Small Wars Journal

DARPA uses the crowd to solve problems

Tue, 04/12/2011 - 4:42pm
At the risk of seriously degrading office productivity everywhere, I bring to your attention the Defense Advanced Research Project Agency's (DARPA) latest request for help. DARPA wants you to help it develop better anti-submarine warfare tactics and it has designed a sub-hunting computer game it wants you to play. DARPA is "crowd-sourcing" the solution to finding and tracking adversary submarines. You play the sub-hunt game and report your tactics back to DARPA. DARPA will then incorporate the collective knowledge it thus receives into the software it is developing for a new sub-tracking robot. Click here to download the game -- and don't tell your boss I sent you.

Sub-hunting is not the only task for which DARPA has turned to crowd-sourcing. In February, DARPA crowd-sourced the design of a new tactical vehicle that would perform either reconnaissance or battlefield delivery and evacuation missions. DARPA hoped to attract the interest of service members, auto enthusiasts, designers, and engineers and offered a $10,000 reward for the top design.

The theory behind these crowd-sourcing experiments is that DARPA will achieve practical designs and concepts faster by engaging a large community, even if most of the participants are non-experts. The traditional method is to engage only experts, who then use experimentation and iteration to refine concepts. Crowd-sourcing aims to perform the required iterations at the beginning and in "parallel" rather than slowly and sequentially. Efficiently processing the initial flood of trials into usable results requires software that can sort through the input, the vast majority of which is presumably of low quality. This sorting capability was likely the barrier to efficient crowd-sourcing in the past. But if anyone would have that capability now, it would be DARPA.

Should DARPA's crowd-sourcing experiments with hardware design and tactics succeed, we should expect the crowd-sourcing technique to spread to a much wider variety of military challenges. Soon DARPA, TRADOC, MCCDC, and others may call on us to be amateur tinkerers and game-players, as these groups task the crowd to solve problems. The bosses may even have to pay us to play.

Comments

ML (not verified)

Tue, 04/12/2011 - 6:28pm

@CBD:
"An option to speed up the passage of in-game time would greatly improve the experience..."

:facepalm:

There is a time scale option at the bottom right.

CBD (not verified)

Tue, 04/12/2011 - 6:04pm

Unfortunately, the game is rather dull. It is clear that, in order to generate sufficient data for an ideal form of the ACTUV, DARPA has sacrificed player usability (Particularly with missions that require 15 or 30 minutes (in real time) for certain tasks, leading to total level times in excess of an hour for even the basic levels...and most of that involves the relatively simple task of finding a sub by one of several methods and then watching for changes in speed and direction every couple of minutes.

An option to speed up the passage of in-game time would greatly improve the experience...

The user interface gets bonus points and it's fun to use the tech to figure out how these vehicles would work...but a computer will still do this type of sim (sweep and locating tactics, reacquisition methods, sense-and-evade, etc) better than the crowd...though it is certainly tempting to have the Web produce the solutions rather than buying time on a supercomputer and a team of specialist software engineers to run the thing. But then again, making this game took software engineers and some small chunk of change.