The Use
of Development Tools in Software Engineering Projects
Dr. Janusz Zalewski, UCF Electrical and Computer Engineering
April 14, 1999
Abstract:
Correlation
Error in Multiple Resolution Entity Simulations
Dr. Robert Franceschini, Institute for Simulation & Training
May 19, 1999
Abstract:
A Multiple Resolution Entity Simulation (MRES) represents a real-world
system using a combination of cooperating simulations that run at
different resolutions. An MRES can dynamically change the resolution of a
simulated entity by transferring the entity from one to another of the
component simulations of the MRES. Correlation error occurs when the
sequence of simulated entity transfers (i.e., the resolution of the
simulated entities) changes the simulation results. An MRES without
correlation error allows a user to arbitrarily select resolutions for
different parts of the simulated scenario. The Correlation Error Problem
seeks to reduce or eliminate correlation error. One possible application
of this work is in battlefield simulations that represent military
hierarchies.
This presentation introduces an appropriate
methodology and infrastructure to build the foundations of an original
theory of MRES. Our Simple Multiple Resolution Entity Simulation (SMRES)
includes two simulations (aggregate and atomic) and allows simulated
entities to be transferred between these simulations. Using SMRES we show
that correlation error has two components. The first component is due to
the initial conditions of the resolution change. The second component is
due to the differences between the aggregate behavior and the average
behavior of the atomic entities during simulation. We propose a new
definition of consistency at the aggregate level independent of variations
in the atomic level. We define correlation error quantitatively and
provide the first techniques for measuring and removing correlation error.
Historical
Overview of Intelligent Agent Architectures
Dr. Douglas Reece, SAIC
July 16, 1999
Abstract:
This talk will briefly describe the development of intelligent agent
architectures over the last 30 years. "Intelligent agents" are
intelligent systems that act autonomously in a complex, dynamic
environment. Intelligent systems started with logic-based problem-solving
systems and grew through planning systems in the 70s and mobile robots in
the 80s. In the mid 80s behavior-based robotics arose to challenge
traditional plan-execute systems as a means to achieve competence and
robustness with primitive tasks. This emphasis on primitive but robust
autonomous behavior was extended by the Artificial Life community. In the
last decade there have been software implementations of cognitive models
and attempts to tie all intelligent activity together in one integrated
architecture. The talk will touch on all of these areas and highlight
important systems in each. Applicability to ModSAF may also be mentioned.
Douglas A. Reece is a Senior
Scientist at SAIC in Orlando. He has been developing physical and
behavioral models for individual combatant CGFs for four years. He was the
Principal Investigator on the project to develop Computer Controlled
Hostiles for the Marine Corps' Team Target Engagement Simulation, and is
now the software architect for DISAF. He received his Ph.D. in Computer
Science from Carnegie Mellon University in 1992.
Learning
Agents in Simulation and Training
Dr. Erol Gelenbe, UCF School of Computer Science
August 13, 1999
Abstract:
Advances
in the last ten years in the area of learning algorithms, and the
resulting tools, now enable the development of autonomous adaptive
entities which respond to external stimuli by changing their behavior in
the course of a discrete event simulation. In this presentation we will
describe learning tools based on the Random Neural Network model, and
describe an application to the design of goal based learning in a specific
simulation environment.
Erol Gelenbe is the Director of the
University of Central Florida’s School of Computer Science. He has
earned a D.Sc. in Applied Mathematics from the University of Paris, a
Ph.D. in Electrical Engineering from the Polytechnic Institute of
Brooklyn, a M.S. in Electrical Engineering from the Polytechnic Institute
of Brooklyn, and a B.S. in Electrical Engineering from the Middle East
Technical University. Dr. Gelenbe is interested in developing biologically
inspired computer and mathematical models that are applicable to the study
of natural and artificial learning systems and in algorithms and
techniques which are biologically based. He has published over 100 journal
articles and four books.
Advancements
in Environment Representation
Michael Craft, SAIC
September 24, 1999
Abstract:
This lecture
will cover two new simulation software methods developed for the WARSIM
project but having wider applicability. First, software techniques for the
encapsulation of metrics (e.g., distance and speed) and of the coordinate
system will be presented. The encapsulation methods are based on software
engineering principles and provide substantial benefits in terms of
reliability and reusability. Second, a new method for the encoding of
terrain triangle information will be described. The encoding is very
compact and avoids some classic page boundary and page fault problems that
often arise in terrain databases. Data driven terrain features will also
be discussed.
Michael A. Craft is
a Senior Software Engineer with SAIC, working on the WARSIM project. His
responsibilities include the design and development of diverse support
facilities for WARSIM and JSIMS, including ECS and the triangle encoding,
overseeing the detailed architecture of the various elements of
Environment support, and primary authorship of the WARSIM software
standards. Mr. Craft holds an M.S. in Computer Science and an M.S. in
Mathematics, both from the University of New Hampshire. His major
interests include software engineering, computer languages, simulation,
and computer protocols.
Optical
Motion Capture for Application to the Virtual Reality Dynamic Anatomy Tool
Dr. Jannick Rolland, UCF School of Optics/Electrical and
Computer Engineering/School of Computer Science
Date rescheduled - TBA
Abstract:
Models
of the Evolution of the Immune System Using Genetic Algorithms
Dr. Rebecca Parsons, UCF School of Computer Science
November 17, 1999
Abstract:
An organism's immune system protects it from invasion of foreign matter
(viruses, bacteria, etc). The antibodies involved in the body's immune
response can be categorized along several axes. One such axis is the time
scale over which the antibody evolved. The body generalist antibodies that
have evolved across generations and specialist antibodies that specific to
the antigens the individual has encountered in his lifetime. Antibodies
can also be categorized on the basis of specificity. The generalists, as
the name implies, tend to recognize broad classes of antigens, while
specialists are tuned to a particular antigen. This talk describes our
work in understanding what evolutionary mechanisms and selection pressures
could give rise to this diverse suite of antibodies. We will present some
theoretical results on diversity within an evolutionary context and them
some experimental results that demonstrate the expected evolution.
Dr. Rebecca Parsons received her Ph.D. in
Computer Science from Rice University. She was awarded a Los Alamos
Director's Post Doctoral Fellowship; while she was at Los Alamos, she
worked with Dr. Stephanie Forrest on genetic algorithms applied to various
problems in computational biology, including immune system modeling and
DNA fragment assembly. She is currently an Assistant Professor in the
School of Computer Science at the University of Central Florida.
Building
Intelligent Synthetic Characters for Computer Games
John E. Laird, Professor
of Electrical Engineering and Computer Science, University of Michigan
December 6, 1999
Abstract:
Synthetic characters in computer games usually fall short
of human players, struggling to exhibit even a modicum of intelligence. However,
progress is being made. Characters in recent games have progressed to include
limited forms of situation-based
reasoning, communication, and cooperation. The continued improvement in the
intelligence of synthetic characters should lead to significant improvements in
game play as well as new gaming experiences. Dr. Laird’s work in developing
characters for computer games using the Soar architecture tries to push even
further toward human-like behavior. In this talk, He reviews research on Soar,
an architecture for building AI systems and psychological models of human
behavior. He also briefly describes development of TacAir-Soar, a real-time
expert system that flies U.S. military air missions in simulation, and is used
for training in the U.S. Air Force. Experience building TacAir-Soar is now being
applied to building human-like synthetic characters for computer games, which
include Descent 3 and Quake II. Two hypotheses underlying this talk are that AI
architectures such as Soar can greatly improve the cognitive capabilities of
synthetic characters and speed development using modest computational resources;
and that computer games provide a challenging (and cool) environment for
research in AI.
Dr. John Laird received a Ph.D. from
Carnegie Mellon University in 1983. He is currently a Professor of Electrical
Engineering and Computer Science at the University of Michigan. He has spent the
last 20 years doing research in Artificial Intelligence. His goal is to develop
human level AI systems and his approach has been to concentrate on the
underlying cognitive architecture - the primitive processing and memory
structures that support cognitive activity, such as reasoning, problem solving,
planning, language, and learning. He is an original developer of the Soar
architecture and leads its continued development and evolution. Soar is used
worldwide by researchers for building AI systems and modeling human behavior. He
led the development of TacAir-Soar, a real-time expert system that flies U.S.
military air missions in simulation. He teaches a senior-level design class in
computer game development and his current research focuses on AI and computer
games. He was an organizer of the AAAI Spring Symposium on AI and Computer Games
and has presented papers at CGDC '98 and GDC '99.
Methods and Tools
in Computer-Supported Taskforce Training
Dr. Johan Jenvald, Chief Scientist, Operational Development
Department, Swedish Armed Forces Naval Centre
December 7, 1999
Abstract:
Efficient training methods are
important for establishing, maintaining and developing taskforces that are
organized to manage complex and dangerous situations in order to serve and
protect our society. Furthermore, the technical sophistication of various
systems in these organizations, for example command, control and
communication systems is growing, while the resources available for
training are being reduced due to budget cuts and environmental
restrictions. Realism in the training situation is important so that the
actual training prepares the trainees for, and improves the performance
in, real situations. The ability to observe and review the training course
of events is crucial if we want to identify the strength and shortcomings
of the trained unit, in the overall effort to improve taskforce
performance.
The research results describe and
characterize methods and tools in computer-supported training of multiple
teams organized in taskforces that carry out complex and time-critical
missions in hazardous environments. We describe a common framework that
consists of a training methodology together with a system architecture for
an instrumentation system that can provide different levels of computer
support during the different training phases. In addition, we use two case
studies to describe the application of our methods and tools in the
military force-on-force battle-training domain and the emergency
management and response domain, respectively.
Our approach is to use an observable
realistic training environment to improve the training of teams and
taskforces. There are three major factors in our approach to taskforce
training that provide the necessary realism and the ability to make
unbiased observations of the training situations. The first factor is the
modeling and simulation of systems and factors that have a decisive effect
on the training situation and that contribute in creating a realistic
training environment. The second factor is the data collection that
supports unbiased recording of the activities of the trained taskforce
when solving a relevant task. The data are received both from technical
systems and from reports based on manual observations. The third factor is
the visualization of compiled exercise data that provides participants and
others with a coherent view of the conducted exercise. The main
contribution of the research presented is the systematic description of
the combination of a training methodology and a system architecture for an
instrumentation system for computer-supported taskforce training. The
description characterizes the properties and features of our
computer-supported taskforce-training approach, applied in two domains.
Dr. Johan Jenvald is an active Naval Officer
and also a member of the MIND system group at the Defence Research
Establishment (FOA), where he is in charge of methods and tools for
computer-supported training. His research interests include team and
taskforce training, simulation, data collection and visualization. Dr.
Jenvald holds a PhD in Computer Science and a MSc in Computer Science and
Technology from Linköping University. Dr Jenvald is a fellow of the Royal
Swedish Society of Naval Sciences and of the Defence Research Science
Society.
|
