Communications of the ACM
Preparing For the DARPA Network Challenge
Ethan Trewhitt, co-leader of GTRI's I Spy A Red Balloon team
In late October of 2009, DARPA announced its Network Challenge, a competition designed to see how online social networks could perform a widespread task. The challenge: to find ten (10) moored red weather balloons spread throughout the continental United States for up to nine hours on Saturday, December 5th.
When my Georgia Tech Research Institute (GTRI) colleague, Erica Briscoe, introduced me to the competition, we chatted idly about competing and quickly recognized the low barrier to entry. Whereas DARPA's previous competitions, such as the Grand Challenge, required years of work and millions in funding, this task required very little investment beyond the voluntary time of the participants. Teams quickly realized that aerial photography, radar, satellite imagery and the like were too costly considering the $40,000 prize, and that crowdsourcing would be the most accessible approach—exactly the conclusion DARPA intended.
Many early teams decided to capitalize on the age-old concept of greed by paying balloon spotters for their reports. However, it seemed unfair that many devoted teammates, who may not run across a balloon on launch day, would be uncompensated for their loyalty and effort. For that reason the GTRI team, known as "I Spy A Red Balloon," earmarked any potential winnings for charity, specifically the American Red Cross. This would also reduce the tax and paperwork headaches of prize money distribution; as I understand it, MIT's winning team is still mired in the red tape of distributing the winnings!
In mid-November we attempted to court a major shipping company in order to have access to the eyes of its drivers, since a huge existing network of trucks traversing most of the US would likely find the balloons in no time. Who wouldn't want to advertise that "if you needed access to every part of the US at once, you'd come to us"? Instead, they declined over "concerns about driver safety and focus on the job." Indeed, it's probably a bad idea to have a bunch of shipping truck drivers texting us balloon positions while driving. A few other corporate opportunities came and went; one month was simply not enough time for most major corporations to make a measured decision about such a contest.
Our Web site, ispyaredballoon.com, went live about three weeks before launch day. In addition, we registered a Google Voice number, (262) I-SPY-SPY, for phone calls and text messages if and when people stumbled upon a balloon during their normal Saturday activities. Our Facebook group provided the best interactive forum for team members, with more than 800 participants prior to December 5th.
Launch day began calmly, even as we passed the 10:00 am launch time. A total of seven GTRI employees volunteered to operate our "war room" in a lab at GTRI, with donuts and bagels for breakfast. A projector showed the internal Google Maps interface that indicated the balloon positions reported on our Web site, with different colors to indicate balloon numbers and an IP address log for forensics. We browsed through our Facebook group and the pages of competing groups, set up Twitter searches for queries like "DARPA" and "balloon," forwarded our Google Voice number to a phone in the lab, and waited. By 10:30 am, reports began appearing on our Web site and Twitter began to show serious traffic.
Coming up in part 2 of this three-part series, on Monday: handling the incoming data on launch day.
No entries found