THE GOAL: To build the world's first truly autonomous robotic land vehicle.
THE PLAYERS: Two dozen teams of robotics pioneers, garage tinkerers and high school eggheads.
THE TEST: A 210-mile off-road race from L.A. to Las Vegas.
William "Red" Whittaker -- doing his best George C. Scott doing his best George S. Patton -- strides before an oversize American flag and addresses his troops. "What we are looking at is a list of top 10 things that need to be done," he barks. "These things will be retired by Friday of next week." It's a November morning in Pittsburgh, this hangar that houses Carnegie Mellon University's Planetary Robotics Lab is drafty, and Whittaker's troops -- 30 CMU undergrads and graduate students -- are overextended and sleep-deprived. In the corner sits a decidedly earthbound military-grade Humvee stripped to its skeleton; the team hasn't even begun outfitting the thing with the sophisticated electronics that will, in theory, enable it, four months hence, to drive all by itself some 210 miles off-road from outside Los Angeles to Las Vegas -- and in so doing win the $1 million DARPA Grand Challenge. One of the Whittaker rank and file has the temerity to raise his hand. The vehicle's isolation shocks need more verification work, he says: "We can't risk damage to a primary component." Whittaker, an ex-marine turned robotics professor, pulls himself up to his full 6 feet 4. "Well," he says, "I reiterate: You've had weeks, you've had mock-ups, and I just decline. This is a question of technical ambition versus solid execution. It's time to kill technical ambition."
A couple thousand miles west, Anthony Levandowski has for a week been catching his nightly ration of sleep in a makeshift basement bedroom, just a few steps below the converted guestroom that serves as his workshop in his south Berkeley, California, home. The 23-year-old University of California, Berkeley, industrial engineering grad student -- and DARPA Grand Challenge aspirant -- has scheduled an important early field test for tomorrow morning. "My initial goal was to have the full remote control and some autonomous control done by last weekend," he says, "but that didn't happen, so now I'm in the basement. It helps keep me focused on the task." The task, as Levandowski has defined it for himself and for a handful of undergraduate helpers, is to build not just an autonomous off-road vehicle, something that has thus far eluded the best efforts of billion-dollar defense contractors -- but to build an autonomous off-road motorcycle. Nine months in, he is beginning to feel the weight of this audacious design decision. Tonight Levandowski permits himself no downtime in his bedroom of shame; he works straight through, struggling with kinks in the program that controls the steering, the crucial link in the software chain. His workshop door is open to the street so he
doesn't asphyxiate himself. "I don't care about the million bucks," Levandowski says. "It's more about realizing that I can actually do something that fairly intelligent people have told
me I can't. Right now, they're winning."
A few days later and 400 miles south, the kiddie corps is on the move. Sort of. "We're not stuck," says Tom Laymon, "but we are getting stuck." It's a crisp, sun-sharpened afternoon, and Laymon has tentatively nosed a stock Acura SUV off a path in California's Mojave Desert and into a patch of loose-packed sand, which has immediately begun to devour the truck's wheels and suspension. Laymon is not a high school student (he is, in fact, a vice president of American Honda) and thus he is not an actual team member of the Palos Verdes High School Road Warriors, the team that has emerged as the against-all-odds Bad News Bears entry in the DARPA Grand Challenge. Laymon is an "adult mentor" to the Road Warriors: one of the many parents who happen to be well connected in the defense, aeronautics or automotive worlds, and who are thus supporting -- and at times threatening to usurp -- their children's efforts. The SUV, though, is packed with legit Road Warriors, shouting advice, lobbing wisecracks, and otherwise enjoying this preliminary manned desert test-drive. "Hey," shouts 15-year-old Katrina deSimone from the far backseat. "I don't know how we're gonna use the car for the race if we destroy it here."
Over the past 20 years, the individual components needed to build an autonomous off-road vehicle have been fairly well worked out: the laser, radar and sonar sensor systems that are the vehicle's eyes; the GPS system that locates the vehicle in the world by triangulating a position from satellite beacons; the mapping systems that allow real-world visual and topographical information to be compressed into onboard computerized maps. But to date, no one has put them together well enough. ("Systems fusion," it's called.) Certainly not well enough for a vehicle to negotiate rugged terrain over long distances all by its lonesome.
When DARPA -- the Defense Advanced Research Projects Agency, the Defense Department's R&D wing -- began cogitating on an ambitious 2001 Congressional mandate to make one-third of the military's ground vehicles unmanned by the year 2015, it decided to throw open the field with the Grand Challenge. Let the academics and garage tinkerers knock themselves out trying to solve the puzzle of off-road robotics, and if terrific solutions emerge, they can be developed and executed by the big defense contractors -- which are, for the most part, hanging back from the race. "Our goal was to jump-start the technology," says Air Force Colonel Jose Negron, Grand Challenge program director, "by galvanizing the interest that was out there, especially among people who previously had little or no connection with the government or the big defense contracting companies." In other words, less Star Wars, more Road Warrior.
On the strength of their race-application technical papers, some 25 teams, drawn mostly from the ranks of universities and entrepreneurially minded engineers, had by December '03 made the final callback to nail the Mel Gibson part in DARPA's $1 million production. Savvy race handicappers might take a fancy to overbot powerhouses like Caltech, Team TerraMax (a collaboration between Ohio State University and Oshkosh Truck Corp.) and, with two vehicles entered in the race, SciAutonics, a private team from the L.A. area with a core of engineers from the Rockwell Scientific Company. Nice teams all. But a close look at three other qualifiers might better reveal the nature of the overall enterprise, the mix of motivations and aptitudes among a group of people who passionately want to do too much in too little time for not enough money.
The Carnegie Mellon Red Team, led by Whittaker, is the consensus race favorite, the robotics establishment team par excellence, grinding away at incremental improvements in the best existing technology in its methodical, quasi-military way. Levandowski's Blue Team holds down the opposite end of the spectrum: It's the brainchild of one self-appointed Big Thinker who's betting his young career on a bold idea -- single-track motorcycle locomotion -- and hoping a million bothersome technical details sort themselves out as best they can. Somewhere in the middle lie the Palos Verdes High School Road Warriors. The team has a split personality -- split between, on the one side,
a 16-year-old genius and his high-tech stage mother together bent on pursuing the visionary technical muse and, on the other, a cabal of competition-minded engineer parents who want to make damn sure the team turns in a decent performance in a race it never figured to be in.
Even if most of the factories in Pittsburgh shut down years ago, the western Pennsylvania landscape of rough-hewn valleys and sharp ravines still feels industrial, freshly excavated. Which is to say, it's Red Whittaker's kind of town: the blue-collar vibe, the aura of flinty Scottish workaholism that is Andrew Carnegie's legacy to the university he founded. "I enjoy that work ethic," Whittaker says. He is leaning forward in his office chair, taking exactly one hour off from his normal schedule to talk to me after the conclusion of the Wednesday-morning Red Team robot class. "Engineering isn't a stepping stone to med school or law school for the majority of the people you've met this morning. The technical life is their means to the grand future."
A man who names his Grand Challenge team after himself isn't averse to a little mythologizing, self- or otherwise. The Red stories are legion, faithfully reported by the Pittsburgh press -- like the time one of his climbing partners died in a fall on Mt. Chacraraju in the Peruvian Andes. When the local authorities declined to retrieve the body, Whittaker flew to Peru, climbed the mountain, and brought his friend down himself.
Clearly, this is someone who likes to see a job through. For the Grand Challenge, Whittaker has broken down the Red Team's preparation into three phases: formulation, development and field-testing. All along the plan has been to get a core team out to Barstow, California, where the race will begin, by January to run the vehicle continuously under race conditions until the inevitable bugs are worked out. What this means in practice is that the team is constantly bumping against internal deadlines, necessitating all-nighters in October and November.
By way of making the point that non-robotic life responsibilities should only fuel dedication to the cause, Whittaker offers up Chris Urmson, 27, just shy of his Ph.D. and the team's resident software genius. Urmson and his wife had the inopportune timing to have had their first child right before the Red Team went into overdrive. "Wow," I mutter, almost to myself. "What do you mean, 'Wow'?" Whittaker fairly roars. "There are people in the military right now who don't see their children for months. There's nothing heroic or special going on here." When I suggest that it might actually be easier to be stationed in another country than to have to come home every night at midnight or 3 a.m. and face your wife, who's single-parenting your newborn, Whittaker fixes me with a look and says, "Maybe you have the wrong wife." Wow. I'm only visiting CMU for a couple of days and already I'm being asked to consider a divorce for the good of the team. He elaborates. "You get right down to those questions, Am I really living the life I choose?" he says. "I am who I am when I do this work. Am I counting the days until I get my life back? No, it's not like that. If this is what you're born for, this is your life."
Whittaker has been one of the few hardy pioneers ("a godfather of the tradition," as he puts it) who have found truly useful niches for autonomous machines: his Dante robot that hunted for meteorites in Antarctica in 1993; the Pioneer robot that assessed structural damage at Chernobyl in 2000; the Groundhog robot that mapped a 3,500-foot mine shaft in Pennsylvania this past May. The race at hand has yet to be run and Whittaker is already thinking about the next series of technical battles that he can lead his Red Team into, what he calls the Grand Challenge afterlife -- smart cars, for instance, that could wrest control of the wheel from an incapacitated driver. "We race for the future of robotics," Whittaker says. "Competition can catapult a culture from the laboratory into the world."
The project began last April when Robertson collected a few of the tech-minded kids to talk about forming a ham radio club. The conversation drifted to GPS navigation, and one of the kids, Joe Bebel, then a 15-year-old freshman, brought up the Grand Challenge, early news of which had just reached his mother, Alice Parker, an electrical engineering professor at USC. Ham radio didn't stand a chance.
A Sunday-afternoon team meeting in a Spanish-colonial-style PV classroom is a disorienting experience. The average age of the somewhat distracted-looking kids milling around the blackboards and school tables is about 15. The parents in attendance have the concerned expressions of adults brought in to discuss their children's academic shortcomings. But the ever-hearty Mr. Robertson gives Bebel a friendly nod: "Hey, Joe, are you thinking about that 'vision problem'? It's never been done, you know." In this upside-down world, he might as well be clapping the back of the high school quarterback on whose varsity-jacketed shoulders rests the outcome of the big game. Bebel wrote much of the technical papers that transformed the Road Warriors from a cute idea into a runaway train -- and the rest
he described in conceptual terms to his
mother, who translated it into high-level programming-ese and ran it by the USC comp sci department for good measure. "But Joe is really the brains behind the team," Alice Parker says emphatically.
The thought occurs, and not for the first time: Is this any way to run the nation's defense?
Actually, the Road Warriors can't quite decide how to run themselves. Many
of the students -- and Alice Parker most
vociferously -- want the kids to march to their own autonomous drumbeat, even if race deadlines are missed in the process. "My goal is to keep this from being 'Dad's science project,'" Parker says. Other parents, mostly corporate engineers and managers, aren't averse to importing grown-up technology if that will enhance the team's chance of making it to the Big Show. "When the kids need to know things, the world's experts are brought in," Greg Larson, PV parent and top Boeing engineer, tells me. "Because they tend to be their neighbors or their parents."
Bebel, a likeable old soul with a flair for diplomacy, parses the issue this way: "Once we've got something built, then it will be easier to have an adult mentor take on a bigger role without jeopardizing the students. As long as we can still say the students did the hardest part, that would still be very impressive."
When Anthony Levandowski first heard about the Grand Challenge, he knew it was for him, even though he
hadn't a clue what sort of design he would bring to the table. Returning to Berkeley from the February '03 conference in L.A. where DARPA first officially announced the race, he had an epiphany. "I was driving out of the mountains," he recalls, "and this pack of motorcycles comes barreling at me and literally splits around me, as rushing water would split around a rock. Right then, I decided this would be the right way to do things."
An autonomous motorcycle? Levandowski had found a way to take the extremely difficult and make it nearly impossible.
In his methodical way, Levandowski can tick off the theoretical advantages of the motorcycle. It's light, tough, agile and at higher speeds very stable -- "the fastest way to get from Point A to Point B in a dense forest," he says. In the big off-road races like Paris to Dakar or Baja, motorcycles race four-wheel vehicles to a rough draw, and the only reason they don't run away with it is because of "rider fatigue," obviously not an issue in autonomous racing. Great. The problem is, as everybody knows, a motorcycle without a rider has a tendency to fall down.
There are solutions to this problem, and Levandowski is working hard at them, so hard in fact that every theoretical issue not related to basic locomotion -- perception, pathfinding, navigation -- is getting the most expedient quick-fix treatment possible. From a competition point of view, this approach is insane, which he readily admits. "Why two wheels?" he says. "At some level, I still don't have a good answer."
At least not an engineering answer. "I always wanted to have an off-road motorcycle as a kid," he says. But growing up in Brussels, Belgium, the son of a divorced French bureaucrat mother, wasn't really conducive to California-style off-road pursuits. Even when he moved to the San Francisco Bay Area to live with his American father and remake himself as an American teenager, he remained a studious, disciplined kid. The best explanation for Levandowski's magnificent two-wheeled obsession is that no one's ever gone this way before. "I think it's just my personality," he says. "I feel the need to be different, not by dying my hair blue but by coming up with more interesting ways of doing things. After I've shown that something can be done, then I'm no longer interested. That's bad, but I'm trying to be honest."
A lanky 6 feet 6, Levandowski projects an athletic grace (as a Berkeley undergrad, he was on both the crew and the volleyball teams) and a Continental self-assurance. He bought his tidy, pleasant house "with judicious planning," he says, in part with earnings from various Internet sideline ventures. He's structured his Grand Challenge Blue Team in a similar spirit of canny enterprise. Levandowski pays the bills, and he owns all the intellectual-property rights, even to work designed by his mostly young, undergraduate teammates. They labor, with varying degrees of diligence, not for course credit or money, but because Levandowski is a persuasive salesman and an autonomous motorcycle is a cool thing. "With no rewards, it is hard to apply pressure," he admits. "My biggest challenge is to come up with a rebuttal to, 'Oh, I had midterms last week.'"
The little cycle actually provides a great demonstration of the physics that Levandowski is up against. The bike doesn't want to stay upright, so when you give it a shot of juice from the remote, it tends to go into a wild wobble and then dramatically crash to one side or the other. Unless you know what you're doing. The trick, as any motorcyclist will tell you, is proper countersteering. If the bike is falling to the right,
you turn the front wheel to the right, more sharply than if you were making a right turn, and that provides the counterforce to bring the bike back to vertical. Turn too forcefully to the right, as I did when I was trying to drive the mini-motorcycle, and you'll overshoot vertical and wipe out on the left side or get locked into a desperate, doomed left-right oscillation that results in a similar slide across the ground.
Finally, though, I got the hang of it -- and sent the miniature bike flying out of sight over a ditch, a thrill straight out of Thelma & Louise.
Time is running out. There are, Red Whittaker reminds his Monday-morning Robotics 101 class (CMU's academic cover for the Red Team), only 109 days left to race day. Whittaker turns his attention to the student responsible for designing how the decals of the various corporate sponsors are going to fit on the car. That may sound trivial, but Whittaker calls it a "hot-button issue." Despite its reputation as the rich kid on the DARPA block, the Red Team's aggressive massage of corporations for money and equipment is the only way it can afford to expend, all told, the equivalent of more than $2 million to win a $1 million prize. "I'm working on it," the decal specialist replies, which even a first-time visitor can sense is not the right answer. "Could you not get 'working on it' but get it done?" Whittaker replies sharply. "It would be a tremendous 'done.'"
If only the other vehicle issues could be solved with a neat wrist slap. Of all the problems that Whittaker and his team face, perhaps the most intractable is one that has bedeviled artificial intelligence for years: how to get the damn thing to see. By the mid-'80s, roboticists at CMU and elsewhere had worked out the dialogue between a mounted video camera and some basic computer steering algorithms that allowed cars to be driven without drivers. Basic pattern recognition, being able to recognize and follow the straight line of a road, is something that computers tend to be good at. But having a clue about what's going on in the dynamic, three-dimensional and completely unpredictable off-road environment is something else again. Dodging a boulder in the middle of the path shouldn't be too hard. But recognizing whether the large object is in fact a boulder or a clump of tumbleweed or a moving Grand Challenge competitor will be. Ditto distinguishing the edge of a desert path from the rough country that lies next to it, or recognizing "negative spaces" like ditches or potholes.
The Red Team has chosen a range-finding laser, or LIDAR, to be the vehicle's primary "eye." The LIDAR shoots out a pulse of light, which hits something and bounces back, providing the onboard computer with a good idea about the size and shape of the world outside. Twin video cameras (for stereoscopic, near-3-D vision) and radar fill out the picture, the radar working on the same bounce-back principle as the LIDAR but with less accuracy.
In fact, every serious Grand Challenge contender has come up with a sensing system that looks on paper very similar to the Red Team's. Whittaker's contention is that it's not about the hardware; it's how you use it. The team is working on software to intelligently "fuse" the sensory data -- for instance, when the vehicle hits the inevitable cloud of sand, it needs to know to stop listening to the LIDAR, which gets hopelessly flummoxed by airborne particulates, and start relying on the radar, a champ in the dust. An enervating number of man-hours have been spent on what in layman's terms might be called a steady-cam system. All the sensors are being mounted on gimbals that keep them taking pictures at a fixed, precisely calibrated angle to the ground, no matter whether the Humvee has its nose in the dirt or the sky. The onboard computer can then construct a precise and reliable 3-D image of the world from the rapid-fire barrage of images it receives. That's the theory anyway. The reality is that the more the sensors bobble around, the likelier the vehicle is to run itself into a ditch.
Of course, it helps that the Humvee, with its racing suspension and 10 inches of suspension clearance, doesn't roll so easily. The route-finding systems don't have to be perfect as long as the beast can steamroller the small stuff.
Everybody loves an underdog, especially if the "underdog" is a group of affluent, well-connected 15-year-olds -- and so the tech corporations of southern California have opened their checkbooks and their inventory catalogs for the Road Warriors. American Honda has been the chief sugar daddy, courtesy of PV parent Tom Laymon, providing the team with an Acura MDX whose excellent computerized handling controls and navigation system are proof that the gap between the car industry of today and the autonomous military vehicle of tomorrow is less than might at first be supposed.
The SUV's inaugural desert test begins inauspiciously as we overshoot the turnoff on Highway 15 east of Barstow that will take us into the Mojave Preserve. This is an off-road vehicle, Robertson says: Just go over the shoulder, pull a U-ey, and head cross-country back to the exit. I'm driving. We cautiously plow into the dirt shoulder and nose down an incline, a large rock slamming into the undercarriage at the bottom. "No problem, we just scraped the exhaust pipe a bit," Robertson says, a blithe spirit if there ever was one.
Among the Road Warriors' strengths is home-field advantage. Palos Verdes is just a three-hour drive from Barstow, and a couple of the adult mentors have been able to recognize exact course locations in some of the desert photos posted on the DARPA Web site, DARPA's effort to scramble the backgrounds notwithstanding. So we visit a few.
jittering along and yells, "Whoo, this is fun!" The other Road Warrior adults along for the ride say a little prayer for the edge-recognition capabilities of the team's as-yet-unwritten software, and for a Grand Challenge course that doesn't go too far into the rough.
We come to a steep, twisty, three-foot descent, "the Chinese Wall," and we first get out of the car to mentally work out a route so the Honda doesn't stick between the uneven rocks or flip altogether. The car manages it like a pro, but how well are any of the competitors' onboard computers going to do this sort of thing?
The kids in my car, none of whom are key technical players, seem not unduly perturbed. "Hey, Jeff, you've been voted off the island," Ashton Larson yells from the way back. Jeff, son of Honda benefactor Tom Laymon, replies, "Too bad, it's my car."
"Dude," Larson fires back, "you're acting like me and my sister. That's a bad thing."
Before jetting off to Atlanta and Houston to beg for free technology, Whittaker convenes an early afternoon council of war with a core Red Team group. The subject, of the meeting and the begging, is mapping -- and it's a big one.
Given the state of the technology, there is no way a race vehicle could average 20 mph over rough desert terrain without some stored knowledge of the route. (Humans are no different. Send a driver onto an obstacle course he's never seen before, and he'll nose around it gingerly. Let that driver practice the course for a year and he'll fly, continually moving back and forth in his brain between what he knows and what he sees.) It is the Red Team's aim to construct the most precise, detailed maps possible, maps not of the course, since its exact coordinates are as yet unknown, but of the entire area the course might run through.
The team starts with U.S. Geological Survey aerial photography of the area. Then it adds additional layers -- USGS vector maps that show roads, paths and riverbeds, and data from a U.S. Bureau of Land Management Web site that gives the team a good idea of what parts of the region legally can and can't be used for the race. The end result, the Red Team hopes, will be a mapped image of every square meter of potential course. For the finishing touch, CMU is visiting areas that have a high probability of being on the actual course -- riverbeds, old paths and the like -- and marking its own GPS waypoints every few meters. The team will then combine all this data to generate so-called breadcrumbs at 1-meter intervals -- information the team will program into its race vehicle.
The ideal here is that when the Red Team gets the true course waypoints on race day, members will crunch some data and come up with a map of their perfect course before the Humvee has even moved off the starting line: a triumph not of racing but of pre-racing. "Over the years," Whittaker admits, "some people in the robotics community have looked at the machines I've built and said, 'That isn't robotics; these aren't machines that are reasoning about their world.'"
So what? he figures. They work.
When I return to Berkeley two weeks after our model-motorcycle race, Levandowski and I make our way to the testing grounds with his full-size bike strapped down in the back of his Nissan pickup. The steering software has been patched together in a fit of last-minute improvisation.
Levandowski's challenge is to design the computer algorithms that will register how far the bike is tilting in either direction and apply just the right amount of counterforce via the electronic steering and acceleration controls that are powered by a 12-volt electric motor mounted atop the bike. His Yamaha, like all the vehicles in the race, is internally monitored by an inertial measurement unit (IMU) that registers, among other things, its orientation in space, or attitude -- its pitch (front-and-back motion), roll (side-to-side) and yaw (left-to-right). For any vehicle, attitude must be taken into account to get a precise IMU read on how far it has traveled along the course. For a motorcycle, the skillful manipulation of attitude, the rate of roll in particular, is the only thing keeping the machine up. Making a computer perform at the necessary skill level is a trick Levandowski has not yet mastered.
In the face of the gulf between his vision and the current reality, Levandowski reassures himself by focusing on the big picture. "For other teams," he says, "the race is what makes their vehicle important. For us, the race is just a field test. Our work starts the day after." Levandowski sees himself in the business of developing the next generation of autonomous military vehicles, this in contrast to many team leaders -- including Red Whittaker, who visibly blanches at the mention of military applications, his own field-general persona notwithstanding. "The other teams are working on perception and terrain recognition," Levandowski says, "and I'm working on increasing the mobility of autonomous vehicles. That's fine. Someday, their systems will be on my chassis."
His blueprint for a roboticized military future is an influential 2002 National Academy of Sciences document that sketches out an entire genealogy of future military unmanned ground vehicles (UGVs). He says that the other teams are, whether they choose to be aware of it or not, helping to lay the technical foundations for UGVs like Donkey and Wingman, which are supposed to perform troop supply and support in the coming decade. Levandowski is under no illusions about where his single-track vehicle concept fits in -- not the Yamaha per se, but something more stable with, say, a double set of front and back wheels. "I'm doing the Hunter-Killer," he says -- the machine whose mission, in the bloodless language of the NAS report, is "doctrinally quite straightforward."
Five amazing, clean technologies that will set us free, in this month's energy-focused issue. Also: how to build a better bomb detector, the robotic toys that are raising your children, a human catapult, the world's smallest arcade, and much more.