Automated crash notification: the car calls for help if airbags deploy
Wireless voice and data networks: facilitate on-road communication
In-vehicle sensors: allow for automated breaking, traction control and rollover protection
Roadway sensors: monitor weather conditions, create automated message signs
Radar: used by law enforcement to track speed, also used in vehicles for crash-warning systems
GPS: Computer-controlled traffic lights
Source: National Highway Traffic Safety Administration
Hotshot Hollywood directors make movies about machines that can predict the future and software programs that can peer ahead in time. Silver screen villains plot to use the predictive power for evil; heroes fight for good.
The drama makes for great movies, but it's not all science fiction: the Tennessee Highway Patrol is already using that kind of technology every day.
It's called predictive analytic software. And it could be the start of a whole new generation of traffic safety, a new tool as revolutionary as seat belts or radar.
"It's the coming thing," said Tennessee Highway Patrol Colonel Tracy Trott.
Tennessee Highway Patrol analysts plug all sorts of factors into the software -- like weather patterns, special events, home football schedules, festivals and historic crash data -- and the program spits out predictions of when and where serious or fatal traffic accidents are most likely to happen.
So the THP can send troopers to problem spots ahead of time, either to stop the predicted crashes from happening, or to be immediately on-hand to help.
The program, officially dubbed "Crash Reduction Analyzing Statistical History" or C.R.A.S.H, breaks Tennessee into five-by-six-mile squares and predicts traffic risks for each square in four-hour increments every day.
"So it might show that between 6 and 10 p.m. the probability of a serious crash is 68 percent in this block," Trott said. "And that's where the captain should direct his resources."
In the six months the highway patrol has been piloting C.R.A.S.H, the system has been accurate 72 percent of the time.
"You can't predict anything 100 percent," said Beth Rowan, one of two THP statistical analysts who work on C.R.A.S.H. "You have some days when the predictions are right on, and other days when they're way off. Mainly what you want to look for is whether the performance of the model is acceptable. And collectively, it's been very good."
The entire C.R.A.S.H. program cost $243,000 and was funded by federal grants through the Governors Highway Safety Office, according to the THP.In addition to the crash-focused model, the THP has also deployed a model aimed at predicting where and when drivers who are under the influence of drugs or alcohol will be on the road. One of the factors that program considers is the location of places that sell alcohol under an Alcoholic Beverage Commission license.
The software can consider any factor, Rowan said, and C.R.A.S.H. can even figure out which factors are irrelevant and automatically filter those out.
"The model itself goes through and identifies, if you will, what the most important characteristics are," Rowan said. "You put everything in that you can, and the model tells you what is important and what's not."
Predictive analytic technology has been around for a few years, said Mike Reade, public safety specialist at IBM, which sold C.R.A.S.H. to the highway patrol. Dozens of police departments across the country, including the Memphis department, already use similar systems to predict crime. Only a handful are using predictive analytics for traffic patterns.
Reade said C.R.A.S.H. is more like a super-smart veteran officer than a crystal ball.
"Oftentimes veteran law enforcement officers will be making those predictions themselves when they're in the field," he said. "What we do is put a lot of data and fact behind it. The volume of factual data we're using can't be done by a human. You need an analytical tool like this to sift through the volumes of data -- years of traffic data -- to come up with this type of foresight."
It's hard to tell how effective the program is at saving lives after just six months, but traffic fatalities are down about 5.5 percent from this time last year, Trott said, which is a strong drop.
Reade added that the best way to refine the model -- especially as troopers target problem spots and hopefully prevent predicted crashes -- is to test the model against itself.
"You can determine the relative accuracy of a predictive model by seeing if it would have predicted something that already occurred," Reade said.
He added that the program can be fine-tuned to track and predict almost anything -- from violent crime to property crime to which parolees are most likely to end up back in jail.
But there's at least one thing it can't do: predict winning lottery numbers.
Contact staff writer Shelly Bradbury at 423-757-6525 or email@example.com with tips or story ideas.