Who are we?
Physical Science Lab
Mission: We will partner with our customers and the University community
to establish leadership in key 21st Century technologies, building on current
expertise to transform our workforce and our marketplace by development
of strategic initiatives through major business segments and potential
spin-off opportunities in specialized products (telemetry) and services
(Balloons, RLV, EM), joint commercial enterprises (microelectronic, UAV),
and centers of excellence (Information Operations and EW/electromagnetic
analysis).
Information Sciences Team
The mission of the Information Sciences Team is to enable U.S. global information
superiority through the unique academia partnership to conceive the IO
workforce of the 21st Century.
Agent Modeling Group
Mission: Support the Information Sciences Team in provided software tools
and modeling expertise to allow rapid exploration of different domain spaces
using agent modeling techniques.
What are we doing?
We are developing a distributed, agent-based simulation and modeling system
which will be totally configuarable through XML programming and flexible
enough to apply to many domains.
A few of the questions we are currently trying to answer:
-
What does it mean to model decision structures?
-
Can we give reasonable decision/command and control analysis in a simulation
to give recomendations concerning tactics and doctrine?
-
Given certain machine/personnel/knowledge, what vulnerablilities exist?
-
How can we take advantage of the competition's vulnerabilities to maximize
on our goals?
-
Can we dynamically find patterns in the Command, Control, and Communications
structure of the competition and use these patterns to make predictions
about their actions?
-
Can we recreate these patterns in the competition's communications structure
to try to force them into a state we want them to be in?
How are we doing it?
We are currently in the middle of developing a prototype agent-modeling
system in a military domain. The demonstration of this prototype is scheduled
for March, 2002. We will use this prototype to explore the domain and what
will be necessary to support in the long run. The end results of our testing
and experimenting with the prototype will be a requirements document which
we will use to design and develop the "final" system for our customer.
Additionally, we will use this as a partial requirements gathering tools
for our future goals in building a more generic agent- modeling toolkit/system
which can be easily applied to multiple domains.
What will this talk focus on?
I thought you would never ask. This talk will be focused on what we currently
think will be necessary to answer some of the above stated questions relating
to decision structures and algorithms. I will talk about some generic ideas
on agent-based decision structures and how we plan on applying them. I
hope to lay out our current prototype concept and possibly get feedback
on other research areas we may not have looked at yet or different ways
of doing things that may help us avoid mistakes early. Different ideas
we've looked at range from BDI models to reactive models and to applying
uncertainty concepts and genetic algorithms.
Author: Jason Robey, Project Facilitator/Domain Expert