Behind schedule and costing about $60 million so far, the government-owned National Advanced Driving Simulator (NADS) operated by the University of Iowa is about to come online. Its sponsors may be gearing up to celebrate, but others still are asking the question that has plagued the project from the start: Why is so much money being spent on something with no proven value for highway safety research?
A longtime skeptic is Institute president Brian O'Neill (see "NHTSA moves ahead with $49 million driving simulator," April 20, 1996). "Not only was this simulator expensive to build, it will cost a lot to operate," he points out. "No matter how sophisticated it is, the subjects in it will know they're being studied, so researchers will never know if they're observing normal behavior or, what's more likely, special behavior. Simulators violate a basic principle of behavioral research — people often behave differently when they know they're being observed."
Leonard Evans, recently retired senior research scientist at General Motors, shares this view. "Crashes have more to do with driver behavior than driver performance, and the only thing you measure on a simulator is performance. By the very nature of the process, people are performing a task." Evans says he knows of no important research findings from high-level simulators.
"There's not much that can be studied on a simulator that can't be done through simpler means," Evans also says. One presumed advantage of simulators is that drivers can be put in situations where crash risk is high. But, as Evans points out, "suppose you had a magic vehicle that had the property that after a crash, all human and mechanical harm was instantly canceled. What list of wonderful research studies could you do with it? I'm not saying the list is empty, but there's not a lot on it."
Crash avoidance researchers at the National Highway Traffic Safety Administration (NHTSA), where NADS originated, disagree. Although NADS won't be officially unveiled until later this year, NHTSA already has requested $4 million for two studies in 2001. One will look at driver distractions, and the other will focus on the effects of alcohol under demanding driving conditions.
Michael Goodman, the human factors researcher at NHTSA who will head these studies, admits validity is a concern but says "we're going to have a minimal problem" with that. "We believe it will be a very compelling driving experience, compelling in the sense of being very realistic. I think subjects will be absorbed in the task and won't focus on the fact they're being observed." He says the agency plans to add a validation component to each project. This will consist of performing tasks on the simulator and then taking corresponding on-the-road measures, checking for differences that suggest the simulation isn't valid.
This approach to validation "makes no sense," O'Neill counters. "If you can observe the behavior on the road, why do you need the simulator? The advocates of the simulator need to go back and study the principles of scientific experimental design. Simulators violate those basic principles."
Quest to build a better machine
While NHTSA's first batch of NADS-based studies still are in the planning phase, the focus has been on the technical challenges of building the machine, housed at the University of Iowa's Oakdale Research Park.
NADS is billed as more technologically advanced than either its U.S. predecessor, the Iowa Driving Simulator, or the world's only other advanced driving simulator, DaimlerChrysler's in Germany. The key technical innovations involve NADS' motion base and the fidelity of simulation — how closely it approximates the real thing.
A distinguishing feature of the motion system is its capability for large excursions, which are made over a grid of rails and belts. These slide the simulator dome, which is big enough to hold a truck tractor, up to 64 feet laterally and lengthwise across the room. The dome itself sits on six hydraulic legs that move up and down to simulate pitch and roll, and a rotating platform inside the dome reproduces spin.
Another advancement is the visual system with its 360-degree field of view so test drivers can see the simulation ahead, to the sides, and in all three mirrors. The graphics are more realistic than those of the best computer video game, according to NHTSA.
High-end computers run all subsystems simultaneously — motion, visual, sound, plus the dynamics specific to the vehicle being simulated. NHTSA says NADS can copy the dynamics of a Ford Taurus, Chevrolet Malibu, Jeep Cherokee, and Freightliner Class 8 truck cab.
$34 million price tag was only the start
Complexity adds to the price. The University of Iowa's software contribution is valued at about $5 million. TRW's contract to build NADS was worth $34.1 million. With "cost growth," this contract now stands at about $50 million, considerably more than NHTSA's initial $32 million estimate.
Operational costs will be hefty, too. Revenues are expected from rental fees of $1,000 per hour. But from the beginning, congressional leaders and others have been skeptical that there will be enough private users to generate the $9 million a year needed to keep NADS running.
Government users, mainly NHTSA but also including the Federal Highway Administration and Federal Motor Carrier Safety Administration, will have access to two-thirds of the rental time. The University of Iowa will get to market the rest of the time to private users. Many of these users probably will be in agriculture or defense, not highway safety.
Keith Brewer, who directs NHTSA's Office of Human-Centered Research, says the motion base and higher fidelity cueing systems added much of the cost. The agency wanted these to enhance realism and particularly to minimize simulator-induced sickness, a major problem with any simulator. Brewer says it isn't known whether this side effect will be eliminated using NADS.
Genesis of a flawed project
The idea of a government-funded advanced driving simulator began with Michael Finkelstein, NHTSA's Associate Administrator for Research and Development during the late 1980s. He recalls that "there was the expectation that there would be a lot of advanced technology moving into cars" such as antilock brakes and adaptive suspension systems. "The issue was, when I put advanced technology into a vehicle, how is the driver going to respond? If I put more capability in, is it going to have a safety benefit?"
Finkelstein says the simulator seemed like a reasonable way to anchor a largescale crash avoidance research program on par with the crashworthiness program. According to NHTSA's 1993 budget justification, the driving simulator "will be to crash avoidance what the anthropometric dummy is to crashworthiness."
But Leonard Evans says this analogy "reveals a profound misunderstanding of what traffic crashes are all about. Research in crash avoidance and research in crashworthiness do not involve the same issues. A vehicle crashes into a wall the same way each time. It's reproducible. There's no behavioral element there. But obviously crash avoidance does have to do with the human element. It's very alarming that those with responsibility for the nation's safety can reveal so deep a misunderstanding."
Once NHTSA officials began considering the simulator, they primarily consulted simulator experts in commercial aviation and defense. This illustrates "another major misunderstanding," O'Neill points out. "Flight simulators are training devices, not research tools. They're valuable for training pilots to fly expensive planes and how to react in specific emergencies such as loss of power. It makes sense to train pilots this way because it's cheaper than using a real plane. But no one contends that it makes sense to train car drivers in a $60 million simulator."
Without any direct evidence that NADS would be useful for research, NHTSA's quest for it succeeded. Now that research is about to begin, the debate turns back to issues of validity and cost. "It's a $60 million gamble," O'Neill says, "and it's not a good gamble at all because it's likely people won't behave the same in a simulator as they do in their everyday driving."