THE SIGNING OF THE FLOYD D. Spence National Defense Authorization Act for Fiscal Year 2001 didn’t warrant a Rose Garden ceremony, a bouquet of microphones, or a write-up in the next day’s Washington Post. The 515-page document was routine legislation, setting the budget for the American military: which weapons it would build, how much veterans would pay for prescription medications, which rusting artifacts would be transferred to museums.
President Bill Clinton, on his way out of office, had his quibbles with the bill, sent to his desk by a Republican majority Congress. But he deemed it fine in the balance, and necessary to the nation’s security. In a statement, he praised the bits he liked: the increased housing allowances for military personnel, the authorized cleanup of a former uranium mill in Utah, funding for the next-generation F-35 fighter jet. The president had nothing to say about Section 220, which read:
For the staffers and lobbyists who wrote the bills on which legislators stamped their names, this sort of mandate was a common tool for getting things done, or at least securing the funding to try. Section 220, along with most of the bill, came from the office of John Warner, the Virginia senator who helmed the Armed Services Committee. Warner had enlisted in the navy as a seventeen-year-old in 1944, joined the Marines during the Korean War, and served as Richard Nixon’s secretary of the navy (marrying and divorcing Elizabeth Taylor along the way). By 2000, he had been a senator for more than two decades, and saw the role robotics could play in the future of warfare. The Predator drone had entered service over the Balkans in 1995, letting American pilots “fly” over dangerous territory without risking their lives.
Warner wanted the US to rely far more on such tools, even if the military wasn’t raring to make such a drastic change. “We wanted to move swifter, more forward leaning,” Warner said. “The Pentagon wanted to follow its usual, more conservative track.” A mandate, he figured, might change that attitude.
With the Predator already in service, the first part of the Section 220 mandate was just a matter of multiplying that success, applying it to more aircraft and pumping out more drones. Unmanned ground vehicles were less developed, but at the time, the advent of trucks and tanks that could drive without a person on board seemed plausible, maybe even imminent. Computers could weave a fighter jet through the air, launch a ballistic missile from a submarine, or destroy a target from a hemisphere away. Researchers in the United States, Asia, and Europe had demonstrated vehicles that could drive themselves in restricted conditions. Les Brownlee, the staff director for the Armed Services Committee, who helped Warner craft the bill, thought that with a fifteen-year window, making robotic vehicles a major presence within the armed forces was doable. And he knew America’s scientists wouldn’t deliver without a push. “We certainly wanted to challenge them,” he said.
It made perfect sense, except to people who happened to know anything about unmanned technology. The sky is virtually empty, so you don’t need much more than a good understanding of aerodynamics to fly a drone like the Predator. Driving demands the ability to find and stick to flat, or at least even, ground, and to contend with rain, snow, and fog that can blind computer vision systems, but that aircraft can fly above. It requires not just avoiding all the things gravity keeps out of the sky—trees, rocks, buildings, people, other vehicles—but understanding what they are, how they’re likely to act, and how one’s own movement affects others’ plans. Driving might be the most complicated task humans undertake on a regular basis, even if they don’t realize it.
Moreover, while Warner’s mandate called for “unmanned” vehicles, the Predator drone’s remote control setup was a nonstarter on the ground. Because flight requires relatively few split-second decisions, latency—the delay between a pilot sending a command and seeing it executed—is more pesky than problematic. When navigating the crowded ground at “tactically relevant speeds” of fifteen or twenty miles per hour, it’s devastating. (Think of elderly drivers with slowed reaction times.) Remote operation might demand a one-person, one-robot paradigm that left major benefits on the table, like reduced need for manpower. Like regular driving, it requires one’s full attention. If a soldier in the field wanted to remotely control a scout vehicle, she might need someone to stand watch for her—doubling instead of slashing staff requirements.
The challenge Bill Clinton signed into law wasn’t to connect a human to a remote vehicle. It was to teach a car to do everything a human can do. Anyone familiar with this technology who read Warner’s mandate knew that America’s unmanned ground vehicles would have to be autonomous. Warner may not have recognized the difficulty of the challenge he put into law. But he knew he wanted it done, and he knew who might be able to do it.
On June 18, 2001, Tony Tether walked into an office on the ninth floor of 3701 Fairfax Drive in Arlington, Virginia. The Stanford-trained engineer had spent plenty of time in this room in the 1980s, always on the side of the desk closer to the door. This time, though, he sat behind the desk, as President George W. Bush’s newly confirmed choice to run the Defense Advanced Research Projects Agency.
Better known by its acronym, DARPA was born into the Pentagon’s sprawling organizational tree in February 1958, as a response to the Soviet Union’s launch of Sputnik 1. That small, pinging satellite, which circled the planet every ninety-eight minutes and was visible from Earth, shook Americans and their government. Dwight Eisenhower wanted an agency dedicated to ensuring the United States would never again be surprised by a technological advance, an agency that stood apart from the army, navy, air force, and Marines. The small outfit was first called the Advanced Research Projects Agency, or ARPA. As the military nature of its mission morphed, the “D” for “Defense” was added in 1972, removed in 1993, and put back in 1996.
Like its name, DARPA had an unsettled, roving history. It started as America’s de facto space agency, then lost the role to NASA when the civilian agency was created a few months later. Without much of a mission statement or specific goals, the runt of the Pentagon spent its first decade focused on missile defense at home and counterinsurgency in Southeast Asia, researching new ideas and funding scientists with promising pitches. Those efforts produced what would become DARPA’s standard mix of results: nonstarters, embarrassing failures, and a heavy helping of projects whose impacts spread beyond whatever anyone had imagined or intended.
First tasked with the devilish problem of defending the United States against nuclear attack, DARPA explored ideas for a particle beam gun that could shoot down an incoming ICBM. That went nowhere. The agency fared better when it worked on the ability to detect Soviet nuclear weapons testing. Along with developing technology to spot such testing in outer space, DARPA installed seismographs all over the planet and funded research to identify tremors as natural—e.g., earthquakes—or the result of an underground nuclear test. That laid the groundwork for the 1963 Limited Nuclear Test Ban Treaty, by giving the US confidence in its ability to spot Soviet cheating. Meanwhile, DARPA’s support of seismographic research proved invaluable to the scientists who presented the theory of plate tectonics.
DARPA’s greatest success started in 1961, when Joseph Carl Robnett Licklider joined the agency to do some behavioral science work and improve the military’s ability to counter conventional and nuclear weapons in times of crisis. Licklider was a psychologist with a deep interest in the budding field of computing. He focused his energy on the command-and-control assignment, which he saw as one of many potential applications for his grand vision: a network of computers that did more than arithmetic. He funded research at places like MIT, Stanford, and the defense-oriented think tank RAND, aiming to connect a few computers in the same room. In 1965, one of Licklider’s successors, Robert Taylor, decided to pursue the idea on a grander scale. In a fifteen-minute meeting, he squeezed a million dollars out of his boss and used it to create the ARPANET—the network that became the internet.
Less eulogized is DARPA’s work in Southeast Asia. In May 1961, the agency launched Project AGILE, a counterinsurgency program proposed by William Godel. An intelligence operative and one of the agency’s first employees, Godel cranked out innovative, often absurd ideas for helping embattled Vietnamese president Ngo Dinh Diem fight the Communists coming from the north. DARPA experimented with portable flamethrowers, mines made to look like rocks, and a near-silent “swamp boat” that could carry thirty men through water just three inches deep. But Godel was especially interested in destroying the crops and jungle foliage that fed the Viet Cong and let them covertly move supplies and launch ambushes. DARPA funded the development of a range of chemicals, millions of gallons of which American C-123 cargo planes would pour over South Vietnam. The best known of these was called Agent Orange. It ravaged the land and left behind a trail of cancers and birth defects that devastated Americans and Vietnamese alike. As antiwar sentiment built up at home, DARPA was moved from its original office in the Pentagon to the Fairfax Avenue building in Arlington—a physical manifestation of its bruised reputation.
These diverse efforts were all born of DARPA’s defining trait: flexibility. The agency worked nothing like the rest of the military. It usually employed no more than a few hundred people and was largely unbound by the bureaucracy that dictated life in most of the government. The director had the office on the top floor, but the direction came from the program managers who made up more than half the head count. These were physicists, chemists, biologists, and engineers, academics and industrialists, civilians and service members. Their job was to come up with potential solutions to stubborn problems they encountered, a new kind of communication device or armor or navigation system. They pitched the director on the program they wanted to run and, if approved, found and funded the companies or universities or whomevers who could make their ideas real. Program managers often lasted just a few years. Few went more than five. DARPA favored constant turnover, prioritizing new thinking over institutional memories, especially of failures. When a project worked, DARPA handed it off to the military or private sector for commercialization, and went looking for the next wild venture.
By the time Tony Tether first came to DARPA in the eighties, this approach—hunting down innovative leaps to solve real problems, dodging bureaucracy all the while—had produced or laid the groundwork for the stealthy F-117A fighter jet and B-2 bomber, the M-16 rifle, the Predator drone, and GPS. Then forty years old, Tether had the look and CV of a defense industry lifer. He wore his hair slicked down and seemed to have stopped buying new glasses around the time he got his PhD in electrical engineering, in 1969. Tether spent four years as the head of DARPA’s Strategic Technology Office, doing work that remains classified. When the DARPA director job opened up in 1985, Tether went for it and lost. He returned to the private sector, where he stayed until Defense Secretary Donald Rumsfeld brought him in for an interview.
Along with his engineering background, Tether’s love for science fiction made him a good fit to run DARPA. As a kid, he had listened to Sputnik beeping overhead on his ham radio. He was enamored of novels like Robert Heinlein’s The Moon Is a Harsh Mistress, where humans colonize the Moon, then start a civil war with those who remain on Earth. “I believe strongly that the best DARPA project managers must have inside them the desire to be a science fiction writer,” he said. H. G. Wells, he thought, would have been a fantastic employee. But by the time Tether sank into the director’s chair in June 2001 and added a few personal touches—he didn’t bother swapping out the old furniture—it hardly mattered whether his deputies had read any sci-fi, let alone written their own. America’s great bogeyman, the Soviet Union, was long dead, and with it had gone the agency’s motivating force. The 1990s had been about the peace dividend, not defense spending. Through the first summer of Tether’s tenure, Americans weren’t watching for an invasion or nuclear attack. The US was the world’s lone superpower, and needed its mighty military only to swat at the occasional militant group in Africa or the Middle East. “DARPA had become a backwater,” Tether said.
A few months later, on a sunny Tuesday morning in September, Tether’s secretary pulled him out of a conference room and directed his gaze out the window, to the east. Black smoke was filling the sky over the Pentagon, DARPA’s former home. Soon America was back at war. In Washington, defense once again took center stage, and the money flowed: From 2001 to 2005, DARPA’s annual budget increased 50 percent, to $3 billion.
Right away, Tether diagnosed the attacks as resulting from a failure of intelligence. He wagered the clues were all there, just not in one place, where any one person or agency could put them all together. Within months, he launched an intelligence gathering project pitched to him by John Poindexter, then senior vice president of SYNTEK Technologies, as “A Manhattan Project for Combatting Terrorism.” Poindexter was best known for his central role in the coverup of the Reagan-era Iran-Contra Affair, but Tether was willing to overlook his shady history. He thought he was the right man to run a project they called Total Information Awareness. But before long, September 11 led to military questions that weighed more heavily on the public’s mind than ferreting out terrorists.
In Afghanistan and Iraq, American men and women in uniform met a vicious antagonist: insurgencies using roadside bombs to kill and dismember the troops traveling local roads. As the hopes for a quick and glorious romp through the Middle East soured, Tether kept thinking about John Warner’s unmanned vehicle mandate, and what DARPA could do to fulfill it.
The dream of a vehicle that drives itself dates back to the early days of the automobile, as people abandoned sentient horses for machines that punished any lapse in attention. In 1926, the Milwaukee Sentinel announced that a driverless “phantom auto” would tour the city, controlled by radio waves sent from the (human-driven) car behind it.
The idea went national with Futurama, General Motors’ exhibit at the 1939 New York World’s Fair. Millions of Americans braved hours-long lines for the chance to sit in the navy-blue mohair armchairs that would take them on a seventeen-minute tour of a “wonderworld of 1960.” During the tour, when they weren’t too busy necking with their sweethearts, they ogled massive dioramas of a national highway system that eliminated crashes and congestion, where radio control systems kept everyone in each of the fourteen lanes going a set speed and staying a safe distance apart. At the height of its power at the time, GM kept at the idea. A promotional video for its 1956 Firebird II concept car explained that “the driver might just push a button, and the car would literally drive itself” by picking up electronic signals from the highway. The automaker teamed up with RCA to build a test track in Princeton, New Jersey, but soon abandoned it as impractical at scale. In the 1960s and ’70s, researchers at Ohio State; University of California, Berkeley; and in Japan and Germany did similar work.
All these concepts, though, were limited in scope to the easiest part of the driving problem, cruising on the highway. With the cars pointing in the same direction, all you needed was a way to keep them in their lanes and away from one another. Given the right mix of infrastructure and in-car tech, the problem seemed tractable, if hard to implement at a national scale. No one seriously considered making a car that could negotiate a more complex environment, with intersections, traffic signals, and pedestrians. Even in America’s most enthusiastic portrayal of the future, the problem went untouched: George Jetson did his own driving.
That’s because the kind of technology that might be able to mimic the human driver—who surveys his surroundings, analyzes their elements, predicts how the scene will evolve over time, and moves accordingly—just didn’t exist. Not yet, anyway. While GM was laying underground cables, researchers in Palo Alto, California, were pioneering a field they called “artificial intelligence.”
The first robot that could move around and “think” about its actions was Shakey, built by the Artificial Intelligence Center at Stanford Research Institute (which has since split from the university and changed its name to SRI International). But the first machine widely recognized as an autonomous vehicle came from nearby Stanford University’s Artificial Intelligence Laboratory, or SAIL. The Stanford Cart had been built in 1961, part of research into how well a human on Earth might be able to control a rover on the Moon. This cart, which looked like a card table riding a quartet of bicycle wheels, spent the next twenty years being passed from one researcher to another, each using the platform for his own application. By the time the Austrian-born computer scientist Hans Moravec adopted it in the early 1970s, it could use a camera to follow a wide white line painted on the ground, in very specific conditions, at not quite 1 mph.
Moravec wanted to make the computer do more, and found his solution by watching some lizards he had caught and kept in a terrarium. Before pouncing on a fly, the lizards would fix one eye on their prey, then sway their head from side to side. Perhaps, Moravec thought, a computer could calculate the distance of objects it saw the same way. So he put the camera on a slider and programmed it to move from one side to the other, taking photos along the way. With this spin on stereo vision, the cart’s computer (which took up most of a nearby room) would pick out spots of high contrast, the things most likely to be objects. By comparing their positions in the sequence of photos, it could fix the location of each in space. Moravec would let the cart loose in a large room or an outdoor space strewn with chairs, trees, and cardboard icosahedrons (imagine overgrown twenty-sided dice). It navigated slowly, pausing for ten to fifteen minutes between one-meter dashes. It bumped into things. But it navigated, covering one hundred feet in five hours. “Similar humble experiments in early vertebrates eventually resulted in human beings,” Moravec wrote in his 1980 dissertation.
After completing his PhD, Moravec moved to Pittsburgh to join the newly created Carnegie Mellon University Robotics Institute, where he and his colleagues fostered that evolution. In 1984, their efforts in the budding field of computer vision—the ability for a machine to see and understand its surroundings—graduated to honest to goodness vehicles with the Navigational Laboratory program. NavLab 1 was a blue Chevy panel van that carried around an extra four-cylinder engine just to generate the power to run its onboard supercomputer, camera, laser scanner, and radar. The van did most of its early testing at speeds around 1 mph and couldn’t do much more than follow the road and spot obstacles ahead. But it was maybe the first robot roomy enough for its makers to work inside it as it moved, making it a viable transportation option. As computers got faster and sensors improved, the CMU team produced a series of vehicles with increasingly humanlike abilities.
When they got to NavLab 5, they decided it was time for a road trip. Researchers Dean Pomerleau and Todd Jochem had developed a program called the Rapidly Adapting Lateral Position Handler (RALPH), which used a camera to look for lane markings, road edges, and discoloration from dripped oil to find the center of the road and stay there. Eager to see how RALPH, installed in a gray Pontiac Trans Sport minivan, would fare on highways with different sorts of road markings and types, in the summer of 1995 Pomerleau and Jochem launched a cross-country drive they called “No Hands Across America.” (Since their focus was on the road vision system, they worked the gas and brakes.) Over nine days, listening to Star Trek books on tape to pass the time, the computer scientists covered the 2,849 miles to San Diego, letting the car do nearly all the steering.
America wasn’t the only country with robots hitting the road. Even before Carnegie Mellon developed its NavLab vehicles, the Tsukuba Mechanical Engineering Lab in Japan created a vehicle that drove itself at 20 mph by using cameras to pick out lane lines. In the 1980s, Germany’s Daimler launched a program it called Prometheus (that’s “program for a European traffic of highest efficiency and unprecedented safety” in German). Collaborating with computer vision pioneer Ernst Dickmanns, Mercedes-Benz’s parent company built a series of vehicles that drove themselves in simple (e.g., highway) settings at various speeds, culminating in October 1994, when an S-Class sedan with video cameras drove itself six hundred miles on a multilane motorway, even changing lanes and overtaking other cars. In an offshoot of the Prometheus program, Italian researchers made a car that drove them more than seven hundred miles.
Apart from GM’s early toe dip, the American auto industry was neither interested nor involved in such projects. The funding for the research being done at Carnegie Mellon and elsewhere in the country came largely from the Department of Defense—and from DARPA in particular. The Pentagon’s Skunk Works arm had long seen robotics as a vital capability for the future of warfare, and exactly the sort of long-view work it was designed to tackle. It had funded the scientists who built Shakey and some of Moravec’s work with the Stanford Cart. In the early 1980s, DARPA started an autonomous land vehicle program, using an eight-wheeled ATV, which by 1987 could follow a curving road made up of different kinds of pavement. Through the 1990s, DARPA collaborated with the army and sponsored research by various defense contractors and universities, focusing on making vehicles that could run scouting missions.
In total, decades of scattered and sporadic work by a variety of parties—chiefly government-funded academics—had produced a solid foundation for the kind of military force John Warner demanded. It wasn’t much more than that, though. The elite robots of the early 2000s were slow, expensive, unreliable, and not a sure bet to beat the old Stanford Cart in a race. The idea looked like the eternal research project, one built of steady, incremental advances that would produce a steady stream of PhD dissertations.
Now Congress had told DARPA it had less than fifteen years to turn these raw robotic recruits into a nimble, hardy, and ubiquitous fighting force. Scott Fish, a program manager running one of the agency’s robotics efforts, remembers the word coming down. “That’s a hell of a challenge,” he thought.
As he pondered how to meet Warner’s mandate, Tony Tether thought of another line from Congress, this one tucked into the National Defense Authorization Act for Fiscal Year 2000:
The Secretary of Defense, acting through the Director of the Defense Advanced Research Projects Agency, may carry out a program to award cash prizes in recognition of outstanding achievements in basic, advanced, and applied research, technology development, and prototype development that have the potential for application to the performance of the military missions of the Department of Defense.
Tether didn’t have to let out yet another series of contracts and hope America’s academics and defense contractors turned up a miracle. He could join a storied history of innovation triggered by competition. He would create a twenty-first-century version of the Orteig Prize, the $25,000 reward offered by a New York hotelier that Charles Lindbergh won by flying solo across the Atlantic.
The idea of a contest had been kicking around DARPA before Tether came on as director. That snippet of law was the product of Rick Dunn, the agency’s general counsel, who worked to enhance the agency’s prized flexibility. He convinced legislators to give DARPA the right to enter into partnerships with commercial companies and to dodge civil service laws so it could better recruit scientists and engineers. The ability to award prizes instead of contracts was just one more element in what Dunn called “an ecosystem of doing business in an innovative way.”
The prize authority didn’t make much sense for most DARPA projects. If the goal was to develop a stealth plane or new sort of rifle, the standard sort of contract with one or a few defense industry contractors was the right bet. But Tether recognized a particular swirl of factors that made an unmanned ground vehicle the ideal application for a contest open to the public.
First, the impetus was there, in the form of the Warner mandate and in the rapidly developing conflicts in the Middle East that threatened to draw in more and more American soldiers. Second, the technological foundation had been laid. Cars and computers were readily available. Anyone with the right phone number and some cash could get the servers, cameras, radars, or laser range finders a vehicle might need. Third, the intellectual foundation was there, put down and published by researchers at Carnegie Mellon, Stanford, and elsewhere.
Tether believed the trick at this point was more integration than invention, finding just the right way to mix together all the existing pieces. His quest was for the secret sauce, and he couldn’t have too many cooks. He wanted to bring in brilliant folks who had nothing to do with the Department of Defense or DARPA. Tether just needed to find them and lure them into the kitchen.
Moreover, the idea of a contest appealed to Tether in a way no contract ever could. He was a scientist, but he was also a salesman. Decades earlier, between finishing college and starting his graduate studies at Stanford, the newly married Tether had gone door-to-door hawking home cleaning products for the Fuller Brush Company. He was good at it, learning to capture people’s attention, to wriggle his way inside and sell them on a vision, even one based around a mop.
A few weeks into his tenure at DARPA, when Vice President Dick Cheney came by for a briefing on the agency’s current projects, Tether planned his lineup like a manager going into the World Series. Ever the salesman, the director started with DARPA’s most exciting work. Figuring Cheney wouldn’t stick around too long, Tether saved his less compelling employees for last. From start to finish, he thought about the story of each program. “I picked them if I thought, ‘Wow! These are really important, and not only are they important, but they can be briefed in a way to show that they’re important.’ ” Tether knew Cheney from the VP’s time in the House of Representatives, and made sure the presentations focused on pictures, not words. By Tether’s account, Cheney—joined by Secretary of Defense Donald Rumsfeld—loved the briefing, and left Fairfax Avenue as a DARPA ally.
In the same vein, Tether changed the tone of the DARPA Systems and Technology Symposium. The agency held this meeting, better known as DARPATech, every two years or so to brief anyone interested on what it was up to. It had long been a sedate, serious affair, held in a city like Dallas or Kansas City. Tether moved it to a Marriott in Anaheim, California, across the street from Disneyland. The location was more than subtext.
“Disneyland is a land of dreams and fantasy becoming reality,” Tether said in his opening remarks at the DARPATech held in August 2002. “That is what DARPA does—and does well.” He had made each branch of the agency pick a theme to fit the occasion. The Information Exploitation Office got the Sorcerer’s Apprentice. The Tactical Technology Office was Frontier World. The Microsystems Technology Office took It’s a Small World. “We were there to tell people what we were doing, and to get people to come with ideas,” Tether said later. “I knew it was a show.”
Before ceding the podium to a succession of deputies who would go over everything from a “self-healing minefield” to “fiber lasers,” Tether made one more announcement. In addition to all its standard programs (if you call monkey-based mind control experiments standard), DARPA was planning a race for fully autonomous vehicles. It would take place at the next DARPATech conference, sometime in 2004, and run from Los Angeles to Las Vegas. Whoever finished in first place would take home a million dollars.
Tether called it the DARPA Grand Challenge.
Tether announced the race before sorting out the details. He hadn’t scheduled the next DARPATech conference yet. He figured the cars could run on the I-15 freeway, but wasn’t sure how, or if, he could take over a major interstate. He didn’t know who might want to build an autonomous vehicle, or how long they would need. He had no idea how to put on a race, let alone one for robots. “We honestly did not know what we were doing,” he said. But he knew he wanted something that would pass what he called “the grandma test.” It had to be so straightforward that anyone could watch and know who won.
To sort out the realities, the director called on Jose Negron. The air force colonel had been assigned to DARPA in September 2001, arriving just days before the terrorist attacks that supercharged the agency’s role in America’s arsenal. Now his job included turning Tether’s idea into a plan.
Negron’s first move was informing his boss there was no way DARPA had the clout to shut down an interstate, let alone one that ran through Los Angeles. They wouldn’t be able to send robots charging into downtown Las Vegas, either. Negron suggested starting in Barstow, an old railroad town about a hundred miles northeast of LA. The race would go mostly off-road, through the Mojave Desert. It would finish in Primm, a sad sack gambling town just over the Nevada border, frequented by Californians too eager to drive the next forty miles to Vegas.
Anyway, Negron thought, the Mojave was a more appropriate testing ground for military vehicles than a freeway. The land between Barstow and Primm—about a hundred miles as the crow flies—included fire roads, dirt paths, and the occasional stretch of pavement. It featured a mix of flat, open terrain and steep hills with narrow paths. In other words, it approximated many of the driving environments of the Middle East. If a vehicle could navigate the Mojave, it could handle Afghanistan and Iraq.
Running through the desert would also let the DARPA crew modulate the difficulty of the challenge. The terrain would be rough, but they would space out the vehicles, so each team would only have to deal with an unmoving world—avoiding rocks, cactus, barbed wire, and the like, not other vehicles. Maybe someday the cars could tackle more complicated tasks like negotiating traffic. For now, just going from one point to another at a reasonable speed was tough enough.
The American desert, though, presented unexpected problems. For all its crazy history, DARPA had never had to worry about running near the burial grounds of Native American tribes. Negron had to convince LA Water and Power that the robotic vehicles wouldn’t topple their towers, and work with law enforcement to close the necessary roads and railroads in the area. And he had to negotiate with one of the Mojave’s most established populations.
Long before Tony Tether had called this place his racetrack, before Chuck Yeager broke the sound barrier in the skies overhead, before the white men came through on their search for gold, even before the Mojave and Chemehuevi tribes settled there, the sands of the Mojave were home to the desert tortoise. By 2003, habitat loss, disease, and hungry ravens drawn into their territory by human activity had turned the ponderous creatures into an endangered species. That status, at least, earned the tortoise an ally in the US Fish and Wildlife Service, which had no intention of seeing them crushed by a passel of off-the-leash robots. To placate the bureaucrats, Negron had to draw up an environmental action plan that would keep the animals safe, going beyond “If one’s on the road, have somebody move it.” The problem with picking up a desert tortoise, he learned, is that if you scare or stress it, it tends to pee. In a part of the world that gets five inches of rainfall a year, that lost hydration can be a death sentence. Even touching the animal is a violation of the Endangered Species Act. So Negron pledged to bring in a team of wildlife biologists to survey the course before and during the race, ready to fence in and guard any tortoises who wandered near the route, and with the training and legal right to move one if absolutely necessary.
When he wasn’t haggling in the desert, Negron was at DARPA HQ, building a new kind of command-and-control system. He wanted a chase car following each robot through the desert, so his crew would need a way to track every vehicle. They would need the ability to remotely stop the robotic cars in case they went rogue, or to keep them from hitting one another. And they wanted to be able to start them remotely as well, for situations where they were just pausing a vehicle momentarily. The result, Tether said, was as complex as any vehicle communication system that the US military was using at the time.
The most foreign challenge, though, was designing an off-road race course. That’s how Negron ended up on the phone with Sal Fish. A Los Angeles native in his early sixties with swept-back white hair and a bushy mustache venturing just beyond the corners of his mouth, Fish was something of a desert racing legend. He’d spent several decades in charge of SCORE International, the body that runs some of the world’s most popular and grueling off-road races. He knew everything about creating courses that would push vehicles to their limits, and nothing about the Defense Advanced Research Projects Agency. So when his secretary told him a man from DARPA was on the line, Fish thought it was a prank call from one of his friends. Then the guy started talking about his plan to send a fleet of vehicles—without drivers inside—careening through the desert. “I have no idea what you mean by ‘autonomous vehicle,’ and I don’t even think I can spell it,” Fish told him. But he was intrigued.
After a trip to Washington to meet with Tether, Fish became an enthusiastic member of the team, scribbling notes onto a yellow legal pad in meetings full of people using laptops. “I had no clue what the hell these vehicles were going to look like, what they were going to do,” Fish said. But DARPA had asked him for a course worthy of a Grand Challenge, and he would oblige. He explored the possible paths between Barstow and Primm, looking for a route that would include open terrain, elevation changes, hairpin turns, and as many dangers as possible.
As the details came together, Negron decided it was time to give the public more details than “autonomous vehicle race.” He organized an informational session for February 2003, inviting anyone interested to come learn more about the rules, ask questions, propose ideas, and link up with partners and sponsors. Negron rented out the Petersen Automotive Museum in Los Angeles, thinking it a good venue for creating the future of driving.
As he made his way through LA traffic to the Petersen that February morning, Tether was sweating the small stuff. So detail-oriented that one program manager wondered if he had an eidetic memory, he was not thrilled to find that Negron had taken over an entire museum—and planned for a huge event. He thought that maybe ten people would show up, that they could have made do with some pizza and beer. What are we going to do with $2,000 worth of food? Tether thought. He planned to give the leftovers to the local homeless population.
As he pulled up to the museum, a hulking concrete building at the western end of LA’s Miracle Mile, Tether realized there wouldn’t be much to donate after all. A line of people, four abreast, stretched down the block and around the corner. Half an hour before the doors opened, hundreds of people were waiting to hear more about the Grand Challenge.
Holy cow. We really might have something here, Tether thought.