|
Important Notes!
AFCEA's Signal Magazine interviews Dr. Gary Horne about Project Albert and Data Farming for their June 2005 issue.
The Marine Corps Warfighting Laboratory's
Project Albert is the research and development effort whose
goal is to develop the process and capabilities of
Data Farming, a method to address
decision-maker's questions that applies high performance
computing to modeling in order to examine and understand
the landscape of potential simulated outcomes, enhance
intuition, find surprises and outliers, and identify
potential options. Data Farming is the method by which
potentially millions of data points are explored and
captured. It could be considered akin to Data Mining
combined with feedback which allows for the more
intelligent collection of more data points. This process is
made possible, in part, by the exploitation of High
Performance Computing assets and methods, and the project
is fully supported by the Maui High Performance Computing
Center. The Project Albert modeling approach is achieved
through the development of a suite of models, called
Distillations, to drive home the point
that these models are produced as an intentional complement
to the very highly-detailed simulations being developed
within the DoD, which by the very fact that they are so
highly-detailed and encumbered, do not permit the
examination of a very wide range of possibilities and
outcomes.
By virtue of their being much easier to run and understand
(think: SimCity adapted to a combat situation), they are
proving to be effective tools that help capture and
scientifically reproduce the ideas of Subject Matter
Experts, such as those thinking about tomorrow's concepts,
doctrine, and requirements. This suite of entity-based
models allow for rapid and highly tailorable changes in
entity characteristics and behaviors, quite amenable to,
and intentionally designed for rapid, repeatable concept
exploration. Project Albert develops a suite, vice a single
model, to allow for the testing of robustness of
observations across modeling platforms, and because each
model has inherent strengths and unique capabilities with
regard to each aspect of modeling how entities think,
decide, shoot, move, and communicate. The Project Albert
suite of models
includes Map-Aware Non-Uniform Automata (MANA),
Socrates, and
Pythagoras. It is of note that MANA is developed by the
New Zealand Defense Technology Agency, and it is used free
of charge by members of the Project Albert Team. Project
Albert is also harnessing existing, and developing new,
methodologies for investigating the results of running such
models. In this regard, a wide range of Data
Exploration and Data Visualization tools and
methods are being employed.
Also being developed is the concept and
initial implementation of Automated Red Teaming, a
methodology that helps identify how Red can beat Blue
through intelligent, effective, and efficient search
through the space of possibilities. This is accomplished in
part through the employment of a variety of evolutionary
algorithms.
Currently, Project Albert is collaborating with
military decision-makers, both inside the USMC, jointly,
and within coalitional arrangements, to apply
these techniques and tools to real-world questions. Some
examples of the work efforts can be found on the Project
Albert International Workshop pages. Another example of
collaboration efforts is Project Albert 's Internship and
Experimental History Program. This program has employed
interns from U.S. Naval Academy, Naval Postgraduate School,
U.S. Military Academy, and civilian academic institutions.
Experimental History describes the process of taking a
historical battle of interest and recreating it in one or
more of our distillations and applying the process of data
farming to it to yield a set of data that address
interesting what ifs. In a sense, it is akin to taking a
single data point (of history) and expanding the
understanding of it to a region of possibilities. Finally,
initial progress is also already being made on the
integration (translation of scenarios) between different
models and C2 Platforms, e.g. from C2PC to one of
the Project Albert models, and on the development of
collaborative environments that assist in the whiteboarding
and creation of scenarios and in the joint understanding of
the data that results from Data Farming the scenario.
AFCEA's Signal Magazine has done an article on Project Albert
and Data farming that can be found here.
Project Albert supports the DoD Command and Control Research
Program (CCRP).
This site is hosted in Virginia.
|
