精确监控城市交通状况

Cities count the number of cars on the road in order to plan everything from the timing¹ of stoplights to road repairs. But the in-road metal detectors³ that do the counting can make errors -- most often by registering that a car is present when one isn't. One common error is called "splashover" because it usually involves an over-sensitive detector² picking up the presence a vehicle in the next lane over -- as if the signal from the car "splashed over" into the adjacent（邻近的） lane.

 

Now Ohio State University researchers have developed software to help city managers easily identify detectors that are prone⁴ to splashover and reprogram them to get more accurate numbers.

 

Benjamin Coifman, associate professor of Civil, Environmental and Geodetic Engineering at Ohio State, and doctoral student Ho Lee describe the software in the October 2012 issue of the journal Transportation Research Part C: Emerging Technologies.

 

For the study, Coifman and Lee monitored 68 in-road detectors in Columbus, Ohio. They found six detectors that were prone to erroneously（错误地） detecting cars in adjacent lanes. Error rates ranged from less than 1 percent to 52 percent.

 

"A host of city services rely on these data. We've known about splashover for decades, but up until now, nobody had an effective automatic test for finding it," said Coifman. "With this software, we can help transportation departments know which detectors to trust when deciding how they should put their limited dollars to work."

 

People may not be familiar with the commonly used loop（使成环） detectors, which are often present at intersections⁵ to activate⁶ a stoplight. When the detectors are visible, they look like rectangular cutouts in the road surface, where underground wiring connects the detector to a traffic box at the side of the road. The same detectors are often present at freeway on ramps⁷（坡道） and exits, to help cities monitor congestion⁸.

 

To see how often splashover occurred in the 68 detectors in the study, the researchers went to the sites, and noted⁹ whether a car was truly present each time a detector counted a car. Then they used those data to construct computer algorithms that would automatically identify the patterns of error.

 

In tests, the software correctly identified four of the six detectors that exhibited splashover. The two it missed were sites with error rates less than 1 percent -- specifically 0.6 percent and 0.9 percent.

 

"We might not catch detectors in which one in 100 or one in 1,000 vehicles trigger splashover," Coifman said, "but for the detectors where the rate is one in 20, we'll catch it."

分享到：