Society for Engineering Education after receiving his
Ph.D. from George Mason University in 2004. Dr. Wiegand's general research interests are in the areas of theory of coevolutionary computation, multiagent systems, and application of coevolution for multiagent learning. He helped organize the coevolution workshop at the 2002 Genetic and Evolutionary Computation Conference and served as an invited
researcher at Dortmund University as part of the Collaborative Research Center‚ Computational Intelligence‚ (SFB 531) in 2003. He has also served as co-moderator of EC-Digest, a research-oriented email list that serves the evolutionary computation community; and as a member of the IEEE Task Force on Coevolution.
Social-Network and Simulation Views of Effective Teamwork
Alex Yahja, Ph.D.
Tuesday, Feb. 13, 2007
1:30 p.m.
3280 Progress Dr, Orlando, FL (Central Florida Research Park)
Real world events are often unpredictable and chaotic, and thus challenging. In particular, combat environments are especially challenging for teamwork. It is crucial to be able to adapt to real world events, and the most adaptive component in the human-robot teamwork systems is the human one. Dr. Yahja will talk about how
humans establish and dissolve communication and social networks from the social-network’s point of view.
He will explore a case of robot teamwork architectures built from a computer science point of view. Then he will describe a potential avenue to make teamwork more effective based on the lessons gathered from both social-network and robotic/computer science teamwork research. The social-network’s view includes the
consideration of organizations, a formal form of social-networks. Additionally, he will explore another view for teamwork: the simulation’s point of view. While simulation is not often thought of as an essential part of teamwork, he will show that it should be. He will also touch briefly on cyberinfrastructure.
Alex Yahja received the Ph.D. degree in computation, organizations and society from Carnegie Mellon University, Pittsburgh, PA in September 2006. He received two M.Sc. degrees, one in robotics and one in engineering and public policy, from Carnegie Mellon University in 2000 and 2004 respectively.
He is a Post-Doctoral Research Associate at the National Center for Supercomputing Applications (NCSA), University of Illinois at Urbana-Champaign, Urbana, IL. In his current work, he is doing research into social peer-to-peer disaster response systems and collaborative recommender systems for the Science of Networks in Communities (SONIC) group at NCSA.
In his graduate years, did research in mobile robotic systems, policy analysis, computational social and organizational science, and modeling and simulation. His dissertation deals with how to automate the validation and model-improvement of simulations. A new conceptualization is developed and the resulting tool is applied to agent-based simulations of
disease spread and in social-network workers-management simulations of impeding strikes.
His research interests span modeling and simulation, model validation and improvement, machine learning, hypothesis formation and testing, social networks, collaboration/teamwork and recommender systems, robotics, computational management and organization science, knowledge-based systems, policy analysis, cyberinfrastructure, and artificial intelligence.
Military Troops and Disabled Veterans:
Using Intelligent Agents to Cope with Complex, Time Sensitive Information
Wayne Shebilske Ph.D.
Friday, Jan. 26, 2007
2:00 p.m.
3280 Progress Dr, Orlando, FL (Central Florida Research Park)
Dr. Shebilske is applying his research into training and intelligent displays to learning with disabilities. His key insight is that disabilities slow and distort perceptions and/or responses to increase time sensitivity and information processing complexity. His motivation for this extension comes from a needs analysis he
did of disabled students at Wright State University and related needs of thousands of soldiers in Iraq and Afghanistan who are surviving severe blasts because of (a) body armor that protects vital organs and (b) prompt, effective care in the field. The Veterans Administration is responding by treating and rehabilitating this new generation of disabled
veterans, whose injuries include multiple amputations, low vision, hearing loss, spinal cord injuries, and other neurological losses. He thinks Human Factors and IO psychologists can play an important role in responding to these needs.
Dr. Shebilske's current research also includes Benchmarked Experiential Systems for Training (BEST) and Intelligent Icons for Meta-information Adaptive Presentation (I2MAP). He is applying this research to Airborne Warning and Control System (AWACS) teams and Dynamic Targeting Cell (DTC) teams of the Air and Space Operations Center and is observing
simulated missions of these Air Force command and control tasks in which information is overwhelmingly time sensitive and complex. He combines training and intelligent displays to help troops cope with the time sensitivity and complexity. Both interventions employ mathematical benchmarking of near-optimal performance.
Dr. Shebilske received his Ph.D. in psychology at the University of Wisconsin, Madison, in 1974. After nine years on the faculty of the University of Virginia Psychology Department (Assistant Professor, 1974-1979: Associate Professor 1979-1983), Dr. Shebilske served as Study Director for the Committee on Vision at the National Academy of Sciences
(1983-1985). Dr. Shebilske was a professor at Texas A&M University (1985-1998) where he was the Graduate Program Advisor (1993-1998). He conducted research at the Aberdeen Proving Ground, Brooks Air Force Base and Lackland Air Force Base. He spent the summer of 1985 at the Center for Interdisciplinary Research, Bielefeld, Germany. In 1992, he was awarded
a National Research Council Senior Research Associateship at the Armstrong Laboratory Intelligent Training Branch. He conducted research at the Aberdeen Proving Ground, Brooks Air Force Base and Lackland Air Force Base. He spent the fall of 1998 at Embry-Riddle Aeronautical University in Daytona Beach, Florida. He has worked closely with government,
military, and private agencies identifying critical issues for study in many areas including design and development of aerospace systems, visual display equipment, medical devices, standards for pilots and drivers, automated instruction for complex skills, and virtual reality systems.
Emerging Semantic Web Trends:
Transparent and Trustworthy Applications and Semantically-Enabled
Scientific Data Integration
Deborah McGuinness, Ph.D.
Friday, Jan. 12, 2007
11:00 a.m.
Partnership II bldg., 3100 Technology Pkwy, Orlando, FL (Central Florida Research Park)
As web applications proliferate, more users (both people and agents) find themselves faced with decisions about when and why to trust application advice. In order to trust information obtained from arbitrary applications, users need to understand how the information was obtained and what it depended upon.
Particularly in web applications that may use question answering systems that may be heuristic or incomplete or data that is either of unknown origin or may be out of date, it becomes more important to have information about how answers were obtained. Emerging web systems will return answers augmented with Meta information
about how answers were obtained. In this talk, Deborah McGuinness will describe an approach that can improve trust in answers generated from web applications by making the answer process more transparent.
The added information is aimed to provide users (humans or agents) with answers to questions of trust, reliability, recency, and applicability. The talk will include descriptions of a few representative applications using this approach explaining cognitive assistants that learn and intelligence analyst tools.
The talk will also briefly highlight work on semantically-enabling access to and integration of scientific data. Examples will be taken from Deborah’s work on the NSF-funded Virtual Solar Terrestrial Observatory and the NASA-funded Semantically-Enabled Scientific Data Integration projects.
Dr. McGuinness is the acting director and senior research scientist at the Knowledge Systems, (KSL) Artificial Intelligence Laboratory at Stanford University. She is a leading expert in knowledge representation and reasoning languages and systems and has worked in ontology creation and evolution environments for over 20
years. Most recently, Deborah is best known for her leadership role in semantic web research, and for her work on explanation, trust, and applications of semantic web technology, particularly for scientific applications.
Past Medin Lectures
2004, 2005, 2006, 2007 Other years
Smaller, Faster & Immersive: New
Directions for Virtual Environment Design
Larry Davis, Ph.D.
Wednesday, Feb. 22, 2006
Room 112, Partnership II building
3100 Technology Pkwy, Orlando, FL (Central Florida Research Park)
The increased computational power of
microprocessors has enabled fantastic miniaturization of devices and
unprecedented functionality. From cellular phones, to the iPod, to the
Blackberry and to personal DVD players, devices are smaller, faster and deliver
an ever-increasing amount of features. Moreover, these features are often
complementary, e.g., a cellular phone that plays music files.
With the integration of these devices into everyday life, the commercial
success or failure of a device is often linked to the transparency of its user
interface. By transparency, we mean the level at which using a device is
comfortable, natural, socially appropriate, and/or intuitive.
Unfortunately, virtual environments are typically not very transparent.
However, by taking advantage of the trend toward smaller devices with
complementary features and advanced sensing, we can begin to increase the
transparency of virtual environments. These increasingly transparent systems
will have direct aid in visualization, training, and augmented cognition.
Dr. Davis will summarize his research accomplishments and how they have shaped
his thinking regarding interfaces. He will share his insights into
miniaturization and its pertinence to virtual environments, and discuss
research directions that may be pursued to increase virtual environment
transparency.
Larry Davis is currently a Senior Instructional Assistant at Valencia Community
College and a Co-Founder of Tirrion Technologies, a start-up company in the
University of Central Florida Technology Incubator. His research interests are
in the area of virtual environments and his work spans a range of topics,
including the optimization of tracking scenarios, the creation of novel virtual
environments, and the implementation of user assessment tools. He has authored
or co-authored more than thirty research papers and presentations on these
topics.
Dr. Davis received his Ph.D. in Electrical Engineering from the University of
Central Florida in 2004. He is a McKnight Doctoral Fellow and has participated
in research programs with Sandia National Laboratories. He serves on the
editorial board of the FEF Journal of Interdisciplinary Research.
Redirected Walking in Virtual
Environments
Sharif Razzaque
Wednesday, Feb 15, 2006
10:00 a.m.
IST Conference Room, Partnership II building
3100 Technology Pkwy, Orlando, FL (Central Florida Research Park)
There are many different techniques for allowing users to specify locomotion in
human-scale immersive virtual environments (VEs). These include flying with a
hand-controller, using a treadmill, walking-in-place, and others. Real walking,
where the user actually and physically walks in the lab, and virtually moves
the same distance and in the same direction in the virtual scene, is better
than flying. It is more input-natural, does not require learning a new
interface, results in a greater sense of presence, and theoretically results in
less simulator sickness.
One serious problem with real walking, however, is that the size of the virtual
scene is limited by the size of tracked area. For example, for an architect to
really walk in a virtual prototype of a house, the tracked area must be as
large as the house. This requirement makes real walking infeasible for many
facilities and virtual scenes.
To address this limitation, Dr. Razzaque has developed Redirected Walking,
which by interactively and imperceptibly rotating the virtual scene around the
user makes the user turn her/himself. Under the right conditions, Redirected
Walking would cause the user to unknowingly walk in circles in the lab, while
thinking he/she is walking in a straight and infinitely long path in the
virtual scene.
Dr. Razzaque will discuss Redirected Walking’s theoretical and physiological
underpinnings, and show that it can be used:
1) To make the user turn themselves
2) Without causing the user to be aware of Redirection
3) Without unacceptably increasing the user’s level of simulator sickness and,
most importantly,
4) To useful effect:
· In head-mounted display (HMD) systems, the user can experience a virtual
scene larger than the lab while also having the benefits of real walking.
· In an open-backed, three-walled CAVE, users can have the increased presence
and input-naturalness normally associated with a fully enclosed CAVE.
Ideas for VE training scenarios that address several limitations of current HMD
VE systems, such as field-of-view and latency will also be discussed.
Sharif Razzaque has just earned a PhD in Computer
Science from The University of North Carolina at Chapel Hill, under Fred
Brooks. His interests are human computer interaction, virtual environments, and
building real, deployable, systems. He has previously worked on physiological
assessment of presence, collaborative satellite-engineering tools at Lockheed
Martin, and cochlear implants at the University of Michigan.
Past Medin Series Presentations
(2005)
2004 lectures 2006 lectures
May 3, 2005
11:30 a.m.
IST Classrooms, 3280 Progress Drive
The Team Performance Lab: R&D for training and evaluating
real-life skills in high-consequence applications
Dr. Florian Jentsch
Dr. Jentsch will discuss the Team Performance
Laboratory at UCF which he directs. Performance is critical in complex and
dynamic environments where teams are required to adapt to uncertain information
and where the cost of team error is high. The Team Performance Lab studies ways
to use modeling and simulation not only to provide safe and cost-effective
training solutions, but also to ensure the possibility of tracking, measuring
and critiquing the effect of training on individuals and teams.
Florian Jentsch, Ph.D., is an associate scientist/scholar in the Department of
Psychology at the University of Central Florida. He is also the current
director of the Team Performance Laboratory, a position he took over in the
Fall of 2003 after he had been Assistant and Associate Director of the Lab
since 1997 and 2001, respectively.
Dr. Jentsch received his Ph.D. in human factors psychology from the University
of Central Florida in 1997. He also holds master’s degrees in aeronautical
engineering from the Technical University of Berlin (summa cum laude) and in
aeronautical science from Embry-Riddle Aeronautical University. His doctoral
dissertation on training for junior commercial flight crew members won the
American Psychological Association’s 1998 George E. Briggs Award for the best
dissertation in applied/experimental psychology, and he was also awarded the
American Psychological Association’s 2002 Earl Alluisi award for Early Career
Achievement in applied/experimental psychology.
Dr. Jentsch’s interests center on research and development that supports the
training and evaluation of complex knowledges and skills, especially in
high-consequence environments. Application areas for his research have been in
team training, human-robot-interaction, aviation human factors, cross-cultural
research, research methodology, and simulation where he has co-authored over
150 publications, presentations, and technical reports. He has received grants
and contracts from the U.S. Army (ARL-HRED, ARO, PEO-STRI, and RDECOM-STTC),
the U.S. Navy (NAWC-TSD, NAVAIR Orlando), several units within the Federal
Aviation Administration (FAA), and from the Transportation Security
Administration (TSA). He also has consulted on system and software development
projects for various federal and state government sponsors, including the
National Institutes of Health (NIH), the National Science Foundation (NSF), the
U.S. Navy, U.S. Army, DoT-FAA, DoT-TSI. NASA, and Workforce Central Florida (WCF).
Feb. 24, 2005
11:00 a.m.
IST Classroom C
What's Wrong with Virtual?
The Whereabouts of Marketing Virtual Reality Value
Dr. Patrick Corsi
• Are we yet “on the threshold of
the Age of Magic” (Fruchterman, 1992)?
From the past hype of the early nineties, Virtual Reality fell well under the
past years promises. While the potential remains intact, a deep transition is
going on mostly underpinned by such notions as presence and engagement. This
necessarily calls for revisiting the demerits of VR.
• Do we yet know how can VR make our economy more competitive?
VR equipment is still expensive if not encumbering, hardly real-time and not so
immersive; user training and side-effects restraint the use to e.g.
entertainment. A shared perception of a relatively immature technology hasn’t
much disappeared. Can VR functional added-value be deployed then in e.g. health
care, industrial maintenance and distance operations? How? That is the
marketer’s address here…
• But wait! There is more: VR is a key enabler that impacts design methods and
models, products life-cycles, distance operations…
Specific benefits include the virtual as a training paradigm substituting to
the real; in a nutshell, VR is to be positively found at the core of the
accelerated timings of tomorrow’s innovations. The virtual paradigm
re-positions users within new business models that may capture the value of
tomorrow markets…
This talk explains how to relinquish the intractable of VR and the difficulties
inherent to VR weaknesses. Can we then hopefully delineate the modalities for
mass field adoption so virtual comes of age for… real?
_____________________
Patrick CORSI first worked for IBM
(San Jose Research Lab; La Gaude telecommunications plant); then THOMSON-CSF
(directing AI group), grew a start-up in artificial intelligence over the 80’s
in Paris; worked at the European Commission in Brussels in charge of monitoring
the European AI portfolio of projects over the nineties. He is now an
international consultant in innovation engineering - KINNSYS, Brussels; an
Associate Professor at ISTIA Innovation, University of Angers and Laval on VR
entrepreneurship and an Executive Advisor to the European Multimedia Forum. He
can be reached at patrick.corsi@skynet.be.
Thursday,
February 3, 2005
11:00 AM
IST's Classrooms
3280 Progess Drive
45 minute presentation followed by 15 minutes of discussion.
Teamwork in the
Control of Unmanned Vehicles: Challenges, Progress, and Lessons Learned
Dr.
Jean MacMillan
During 2001-2003,
the Defense Advanced Research Projects Agency (DARPA) funded a program on
“mixed initiative control” of teams of unmanned robotic vehicles. This
presentation will discuss the goals of the program, the cognitive engineering
challenges associated with mixed initiative control of teams of unmanned
vehicles, the progress that was made in developing human control concepts and
some of the lessons learned from the effort that can inform future research.
_____________________
Jean MacMillan
is the Chief Scientist at Aptima. Jean holds a Ph.D. in Cognitive Psychology
from Harvard University, an M.C.P. from Harvard University, and a B.A. from
Antioch College. Her 20-year research career has spanned a broad range of
topics in human-machine interaction and user-centered system design, including
adaptive instructional design, team decision-making, command center design, and
human performance measurement in simulation environments.
At Aptima, Jean
has served as the Principal Investigator for projects focusing on the
development of reliable and valid performance measures for teams of F-16 pilots
training in a distributed simulation facility, the design of synthetic entities
that function as team members for simulation-based training of teamwork skills,
the assessment of team performance and development of team performance measures
for AWACS Weapons Director teams, and the development of optimized model-based
manning reduction strategies that will allow the operation of Navy ships with a
reduced number of personnel. Jean is also co-designing a computer-based
adaptive tutor for improving the reading comprehension of adult readers.
Jean is a
member of the Human Factors and Ergonomics Society (HFES), and the Cognitive
Science Society. She was one of the founding members of the Cognitive
Engineering and Decision-Making Technical Group within HFES. Jean serves on the
Editorial Board of the Human Factors journal.
Past Medin Series Presentations
(2004)
2006 lectures 2005 lectures
Nov. 30, 2004
1:00 p.m.
IST Classrooms
Virtual Reality Assets for Assessment, Therapy and Rehabilitation
Albert “Skip” Rizzo, Ph.D.
Institute for Creative Technologies and The School of Gerontology
University of Southern California, Los Angeles, CA.
arizzo@usc.edu
Virtual Reality (VR) has now emerged as a promising tool in many
domains of assessment, therapy and rehabilitation (Rizzo, Schultheis, Kerns &
Mateer, 2004; Weiss & Jessel, 1998; Zimand, Anderson, Gershon, Graap, Hodges, &
Rothbaum, 2002; Glantz, Rizzo & Graap, 2003). Continuing advances in VR
technology along with concomitant system cost reductions have supported the
development of more usable, useful, and accessible VR systems that can uniquely
target a wide range of physical, psychological, and cognitive rehabilitation
concerns and research questions. What makes VR application development in the
therapy and rehabilitation sciences so distinctively important is that it
represents more than a simple linear extension of existing computer technology
for human use. VR offers the potential to create systematic human testing,
training and treatment environments that allow for the precise control of
complex dynamic 3D stimulus presentations, within which sophisticated
interaction, behavioral tracking and performance recording is possible. Much
like an aircraft simulator serves to test and train piloting ability, virtual
environments can be developed to present simulations designed to assess and
rehabilitate human functional performance under a range of stimulus conditions
that are not easily deliverable and controllable in the real world. When
combining such assets within the context of relevant, ecologically enhanced
Virtual Environments, a fundamental advancement could emerge in how human
performance can be addressed in many rehabilitation disciplines. This talk will
provide a brief introduction to VR and describe the assets that are available
with VR applications in assessment, therapy and rehabilitation. Within that
context, examples of virtual environments that illustrate each asset will be
presented. The value of a multidisciplinary approach for the design and
implementation of VR will be emphasized with examples spanning the fields of
rehabilitation, psychology, neuroscience, physical therapy, occupational
therapy, special education and social work.
June 9, 2004
11 a.m.
IST Classrooms
45 minute lecture followed by 15 minutes of discussion
IHMC - A Habitat for Innovation
Kenneth M. Ford
Kenneth Ford will survey the activities at the Institute for Human & Machine
Cognition (IHMC), established in the early '90s as a research unit of the
University of West Florida. At its inception, IHMC consisted of two part-time
employees and occupied one small room. In its first decade, IHMC has grown into
one of the nation's premier IT-related research institutes with over 115
researchers and staff. Researchers at IHMC pioneer ground-breaking technologies
aimed at leveraging and extending human capabilities.
Our human-centered approach results in systems
that can be regarded as cognitive prostheses, much as eyeglasses are a sort of
ocular prosthesis. Our systems fit the human and machine components together in
ways that exploit their respective strengths and mitigate their respective
weaknesses. The design and fit of computational prostheses require a broader
interdisciplinary range than is typically found in one organization, and IHMC
staff includes computer scientists, cognitive psychologists, neuroscientists,
physicians, philosophers, engineers and social scientists of various stripes,
as well as some people who resist all attempts to classify them. Innovation is
place-based.
IHMC's presence in the heart of historic
downtown Pensacola allows continuous interactions and frequent collisions of
people and ideas. This stimulating, human-scale urban environment is rapidly
becoming a habitat for innovation. IHMC currently occupies 32,000-sq. ft. of
space in two buildings renovated building in the historic district. In
addition, IHMC has recently opened an office in Silicon Valley.
See
http://www.ihmc.us/index.php and
http://www.uwf.edu/uwfMain/
Kenneth Ford is Founder and Director of the Institute for Human &
Machine Cognition (IHMC). Ford's research interests include: artificial
intelligence, cognitive science, human-centered computing, and entrepreneurship
in government and academia. Ford is the author of over a hundred scientific
papers and five books. He received a Ph.D. in Computer Science from Tulane
University. He is the Editor-in-Chief of AAAI/MIT Press and has been the editor
of several journals. Dr. Ford is a Fellow of the American Association for
Artificial Intelligence.
In January 1997, Dr. Ford was asked by NASA to
develop and direct its new Center of Excellence in Information Technology at
the Ames Research Center. He served as Associate Center Director and Director
of NASA's Center of Excellence in Information Technology. Dr. Ford was awarded
the NASA Outstanding Leadership Medal in 1999. Dr. Ford serves on numerous
charitable, government, and corporate boards. In October of 2002, President
George W. Bush nominated Dr. Ford to serve on the National Science Board. His
nomination was confirmed by the U.S. Senate in March of 2003.
May 10, 2004
11:00 a.m.
IST's Classroom C
A 45 minute presentation will be followed by 15 minutes of discussion.
Evolution of the Standard Simulation
Architecture
Jeffery S. Steinman, RAM Laboratories
This seminar provides an
overview of the proposed Standard Simulation Architecture (SSA) that addresses
interoperability between (1) simulations in a federated environment, (2)
entities within a simulation potentially executing in a parallel and/or
distributed environment, and (3) components residing within entities. These
three categories of composability have different peer-to-peer interoperability
overheads that must be addressed specifically. The SSA defines a highly
flexible infrastructure to map federates, entities, and components to hardware
systems. The proposed Standard Simulation Architecture is defined in abstract
layers of functionality with well-defined interfaces to promote technology
infusion from R&D organizations such as universities, government laboratories,
and industry. This seminar will provide an overview of these layers and their
supported composable modeling approach.
Dr. Jeffrey S. Steinman
received his Ph.D. in High Energy Physics from the University of California,
Los Angeles in 1988. He has been involved in numerous parallel and distributed
simulation programs at the Jet Propulsion Laboratory, Metron Inc, and RAM
Laboratories. As the principle developer of the Synchronous Parallel
Environment for Emulation and Discrete Event Simulation (SPEEDES) framework,
Dr. Steinman has provided technical oversight on a number of very large DoD
simulation programs including: JSIMS, MDWAR, EADTB, JMASS, HPC-RTI, JSAF, and
others. This work resulted in five patents and more than 40 technical papers in
the area of parallel discrete event simulation. Dr. Steinman is currently
serving as Vice President and Chief Science & Research Officer with RAM
Laboratories, Inc.
January 29,
2004
10:00 AM
IST's Classrooms.
A 45 minute presentation will be followed by 15 minutes of discussion.
Human Language Technology at the Naval Research Laboratory
Elaine Marsh
The Navy Center for Applied Research in Artificial
Intelligence (NCARAI) has been involved in both basic and applied research in
artificial intelligence since its inception in 1982. The research program of
the center is directed toward understanding the design and operation of
computer systems capable of improved performance based on experience; efficient
and effective interaction with other systems and with humans; sensor-based
control of autonomous activity; and the integration of varieties of reasoning
as necessary to support complex decision-making. The emphasis at NCARAI is the
linkage of theory and application in demonstration projects that use a full
spectrum of artificial intelligence techniques.
The center's Multi-Modal Multi-Media Systems
section performs exploratory research in the area of man-machine interactions,
specifically natural language and multi-modal interfaces, natural language and
speech system evaluation, and textual knowledge management. This talk will
present an overview of the various research projects developing natural
language man-machine communication capabilities.
(See http://www.aic.nrl.navy.mil)
_____________________
Elaine Marsh is section head and group leader for the Multi-Modal Multi-Media
Systems section in the Navy Center for Applied Research in Artificial
Intelligence at the Naval Research Laboratory, where she has been employed
since February 1983.
She has conducted basic and exploratory research
in the areas of interactive systems, specifically natural language and
multi-modal interfaces, natural language and speech system evaluation, and
textual knowledge management.
Prior to joining the Naval Research Laboratory,
she was employed as a research scientist on the
Linguistic String Project at New York University. Her current research
interests include discourse analysis in natural language, human-machine
interfaces, speech for virtual worlds, knowledge representation for narrative
texts, and human-language system evaluation.
IST Research Grant Series Lectures
The Simulation Dependence Graph
Keith Garfield
Tuesday, March 14
1:00 p.m.
IST Progress Drive building
The Simulation Dependence Graph (SDG), is a novel
graphical representation of simulation systems. It was developed at IST by
adapting the Program Dependence Graph (PDG) to the simulation domain. The PDG,
and the related Java System Dependence Graph (JSDG), is a popular program
representation used in computer science and software engineering that allows
sophisticated program analysis. The SDG variation of the PDG captures critical
dependences among simulation attributes to allow analysis of the computations
comprising the simulation and their relationships to one another. Based on the
success and widespread use of the PDG, development of the SDG will provide a
powerful analytical tool to the Modeling and Simulation Community at large.
SDG
is composed of a node set and three edge sets. The node set represents
attributes of the real world entities being simulated, their allowable value
ranges, and the functions that update the attribute values during the course of
a simulation. The three edge sets represent data dependences, entity
memberships, and sequencing constraints. The data dependence edges explicitly
model input-output interactions between attributes; this information can be
used to analyze and optimize system behavior in the same manner that modern
optimizing compilers perform program data-flow analysis. The entity membership
edges explicitly associate attributes belonging to the same physical entity;
this information can be used to distribute entity-dependent behaviors and
constraint across a set of nodes representing a single simulated object. The
sequencing constraint edges explicitly represent the order in which attribute
values are updated; these edges may be used to demonstrate the efficiency of
one manner of scheduling tasks over another. The sequencing edges may model
uni-processor or parallel processor environments.
The advantages of representing simulations in
this format are many. Graphs are well-known and understood mathematical
structures that have been proven to be useful in a many problem domains. More
specifically, PDG’s have been shown valuable in the areas of automated program
optimization and performance enhancement. Program analysis techniques
developed using the PDG may be leveraged to allow static and dynamic analyses
of simulation systems. Candidate problems are listed below:
-
Syntactic
composability can be determined from the graph structure.
-
Semantic
composability may be approximated by extending PDG analysis techniques such
as “program slicing” to the SDG.
-
Improvements to
event scheduling may be attained by adapting optimizing techniques used in
compilers. These techniques involve specialized analysis such as “program
slicing” to identify past computations affecting a current computation.
|
