Designing a car to pilot itself across the desert from Barstow, CA to Las Vegas is hard enough. Now try doing it with only two wheels. Or barely four.
Among the twenty-five teams competing in the DARPA Grand Challenge autonomous vehicle race next week is "The Blue Team," a collection of U.C. Berkeley engineering students entering a motorcycle, and Team LoGHIQ, a self-funded family team with graduates from Rensselaer Polytechnic Institute.
The Grand Challenge is the brainstorm of the Defense Advanced Research Projects Agency, the R&D arm of the Department of Defense. After passing a preliminary test run next week, the vehicles will be asked to navigate roughly 250 miles of untracked desert by themselves. The human members of the team will be permitted to watch, but not control, the cars, following a limited time to program geographic data into the onboard computers.
The prize? One million dollars, making the race a cross between The Cannonball Run and a robotic version of Survivor. DARPA hopes to use the fruits of the public-private partnership to improve its small cadre of electronic soldiers, such as the Predator drone.
DARPA, meanwhile, has remained relatively silent on the obstacles teams are expected to face. Teams will not only have to pass waypoints, but also travel in predefined paths of varying width, across a variety of terrain. During the majority of the trip, the vehicles will be driving on unpaved roads and trails. However, vehicles will also encounter brush and dry washes, as well as sandy and hard-packed ground. Navigating water is a possibility, and the vehicles will likely have to travel through at least one underpass, potentially blocking GPS signals.
Team members say they're doubtful that any vehicle will navigate the course in the ten-hour time limit. "I would be very surprised if anyone wins it this year," said Seth Cabe, the leader of Team LoGHIQ. "It's tough enough to do with a human driver, but 250 miles without a human…"
But both "The Blue Team" and Team LoGHIQ will be competing against teams funded with sizeable budgets and corporate backing; "The Red Team" from Carnegie Mellon, for example, strapped a full-fledged Intel Itanium server to the back of a Humvee and added multiple cameras, radar, and laser range-finding sensors to make the job of navigating easier. Team Digital Auto Drive uses experience gathered from the TV show "Robot Wars" (with help from Toyota), while Axion Racing has signed up a pair of spokesmodel twins to entice its vehicle over the desert floor.
Instead, ExtremeTech sat down with two true do-it-yourselfers: LoGHIQ's Cabe, as well as Anthony Levandowski, team leader of the "The Blue Team" and its unmanned "Ghost Rider" motorcycle.
Team LoGHIQ already has the distinction of being the team whose car has traveled the farthest just to get to the start. Based in New York, the six-member team boasts Cabe, a mechanical engineer, as well as his father, brother, uncle, and two friends. Together, the team includes a physicist, two computer systems engineers, and a physics teacher.
"I had a recently graduated college with a degree in mechanical engineering, and I was looking for a job," Cabe said. "I came across the Grand Challenge web site, and I said to myself, I wish there was some way I could do this, but I don't even have a job."
Soon after, Cabe landed a job at Bernstein Display, a small forms manufacturer, and started roping friends and family into his new project. Cabe's father, who had worked on solar cars, was brought in. Still, Cabe's new project needed funds, and big sponsors weren't biting. So Cabe went hunting on eBay, where he began finding used parts at a fraction of their retail cost. All told, LoGHIQ had less than $50,000 to spend.
"We had to sort of design around the budget," Cabe said. "I think we came up with a vehicle that works, given the money. This is what you're supposed to do, start thinking out of the box."
Team LoGHIQ didn't even adapt an existing vehicle. Instead, the team mounted a 5-kW electrical generator that charges a 96-volt battery atop a bare-bones chassis. The battery, in turn, powers an electrical motor capable of 8 electrical horsepower, the rough equivalent of a 30-hp gas-powered engine. Cabe said the 1,000-pound car has a top speed of about 20 to 25 miles per hour, although speed will not be the driving factor, so to speak. The suspension is cobbled together from the RockShox coils typically found on mountain bikes.
Navigation, on the other hand, is what's at stake, and the LoGHIQ team needed a computing platform. Cabe said he originally posted to a forum operated by Via Technologies Inc., the Taiwan microprocessors and chipset manufacturer. Timothy Brown, who runs VIA's nascent robotics program, joined the team. "I guess they liked us being the little guy using the open platform…hobbyists, I guess," Cabe said.
Via donated one of its EPIA mini-ITX motherboards and a 1-GHz C3 processor. The chip powers Slackware Linux, which the vehicle uses to track its internal progress using open-sourced GPS code the team hunted down on the Internet. A combination of a simple laser rangefinder, the GPS, and some "cheap" stereoscopic cameras completed the vehicle's sensory apparatus.
"I like to think of our vehicle as a cockroach scuttling across the floor," Cabe said. "It goes pretty quickly; it's not that intelligent; it doesn't see everything in great detail, and it doesn't see anything directly in front of it."
The LoGHIQ vehicle uses a system of dead reckoning to find its way around; when the vehicle approaches an obstacle, the onboard digital compass picks a direction to move around the obstacle.
The vehicle has one advantage, however; each wheel is driven so the vehicle can spin in place without needing to back up. While it may not seem like a big deal, the design means that LoGHIQ doesn't need to buy an extra camera.
Not surprisingly, Cabe said he expects the weak point in his first autonomous racer to be the software. "There's glitches in this kind of stuff," he said. "Barbed wire fences, brush.
"It's hard to tell what DARPA will do," Cabe added. "They could make us start on salt flats with no obstacles for 50 miles." On the other hand, they could also start the card within 50 feet of a ravine, or up against scrub and brush that is difficult to detect.
In all, though, Cabe said he looks forward to the race. "We've accomplished an amazing amount in one year, more than in the last 20 years," he said.
Let's get one thing straight: "The Blue Team" isn't out to win the Grand Challenge. Their motorcycle – Levandowski won't name the manufacturer, since none of the bike manufacturers the team approached were willing to take the risk – isn't programmed to go fast enough to complete the 250-mile race in the allotted time.
But Levandowski says their goal isn't to win the $1 million bounty, but to chase the ultimate prize: a DARPA contract, and eventual employment. "The point of our project is that the four of us had an idea who are graduating into the research and manufacturing to design something for the Army," he said. "This is a good way for us to get noticed and get hands-on experience. We're having a blast."
In January of last year, Levandowski's mother, who works for the European Union in Belgium, sent Anthony an email announcement of the race. "It sparked my interest," he said. "I thought, wow, this is so much more sophisticated an endeavor than launching a rocket to win the XPrize. It takes much more ambition, more hardware, planning -- a massive amount of thought."
For Levandowski, a post-graduate student at U.C. Berkeley, the epiphany came when the team drove back from an engineering conference in L.A. along the endless straight stretches of California's Highway 5. Previously, the team had considered a tracked vehicle, such as a bulldozer, for its DARPA entry.
Suddenly, a pack of motorcycles "descends upon us," Levandowski said. "To the left of us, to the right of us. Look at the agility of these motorcycles, I said. This is what the future will be like. We thought about it, asked ourselves, is it feasible? Well no, we said, probably not."
Obviously, the biggest problem for a motorcycle is simply keeping it upright, achieved through a combination of speed and balance on the part of the rider. Increasing the motorcycle's speed, however, also increases the risk of a crash, especially when the "eyes" are mechanical, not human.
