WASHINGTON: If you’ve ever daydreamed of designing your own tank — okay, “infantry fighting vehicle” — then DARPA wants to give you your shot.

The Defense Advanced Research Projects Agency has a long history of long shots, including such high-risk, high-reward projects as the first stealth aircraft and the earliest version of what became the Internet. DARPA also has its share of flops, like a 1960s sensor system called Project Agile that was supposed to locate the elusive Viet Cong (it couldn’t).

This week, under an initiative called FANG (Fast Adaptable Next-Generation Vehicle), the agency opened registration at an online “Vehicle Forge” for a nation-wide competition using a free suite of software tools. The challenge: design an amphibious, armored troop carrier for the Marine Corps — replacing the Expeditionary Fighting Vehicle program, which went through the traditional procurement process only to run badly over budget and be cancelled in 2009 — and to build a working model in three years, a fraction of the normal decade-plus development process.

The winner of the first phase, to design the new vehicle’s engine, suspension, and other motive systems, will get a $1 million prize. Should a team win all three phases — drivetrain, chassis, and the whole vehicle — it would take home $4 million. Ultimately, if DARPA’s winner proves superior to the results of a traditional Pentagon competition among major defense firms, it could become the Marine Corps’ new Amphibious Combat Vehicle.

There is, of course, a host of caveats.

To start with, the actual competition hasn’t started, just registration. The online design tools “are undergoing beta testing as we speak,” said Army Lt. Col. Nathan Wiedenman, the DARPA program manager, and should be ready for the scheduled launch in January. By then, Wiedenman told reporters this afternoon, DARPA will have populated its online library with “thousands of unique parts,” each a detailed computer model of components available in the real world, that users can choose from and connect to make their design. Ultimately, he said, for the third phase in which teams will compete to design the entire vehicle, “it’s on the order of tens and thousands of unique parts.”

It’s a lot like Legos or Tinkertoys, only vastly more complicated. While any U.S. citizen can register, Wiedenman said that in practice using the software competently will require “engineering skills and training.” Past DARPA competitions have mainly attracted teams from graduate schools and research firms.

The difference with FANG, Wiedenman argued, is that competitors won’t have to build a physical prototype, which should open the field to a much wider array of contestants. The traditional process of designing, prototyping, and refining military hardware is so expensive that only a small number of large companies can compete, restricting the pool of engineering talent in any given area to “a few hundred brains,” Wiedenman said. “In a nation of 300 million, we can do better.”

Currently, everything in the system for the first phase of the contest, to design the drivetrain, will be unclassified. The second phase, to design the chassis, may well touch on classified requirements for armor and shock protection, and DARPA hasn’t quite decided how to handle that. “We’re looking at the right options to pursue that now,” Wiedenman said, from simply requiring every team to have security clearance — which would undermine the goal of widening the talent pool beyond defense firms — to letting teams design “blind” without telling them all the details of the requirements they’re trying to meet. The various entries will be assessed automatically, not by a human, but by software rating them on a “multi-attribute utility function” based on Marine Corps input.

DARPA’s design approach takes inspiration from the manufacturers of computer chips, who now rely heavily on computer modeling and simulation to design new integrated circuits and make sure all the components work together before they physically put anything together, a concept called “correct by construction.” Said Wiedenman, “the first one we build is not a prototype, it’s a production model.”

Of course, a military vehicle is much larger and more complex than the most sophisticated integrated circuit, and whether it works depends on countless interactions between components that can interfere with each other by overheating, vibrating, or generating electromagnetic interference. Simulating all those interactions is effectively impossible. “That would take decades of computational work,” Wiedenman said. Instead, “we’re looking at ways we can intelligently deal with higher levels of abstraction… I’m not looking at every single bolt.”

The potential for unforeseen problems to pop up is why major defense programs build and test physical prototypes in the first place; historically attempts to short-cut that process and rely on simulations have not gone well, as both the Pentagon’s top acquisition official and the former director of the Pentagon’s largest program, the troubled F-35 Joint Strike Fighter, have admitted.

DARPA willingly admits its design contest may not lead to a mass-produced combat vehicle. The competition aims at meeting the Marine Corps’s requirements, but it is doing so in parallel to the traditional acquisition process: After the cancellation of the Expeditionary Fighting Vehicle program, the Marines went back to defense industry with orders to build a more affordable, less technologically ambitious war machine, which is being called the Amphibious Combat Vehicle. So while DARPA does plan to build a physical vehicle from the winning design, sometime in 2015, the Marine Corps may or may not decide to buy it.

What DARPA is sure of, though, is that its approach will inspire a lot of new ideas. Traditional development projects at big defense firms come up with one specific design, build prototypes, analyze the tests, and then go back to the drawing board to modify the original design. DARPA’s contest will let lots of people come up with lots of designs, and while most of them may well fail ludicrously, at least a few will come up with concepts no one ever considered before. Historically, “we’re only exploring a small portion of the design trade space,” Wiedenman said. “We are fundamentally changing how you approach system design by allowing people to populate that design trade space much more broadly.”

In essence, DARPA wants to let a hundred flowers bloom. Then it’ll be up to the Marine Corps to pick one — or not.