ANN ARBOR, MICH.—Driving is a crazy idea. Every day in the U.S., millions of people get behind the wheel of a car, truck or bus and attempt to navigate the two-ton machines at speeds often exceeding 110 kilometers an hour, hoping they will avoid each other as well as stationary objects. Every year more than 10 million of them fall short of this goal, resulting in accidents that claim tens of thousands of lives (pdf), according to the U.S. National Highway Traffic Safety Administration (NHTSA).

With so many other aspects of our lives guided by computers that never get sleepy or distracted, NHTSA is considering how to subtract at least some human error from driving.

In Ann Arbor this week NHTSA and eight automakers kicked off a yearlong test to determine whether vehicles that talk to each other can prevent accidents. The Safety Pilot Model Deployment project brings together about 2,800 cars, trucks and buses equipped with dedicated short-range communications (DSRC) devices that constantly transmit "here-I-am" signals to vehicles around them.

This deployment is the culmination of 15 years of work by the Crash Avoidance Metrics Partnership, a public–private research consortium that includes Mercedes–Benz, General Motors, Toyota, Honda, Ford, Nissan, Hyundai/Kia and Volkswagen. The auto companies have been working with NHTSA to develop technology that will help cars, trucks, buses and other vehicles avoid crashes by communicating with nearby vehicles and roadway infrastructure, including traffic signals, dangerous road segments and grade crossings.

Now, they think it's ready to hit the road.

The Ann Arbor test, managed by the University of Michigan Transportation Research Institute (UMTRI) will test several different accident-prevention approaches. The most sophisticated includes 64 cars, three buses and three trucks driven by the public and specially built by the automakers with a sampling of six safety-warning features, along with video cameras to capture traffic incidents.

The warning systems can alert drivers to hazards such as a pedestrian ahead or a car moving into an intersection from the side. An additional 100 cars and eight trucks owned by study participants have been retrofitted with the aftermarket equipment necessary to detect other cars and deliver warnings if they get too close to each other.

The rest of the cars, trucks and buses participating in the safety study have been retrofitted with only DSRC radios that broadcast basic safety messages about surrounding automobiles' speed and location 10 times per second among this smaller fleet. Messages travel on a special wi-fi spectrum designated for vehicle-to-vehicle (V2V) communication. The U.S. Federal Trade Commission has reserved 75 megahertz (MHz) of spectrum in the 5.9-gigahertz (GHz) band specifically for DSRC-equipped intelligent transportation systems.

A car's DSRC signal extends more than 300 meters in all directions, so, unlike unidirectional radar or a sensor, it can pick up the signal of another car approaching too closely from any angle. The "here-I-am" message cannot be blocked by another car, so it can, for example, detect a hazard two cars ahead.

When one of the 178 vehicles with complete warning systems picks up a signal from another that is too close, a signal alerts the driver to take action, with the hope of turning a collision into a near miss and a near miss into an even bigger miss.

Each carmaker and device manufacturer has designed its own proprietary warning system, although these systems adhere to a specific standard governed by the IEEE (Institute of Electrical and Electronics Engineers) to ensure they can communicate. The systems include bells, flashing lights and even seat vibrations to alert drivers. For this test, boxes containing the V2V equipment have been installed in the cars' trunks. If this technology comes to market, it will be more fully integrated along with all the other systems.

"The idea is to give that driver who might possibly be distracted—maybe by talking to another passenger or fiddling with the radio—that extra little buffer to get his attention back on the road and react accordingly," says Mike Schagrin, program manager for the U.S. Department of Transportation's Intelligent Transportation Systems Safety Program.

NHTSA estimates that vehicle-to-vehicle communication, or V2V, could eliminate 81 percent of all crashes where the driver is not impaired, saving lives not to mention billions of dollars in crash-related costs.

NHTSA plans to decide whether to require V2V systems on new cars in 2013 and new trucks in 2014. Eventually, the technology could be required in all vehicles, at a cost of approximately $100 per car, according to Christopher Wilson, principal of Christopher Wilson Consulting, who was involved in early research into connected vehicles.

Smarter cars—even autonomous vehicles that can navigate the roads without the guidance of a fallible human—have been in the works for years. In fact, many high-end autos already come equipped with radar and sensors that alert drivers when they drift out of their lanes or if their car approaches another vehicle too closely.

But in order to get a broad safety benefit, a significant number of vehicles must have DSRC, and achieving that penetration would likely require some kind of government mandate. The Ann Arbor study aims to, among other things, determine just how much penetration will be enough to get the safety benefit.

Nobody is arguing with the theory that cooperative driving could save 30,000 lives a year and billions of gallons of gasoline, Wilson says. But the car companies have to get on board. "All the automakers, and the government, too, realize that everyone has to do this together, or it's not going to happen. The mandate process pretty much won't go forward if the automakers scream too much," he says.

UMTRI was one of several entities that submitted proposals for the $20-million project. Most vehicles in the test are driven by employees of the University of Michigan Health System or nearby residents. The theory is they will provide a highly concentrated environment of vehicle-to-vehicle interactions.

Most participants are being paid $200 to have the device installed and show up to have data uploaded a few times during the year. If participants drop out, do not drive as much as expected or if the data stream is low, UMTRI may add drivers to the program.

DOT will begin crunching all this information after six months; it hopes to have a preliminary cost-benefit analysis completed in the fall of 2013.

If the pilot succeeds, the first DSRC-equipped cars could hit the road in 2018, according to auto industry consultant Richard Bishop, principal of Bishop Consulting. But even a successful test, Wilson says, will not answer the most pressing question for carmakers: "What are the ongoing costs of running this system, and who will foot the bill?"

View a slide show of the Safety Pilot Model Deployment project launch.