|Home > News & Events > MITRE Publications > The MITRE Digest >|
Project Albert Reaps Results with Data Farming
The problem with planning major battles is that battles seldom follow plans. Human beings in particular cause unpredictable effects because combatants on both sides often react differently than expected. This means in the age of digital communications and asymmetric threats, victory often goes to the agile—not necessarily the best prepared.
This need for agility—to zig when the enemy zags—underlies the U.S. Marine Corps' Project Albert. The program, named for Albert Einstein, is the Marines' way of rethinking traditional war-gaming strategies by using high-powered computers and a unique form of modeling and simulation. The project's motto is "Understand the unexpected."
MITRE's Gary Horne is director of Project Albert and creator of "Data Farming," the concept that lies at the heart of the project, which gained congressional funding in the late 1990s. Although largely a partnership between the Marine Warfighting Laboratory in Quantico, Virginia, and the Maui High Performance Computing Center (MHPCC) in Hawaii, Project Albert also draws participants from the U.S. joint services, several international military organizations, national laboratories, and corporations. MITRE joined the effort in 1999. Horne—who works on Project Albert from our Quantico site, along with a handful of other MITRE employees—says our role is not large in staff years, but is exceptionally important: "We provide technical guidance and oversight for the many organizations that interact with us. We act as the technical glue for Project Albert."
Data farming centers around a type of computer simulation called a "distillation." Just as a fine glass of Scotch results from careful filtering until only the refined liquor remains, distillations in data farming are models honed to their key components—the very essence of the questions you want answered. Compared to other forms of high performance computer-based experimentation, data farming is novel because it uses fewer variables than many other computer models. Albert Einstein himself warned against trying to juggle too many variables at once, saying "scientific method in most cases fails" because the super-abundance of detail makes it difficult to judge cause and effect.
Growing Answers from Key Questions
Horne explains that traditional methods of experimentation and simulation often fall short for this very reason. "Over the past 60 years, those of us in the field of operations research have put more and more details into our computer models as we've gotten more computing power," he says. "There's nothing wrong with that—but we still don't have the answers to many of our important questions. That's because military operations involve human beings, and there are so many different possibilities involving people that don't happen with physical systems. Granted, there are physical systems—tanks, weaponry, radios—involved, but all the Project Albert questions also involve the actions of human beings."
"By trying to capture information simply, we're addressing important military questions current simulation technologies can't meet," adds MITRE's Ted Meyer, Project Albert's information architect and visualization expert. "We use our access to high-performance computers and run the models many times—thousands of times. We've also developed visualization tools that let our sponsors see the results of the data farming, so they can start formulating answers to their questions."
To the Project Albert team, it's all about the questions. In this case, the questions the Marines (and other military groups) ask are very much products of the post-Cold War era: How do we protect our nation from martyr-based attacks? What kind of force mix should we use when operating along a coastline? Many of these types of questions are open-ended, lending themselves to a variety of possible answers.
"For example, a Marine commander may have 60 things he thinks he wants us to study, but with the Data Farming Environment we can narrow the variables down to the critical five or six," Horne says.
Project Albert focuses on three main elements of modern warfare—ones generally conceded to be among the hardest to account for in traditional battle modeling and simulation:
Intangibles. These include such human factors as leadership ability, troop morale, and overall trust within the organization.
Co-evolution. This involves changing your actions based on what you know about your opponent and what you think your opponent knows about you.
Thanks to the power of the MHPCC Data Farming Environment, the Project Albert staff can run models in continuous loops, incorporating both the big-picture background and small—but often crucial—details. After much iteration, the data points can be "harvested" for results, using the visualization tools developed by the team.
Surprises Yield the Seeds of Ideas
"Data farming gives us two things," says Horne. "First, it gives us an idea of the huge landscape of possibilities. Second, it gives us the outliers—the surprises. I love surprises! Sometimes, for instance, you run a computerized experiment 100 times, slightly changing the variables, and the outcome is always the same—except once. When that happens, you want to analyze what was different. How did that variable make the outcome change?"
Horne and Meyer cite an example of Project Albert's real-world application, a USMC experiment called "Gunslinger," which modeled the effects of having an escort armed with tracer bullets accompany a truck convoy. At the MHPCC Data Farming Environment, the Project Albert team ran a simulation with slight variations thousands of times. They discovered that using tracer bullets to draw enemy fire enabled the convoy to protect itself better against possible ambush because they proactively sought out the locations of the opposing forces and could focus their return fire accordingly.
"This finding would have been discovered in the field eventually but our experiments allowed the Marines to learn it early," Meyer says.
He explains further that outliers can lead users to the right answer—in a computer-generated environment—providing pathways that might lead to success in the real world. "But you need to follow up on the analysis," he says. "This work points the way, but it can't necessarily come up with the final answers. Yet if we hadn't learned a possible solution exists, that you can win at least in one scenario, then we wouldn't even know to start looking for an answer. And believe me, when it comes to Marines, if they think there's even a possibility of success, their attitude is: 'We will make it happen.'"
Making Code Available
Meyer notes that as interest continues to grow in Project Albert, so do the opportunities to increase its potential. "We are 'open sourcing' [making the software code publicly available] all the Project Albert suite of capabilities, within legal limitations. The Marine Corps' intent has always been to make the software tools and specific modeling techniques available to others who want to use them. By expanding into a much greater development community, we can increase Project Albert's use and effectiveness." The main stipulation for using the Project Albert capabilities is that users share their data, which many do at bi-annual Project Albert international workshops. The most recent, for instance, took place in Wellington, New Zealand, and attracted dozens of participants from eight nations.
Proof of Project Albert's growing popularity can also be found within the pages of each edition of Maneuver Warfare Science, a yearly compilation of reports from around the world detailing recent uses of the Data Farming Environment. (The name stems from the fact that the Marines call their agile, decisive warfighters "maneuver warriors.") Additionally, a popular business book from 2003, It's Alive, devotes an entire chapter to the program.
"People understand data farming might be able to help them," Horne says. "The Project Albert team doesn't claim to have all the answers to the military's toughest questions, but we do promise to use our capabilities to explore them. Above all, we are passionate about answering those questions."
—by Alison Stern Dunyak
Articles and News
Page last updated: February 3, 2005 | Top of page
Solutions That Make a Difference.®