Agency Seeks Urban Automatons
Newest multimillion-dollar contest challenges inventors to tackle traffic.
Autonomous ground vehicles were placed between concrete barriers at the Grand Challenge 2005 starting gate so that human-robot interaction could be controlled. In the upcoming Urban Challenge, however, the Defense Advanced Research Projects Agency will be looking for autonomous vehicles that can handle this interaction on their own.
Daring the world’s robot builders and visionaries to design autonomous ground vehicles that could traverse the treacherous terrain of the
After awarding the Grand Challenge 2005 $2 million prize to Stanford University’s Volkswagen Touareg “Stanley” as the first autonomous vehicle to complete a 131.2-mile rugged but desolate course, the Defense Advanced Research Projects Agency (DARPA), Arlington, Virginia, says it’s time to raise the stakes in more ways than one. First, the Urban Challenge, scheduled to take place in November 2007, invites inventors to build ground vehicles that can autonomously obey traffic laws while merging into moving traffic, navigating traffic circles, negotiating busy intersections and avoiding obstacles. In addition, DARPA is sweetening the participation pot this time around by increasing the number of prizes and offering teams the opportunity to win awards without even revving an engine.
Incorporating lessons learned from the challenge events that took place in March 2004 and October 2005, DARPA is offering teams two ways to participate in the upcoming event. First, teams can submit a detailed proposal to a Broad Agency Announcement. Compliant proposals judged as best in technical approach, management and funding planning, and team strength could receive up to $1 million in technology development funds.
Teams interested in participating in the event but that either do not submit a proposal or are not selected to receive development funds in the proposal-only track may compete in the second track of the Urban Challenge. Track two will follow the previous Grand Challenge events’ format: Teams first submit applications then participate in a series of qualification activities culminating in the National Qualification Event (NQE). Each team chosen to participate as a semi-finalist in the NQE will be awarded $50,000. In addition, successful NQE teams will receive $100,000 and compete in the on-the-road portion of the Urban Challenge.
Although the final race will be nearly identical in format to the first two Grand Challenge events, the 2007 competition features one big difference: Urban Challenge is not a one-winner-takes-all proposition. Perhaps the biggest change—and most tantalizing incentive to participate in next year’s event—is that not one but three cash prizes will be awarded. Teams from the proposal-only track that are not awarded technology development funds as well as those that compete and qualify in the NQE will be vying for a $2 million first place award, a second place $500,000 award and a $250,000 third place award.
But before teams interested in participating in the event decide to begin writing proposals or to put engines in chassis, they need to understand just how arduous the obstacles of this challenge will be. The 60-mile course must be completed in no more than six hours and will test an autonomous vehicle’s ability to travel through busy intersections and follow or pass moving vehicles safely. In addition, a vehicle will need to pass a stopped vehicle and make a U-turn safely as well as drive an alternate route when the primary route is blocked.
Norm Whitaker, Urban Challenge program manager, DARPA, explains that accomplishing these feats will involve overcoming a number of technical hurdles that to date scientists and engineers have not been able to solve. However, the Urban Challenge dovetails nicely on technologies—particularly obstacle-avoidance capabilities—that were demonstrated in the first two events. The significant twist is the upcoming event’s emphasis on safety issues, he says.
“There are a number of military missions that are out in the middle of nowhere, and robots are good for those missions. But sometimes at the beginning, at the end or in the middle of a mission, they’re going to be interacting with people and that’s a critical moment. We need for them to be able to do that safely. We don’t want to have to put concrete barriers every place that a robot goes to keep the people and the robot separated. The robot should be smart enough to know that it’s not going to run over someone’s foot just because they’re standing next to it. So we’re building toward that happy relationship between humans and robots. We’re on that path,” Whitaker says. In the long term, perhaps robots can be “taught” to respond to voice commands, so humans can tell them to stop and they will obey, he adds, noting that the capability is most likely years away.
Dividing the overall technology challenge into two categories quickly reveals that while developing robots that operate in wide open spaces is relatively easy to do, solving the robot-human interaction problem is not as simple, he says. In the former scenario, for example, autonomous vehicles could be easily designed to plow snow off airport runways because airfields are fairly controlled environments.
But the volatile urban environment introduces a huge new set of safety issues that must be addressed, and the solutions are quite perplexing because they call for unraveling the intricate mysteries of intimate communications. “A large part of driving your car is communicating with other vehicles either by positioning your car or using overt hand signals. There’s a human-to-human interaction going on that the vehicle does not know how to do and doesn’t even know how to begin to do, so all those cues will have to be learned. Dealing with humans on the humans’ terms is part of what the Urban Challenge program’s all about,” Whitaker explains.
Development teams competing in the Urban Challenge also will have to find ways to teach autonomous vehicles essentially to make decisions on the fly. In barren, uninhabited environments, a robot that comes upon an unexpected obstacle has the luxury of time when deciding how to proceed. But Whitaker points out that this is not the case in a dynamic locale where the robot must interactwith people and its actions are driven by events. “The complexity grows tremendously,” he states. To his knowledge, this capability has not yet been demonstrated, he adds, and may not be achievable for many years.
Whitaker believes that designing, developing, building and demonstrating these extraordinary capabilities require collaboration between dreamers and hard-nosed technology problem solvers. If the amount of enthusiasm the event already has drawn is any indication, visionaries and engineers are up to the challenge. “We’re getting interest from all around the world. Some of the best teams in the
DARPA Urban Challenge: www.darpa.mil/grandchallenge