Someday, the winding roads of mountain passes may rarely have to close, even in the most blinding blizzards.
For years, Jay A. Farrell has been developing algorithms and methods of analyzing navigation data that may, one day, allow snowplow drivers to clear roads without actually being able to see them.
Farrell is a professor and chairman of the electrical engineering department at the University of California at Riverside researching autonomous vehicle control.
If you ask Farrell, he'll tell you that solving these types of problems always comes down to the math.
And having the chance to solve these problems can motivate students to learn the math.
"It's very fun," Farrell says. "When students get involved, they have a lot of fun doing it. They just want to build something -- and they learn to do the math."
Farrell describes the challenges of working with navigational data from a large third-floor office overlooking a courtyard surrounded by the three buildings used by the school. When he started at the university in 1994, Farrell was the seventh faculty member hired to join the emerging program. Now, the school has 90 faculty members.
After completing doctoral work at the University of Notre Dame that he described as "pure theory," Farrell went to work on applications in the autonomous vehicles group at Draper Laboratories in Massachusetts. There, he helped design underwater vehicles.
A few years later, after joining the faculty in Riverside, he worked on one of seven teams that competed for a DARPA grant to develop methods to trace chemical plumes underwater. The team included a biologist, computer scientists and engineers.
The Riverside team won the grant, and the scientists took the work from simulations to ocean experiments in which an underwater vehicle used the team's algorithms to trace a plume for 975 meters to its source.
Considering the underwater challenge, it's not surprising that, during this time, Farrell took an interest in a request for proposals for a very precise vehicle tracking system. The request came from Caltrans, the state transportation department, and called for a system that could track a vehicle at highway speeds to a centimeter in accuracy.
"Almost everybody was a skeptic\ that such a system could be built, Farrell said.
His work started in this area about 20 years ago, long before people were routinely propping GPS devices onto their dashboards.
In addition to GPS satellites, Farrell's system includes inertial sensors to identify where the vehicle is, and to predict where it will be.
The problem, he said, is that inertial sensors have errors, and the challenge was to find ways to make the effect of those errors go away. Once the device was built, he tried to find a merry-go-round for a demonstration. The idea was to demonstrate the accuracy at various speeds by subtracting the radius of a circle.
But that was in 1996, when cities were doing away with merry-go-rounds in parks because of liabilities. So he took the system to a small amusement park in the area, where the managers cheerfully removed a car from a ride and let Farrell install the 2-foot by 2-foot device for a demo.
Farrell is giving demonstrations now in which the instruments on a car compare the vehicle's position to a database that contains features along the road such as signs and the road's edges.
Using the data from the actual survey of the road, the system considers the car's position, velocity and acceleration to show which lane it's in.
Working on navigational systems was an afterthought for Farrell; initially, he had proposed developing systems to control vehicles. But he shifted the focus to navigation at the suggestion of another researcher.
"They said, 'Throw away this control stuff and do the navigation,' " Farrell says. And now, the bulk of Farrell's funding comes from that area.
Since that first project, he has seen the navigation devices change substantially, from 2-foot-wide boxes on military vehicles to tiny sensors that fit inside cell phones.
"I think the navigation stuff took off because the technology was ripe to go from very expensive military stuff to consumer technology," he says.
These applications work well enough for other researchers to have sent automated cars on long drives across the United States and Europe. Technologies that advance autonomous capabilities are appearing in commercial vehicles.
But on the topic of the driverless car, Farrell is cautious. He believes that the advent of the truly driverless car is far in the future.
"Taking the driver out of the loop is a big liability," he says.