After decades spent designing smarter airbags, antilock brakes, and collision-avoidance systems, Detroit has set its sights behind the wheel. Now the cutting edge of car safety is the driver. Brace yourself.
Engineers have worked for a century to keep cars safe. They've created taillights and padded dashboards; developed multi-airbag interiors and whiplash-proof headrests; simplified gear shifting and invented antilock brakes. By the late '90s, hundreds of millions of dollars of R&D brought them to the ultimate safety solution: crash-warning systems. Detroit had found a way for cars to tell us when danger was imminent.
To those who developed the new technology, the idea was a no-brainer - radar-based systems that recognize threatening situations and notify drivers in time for them to avoid trouble. Yet, where safety engineers saw obvious benefits to the collision-warning devices, a new set of researchers saw problems. Cognitive scientists, whose opinions were traditionally neglected in questions of auto safety, suddenly found a leading role in Detroit. At issue: Would the new anticrash gadgets freak out more people than they'd help? The safety gurus found themselves deluged with questions from the brain guys. When would the alarms blare? What would they sound like? How would drivers react?
"We wondered, could this only complicate things?" says Raymond Kiefer, a veteran General Motors human-factors/safety engineer. "Are drivers going to know what to do?"
Then word came down to Kiefer from the GM brass: Find out. In the winter of 1995, a team of researchers from General Motors and Ford embarked on an unusual joint project. As the engineers developed collision-warning systems, scientists with the Crash Avoidance Metrics Partnership began exploring how people would react to the bleating alarms and flashing warnings triggered by the new systems.
Kiefer's team devised a seminal driving study. Because collision warning makes the leap from a passive to an active safety system - the alert must be followed by driver action - the CAMP staff couldn't recruit human crash-test dummies. "A potential accident wasn't something we could simulate," says Kiefer. "The simulator can't deliver the feeling that you're about to hit something with mass." But a near miss can fool test subjects into thinking they'll hit something with mass. So Kiefer engineered a "safe" crash. First, his team built a rubber model of the back of a Mercury Sable, complete with working brake lights. Then they attached the faux back end of the Sable to the rear of a lead car by a 40-foot pole that acted as a shock absorber. A third vehicle - with a driver and a test administrator inside - followed the Sable as if it were just another car on the road.
In a video from the study, a woman is driving behind the Sable on a GM test track when an administrator in the backseat tells her to look for a nonexistent light on the dashboard. As her eyes wander downward, the administrator brakes the lead car - and the Sable slows with it. The test driver's vehicle quickly closes in, urgent chirps sound, and a green icon of two cars nearly colliding flashes on the console. When the woman looks up her face conveys true fear: I'm going to crash.
Kiefer's team found that drivers prefer alarm clock-style warnings to bellowing recorded voices, and that even a last-second alert gives people enough time to hit the brakes. Still, the scientists called for more research: Should the device expect drivers to behave differently on slippery roads? What happens if two collision-warning alarms - one for forward collisions, another for rear-enders - sound together? Could the new technology encourage recklessness? Those questions and others explain why the systems remain at least two years from production.
"It's not the technological components that need so much attention," says Jim Sayer, a human-factors scientist at the University of Michigan Transportation Research Institute. "It's the human element that we're grappling with."
Ninety percent of accidents are attributable to human error. A key problem is that drivers can get easily distracted - now more than ever - as we load cars with gadgets like cell phones and GPS systems. So the experts in Detroit want to know what the tipping point is. They're exploring every corner of driver behavior, from how much risk we're willing to tolerate to what we think as we blab on mobile phones. They've built machines, made wired roads, and devised experiments to create natural testing environments. They're turning cars into rolling candid cameras. The goal: to make driving safer. After all, in 2000 there were nearly 42,000 traffic fatalities and 3.2 million driving-related injuries in the US alone.
Of course safety isn't carmakers' only motivation. They also want to sell automobiles. Companies are looking to capitalize on two technological innovations: telematics and collision-warning systems. According to Forrester Research, telematics services - including voice-command emailing, Internet access, and customized traffic advisories - will grow from a $3 billion to an estimated $20 billion industry over the next four years. Likewise, automakers know the allure of crash-avoidance systems. J. D. Powers and Associates reports that more than two-thirds of new-car buyers specifically look for safety features - and they're willing to pay for them.
All of which raises the question: How much is too much? At what point will new bells and whistles - the blinking of a warning light, the "you've got mail" drone - become as distracting as split-screen TV?
Permanently parked in a tall building at Ford's Scientific Research Lab in Dearborn, Michigan, the Virtual Test Track Experiment, or Virttex, is a white, 24-foot, fully enclosed dome perched atop six hydraulic supports. Inside sits the entire body of a Ford Taurus sedan. Five projectors above the sedan display a 300-degree, computer-generated view of the road, traffic, and scenery on the interior walls, and four speakers provide highway noise. The simulator responds to the Taurus' accelerator, brake pedal, and steering wheel to provide the ultimate virtual drive.
"We may not have any kind of firewire port that plugs into the back of a driver's head and downloads his current state of awareness," says Jeff Greenberg, a Ford safety researcher and Virttex project manager, glancing at the contraption from a windowed control room beside the simulator. "But we do have this."
Researchers like simulators - and the $10 million Virttex in particular - because they're an effective way to tease out people's real driving habits. After five minutes at the wheel, most subjects call on learned responses and behave as if they're really on the road. Greenberg settles in front of a bank of monitors in the control room as a bearded man climbs into the simulator. "Let's see how this guy does," he says.
One screen displays the driver's face, his eyes darting between the digital road and the chatty Ford researcher who's sitting in the front seat beside him. Other screens show the subject's foot on the pedals and the view out the front and back of the Virttex's Taurus. Computerized traffic appears behind the vehicle, and the guy starts talking to the researcher.
A couple of cars surge past the Taurus. Vehicles behind the test subject begin changing lanes erratically. A car abruptly pulls in front of the Taurus. The driver hits the brakes, and the simulator shifts and hisses like a mechanized spider, its pressurized limbs tilting to generate realistic motions.
Greenberg leans toward the monitors as the traffic closes around the Taurus. The researcher in the passenger seat continues talking, but the guy is no longer paying attention. His mouth is shut and his eyes are riveted to the road. The concentration pays off: As the traffic relents, the test subject pulls over without incident.
"Create a compelling environment and people try doing well," says Greenberg. "They don't always crash." Throw in a few more distractions, though, and things get interesting. Another Ford simulator test required the driver to operate an onboard navigation system, a CD player, and a mobile phone. Training infrared light on the subjects' eyes and using a special camera, Greenberg found people glancing away from the road about 1.5 seconds at a time. The tasks required as many as 40 glances to complete, which at highway speeds translates cumulatively to traveling a mile with your eyes shut.
"We learn what people can do by pushing them to failure," says Greenberg. "At least in here, nothing happens. No cracked windshields." But the virtual crashes do get noticed. When Ford incorporates navigation systems into its cars in 2003, drivers will have to stop their vehicles before they tap out destinations.
Score one for the cognitive guys.
In the early days of autos, the biggest behavioral concern was whether windshield-wiper motion mesmerized drivers. When Motorola's AM car radio debuted in 1930, its critics worried about the distraction of Benny Goodman's clarinet. Back then, people watched a radio as it played.
For decades after, the cognitive testing of drivers centered on basic ergonomics. In the 1960s, scientists developed the "occlusion technique" - opening and closing mechanical shutters in front of drivers to determine how often they needed to see the road. On an empty interstate, shutters could close for up to nine seconds.
The arrival of CB radios and 8-track tape players introduced new concerns. Could drivers handle both the newfangled audio systems and their cars? By the mid-1980s, a prototype Advanced Traveler Information System (ATIS), or onboard GPS navigation device, spurred cognitive researchers to begin focusing on what they termed "spare capacity."
"ATIS was the first complex task to show up in a car," says Tom Dingus, a human-factors expert and director of the Virginia Tech Transportation Institute in Blacksburg. "Radio functions were one thing. This was a screen."
Turns out the screen made driving safer. Cognitive scientists compared driver performance using ATIS and paper maps, and ATIS's simple visual and auditory prompts came out on top. People who read paper maps while driving were seven times more likely to drift from their lanes than those who used the computerized system. The ATIS studies reinforced the idea that drivers could squeeze in tasks besides operating the car - a notion that's now put to the test every time someone answers a call or wriggles a coffee mug into a cup holder.
But the brain guys still puzzle over exactly how much spare capacity we have. When are drivers too busy to multitask? We do know that, these days, looking away from the road for nine straight seconds is out - highways of the 1960s were calmer places. Eye-glance studies show most people can successfully glimpse down for two seconds at a stretch and take 10 glances per task, enough time for the average driver to futz with what in the majority of cars is the most complex in-dash device - the stereo.
Whether that's enough time for people to safely use mobile phones remains to be seen. Simulator experiments show that handheld cellulars pose a bigger distraction than do multi-CD stereos, but real-world crash data doesn't support the research. In the National Highway Traffic Safety Administration's lone study on driving and cell phone use, mobiles played a role in only 40 of about 40,000 traffic deaths in 1995. Driver distraction-related accidents - including cell phone mishaps - were obviously woefully underreported then, as they are now. "When you get into an accident and a cop asks, 'Were you dialing your cell phone?' You're going to say, 'No, no, no - not me,'" says Vicki Neale, a human-factors expert at Virginia Tech. "No one's going to admit to that."
Lying to cops - and frustrating cognitive researchers who are trying to pinpoint spare capacity - will only worsen as people spend more time reading email and downloading traffic reports while they drive. "We do more multitasking now than ever before, and based on the notion that practice leads to improvements, perhaps we'll get better," says Art Kramer, a cognitive neuroscientist at the Beckman Institute in Urbana, Illinois.
One recent study suggests that there might be a limit to the number and kinds of tasks we can juggle. Groups of test subjects were asked to watch for changes in a series of still images. They were shown a typical street scene, followed by the same scene with a kid running into the road. The group that was also simultaneously engaged in hands-free phone conversations made 50 percent more errors than those who were watching the pictures without talking. And nobody in the study was actually driving.
One approach to the spare-capacity debate is zero tolerance. Last year, New York drew nationwide attention by banning the use of cellular phones in cars. But some experts say that clampdowns won't make a difference. "People are going to allocate spare capacity to doing something," Dingus says from a second-story conference room at VTTI. "Some drivers will want to fill that time by reading a paper or working on a laptop. They want to stay busy."
At Virginia Tech's Smart Road, Dingus and his colleagues are trying to figure out how busy is too busy. NHTSA and the Federal Highway Administration buy time on the Smart Road, and last year GM committed up to $4.8 million to VTTI for three years of advice and access to its track.
The closed-loop, 2-mile strip has variable lighting and surface sensors that allow researchers to monitor road wear. Spindly water towers can make rain, snow, or fog. Dingus adds and subtracts real-world annoyances for test subjects cruising up and down the Smart Road. I hop into a waiting Cadillac, and Jim Schell, a GM safety flack and my tour guide, drives slowly down the strip. We cruise past a wooded southwestern Virginia hillside on a clear day. Researchers prefer the VTTI facility to sterile test tracks because this strip feels like an interstate, with grades, curves, and guardrails.
"This is beautiful," says Schell, craning his neck to see the valley below. "This is nice." Our Cadillac is a production model with no telematics, but still my chauffeur falls prey to the number-one driver distraction: scenery.
Schell glances up and realizes that he's drifting into the other lane. He sheepishly rights the Caddy, then instinctively looks in the rearview mirror.
Coast is clear. No cops. Yet.
For all the new simulators and smart roads, there's no substitute for studying real drivers in real situations. Using advanced electronics and telematics, cognitive scientists will soon field key so-called naturalistic studies. The researchers' nagging concerns: How do people adapt to collision-warning technologies? By going faster? Paying less attention? And how do they act when impact is imminent? Do they brace themselves? Swerve into trouble? Freeze?
These are the exact questions GM and NHTSA are asking. Ron Colgin, a research engineer based at the nearby General Motors Tech Center in Warren, Michigan, shows me how they plan to answer them. "A camera like this will record the face of the driver," he says, pointing to a small cylinder bolted to the dashboard of the Buick LeSabre he's steering down a Detroit metro highway. "We'll capture what the subject's doing. At the end of the test, we'll dump the data from the hard drive."
Colgin nonchalantly refers to the Buick as "the red one," but the car we're test-driving is unique. GM is set to build 13 cars just like it in the next few months. Cross sections of citizens with good driving records will each get the test cars for a month as part of a yearlong, $35 million naturalistic study funded by the carmaker and the traffic safety agency.
The prototype has adaptive cruise control - a device similar to conventional cruise control, save its ability, with the help of radar, to maintain a fixed following distance from the car ahead of it. Colgin attempts to trigger the LeSabre's forward-collision warning system, which uses radar as well as GPS mapping, an outward-facing camera, and a "scene-tracking" program that assesses the directional movement of surrounding vehicles. "I'm a bit nervous doing this in traffic," he says, closing in fast on the red BMW in front of us. "Probably wouldn't make for a good article if I smashed into someone."
But it might provide a dose of reality. After all, one concern about the new technology is that it will lull drivers into a false sense of security. Studies suggest adaptive cruise control makes drivers unduly passive. Sometimes this isn't a problem. In 1998, the University of Michigan Transportation Research Institute found that drivers logged thousands of flawless miles using ACC. The fortified cruise control always responded to the vehicles ahead. In reality, however, no computer is fail-safe.
"What happens if a system's reliability isn't perfect?" asks Tom Rockwell, a human-factors consultant and former director of the Ohio State Driving Research Lab. "A guy's finally going to snap to attention half a second before he has an accident?"
Colgin and I are practically drafting the BMW at 70, raising my fears of software glitches - and prompting images of a mangled Buick and the Jaws of Life. But all systems are go. The collision-warning device kicks in only when it calculates an impact could occur based on the following distance and the difference in speed between the two objects. Unbelievably, there's still room for error.
As a warning icon flashes, Colgin says with a wobbly smile, "A good system alerts you only when you're truly in danger. It won't give false alarms." False alarms, which occur when the system's electronics mistake, say, an overpass for an 18-wheeler jackknifed on the highway, continue to dog the technology. Kiefer estimates people will tolerate two false alarms weekly.
By the time the GM research winds down, a more ambitious naturalistic study funded by NHTSA will begin. A hundred cars used by everyday drivers in the northern Virginia area will hit the road, outfitted with four hidden cameras. They'll capture how close people follow other vehicles, how well they stay in their lanes, and more. If all goes as planned, the yearlong pilot project will launch future studies fielding thousands of test subjects.
"It will give us knowledge of everything going on inside the car," says Michael Goodman, chief of driver behavior research at NHTSA. "We'll watch people in traffic jams. And learn how someone acts on an open road. You'll be able to see airbags deploy in real situations," he says. "We don't have access to that. We can't crash people in research."
Cognitive scientists know they can't stop the onslaught of distractions in today's cars. Gizmos like fax machines, refrigerated coolers, and video players no longer raise eyebrows. There are those in Detroit who believe it's fruitless to fight these developments. Don't even try to make driving free from distractions - make cars that can operate safely despite them. That's the impulse behind a futuristic project at Ford's Scientific Research Lab.
The building is littered with prototypes and bisected cars. At one end sits a metallic-brown Lincoln with several antennas on the trunk. Inside the car, a touchscreen menu has an unusual option called e-Proximity.
Ron Miller, a company safety engineer, puts the Lincoln into gear and we drive onto a lab lot. Miller creeps toward a Ford Escape SUV parked several hundred feet away. His seat belt isn't fastened.
"Proximity," he says, and a soft, computer-generated female voice inside the car repeats the word.
"We're talking to each other," says Miller, nodding toward the SUV and then gesturing down to the Lincoln. "Speed, location, heading, braking."
Miller pulls the Lincoln within 4 feet of the Escape.
"Warning!" urges the computer.
Miller gets closer.
"Alert!" she says. "Deploying countermeasures!"
What's supposed to happen next will change the way we drive - as soon as the engineers can work out the details. The SUV's nose would instantaneously dip to lower its center of gravity; the Lincoln's bumper would extend to absorb more force. Seat belts would tighten based on the estimated severity of impact. "We're developing systems so the cars won't collide," Miller says. But if two cars can converse, he continues, ultimately all vehicles will, and not just to ready for accidents.
Engineers believe that such a giant and cooperative wide-area network would make traffic smoother. Cognitive researchers see the "driverless" car as a way around all the mortal variables that make driving unpredictable. But as enormous as the technological challenges are, the psychological challenges are likely greater. Will people give up "driving" - even in exchange for more time to read the paper? Or will they meddle with the microprocessors to regain control? A century of conditioning requires undoing.
"Moving from driver-aided systems to driver-automated systems may be the logical evolution," says Rockwell, who's studied driver behavior for 40 years. "I'm just not sure drivers will evolve at the same speed."
