University of Minnesota Driven to Discover
U of MNUniversity of Minnesota
Center for Transportation Studies

Ramp Metering for Postponing Freeway Breakdown

Presentation by Lily Elefteriadou, Director, Transportation Research Center, and Professor of Civil Engineering, University of Florida

October 22, 2009

As cars queue up at ramp meters during rush hour, the timing of the meter’s signal depends on capacity—a calculation of how many cars the road can bear. Early in rush hour, cars are released quickly. Later, at peak flow, they are released more slowly so the road doesn’t exceed a set number of cars.

But what if traffic is moving smoothly, despite increased density? Could more cars be released, reducing ramp waiting times without causing a traffic breakdown? A new study, led by Professor Lily Elefteriadou at the University of Florida, looked at whether using real-time traffic congestion data in ramp meter decisions could delay breakdown while increasing freeway capacity. Elefteriadou presented the study’s results at an ITS Institute Advanced Transportation Technologies Seminar on October 22.

The study takes into account one of the underlying rules of traffic: that it’s a random system affected by hundreds of small decisions and factors, ranging from a stalled car to weather to a platoon of vehicles entering a crowded freeway.

“The main issue is that breakdown doesn’t happen at the same flows or same combination of flows every day,” Elefteriadou said. “Ramp metering is based on capacity, the maximum that can be sustained at a particular facility, but that maximum varies by quite a bit from day to day.”

Using a year of sensor data gathered at six sites, including a stretch of Interstate 494 in Minnesota and the Queen Elizabeth Expressway outside Toronto, Elefteriadou’s research team calculated the likely time of traffic breakdown in key bottlenecks.

The researchers defined traffic breakdown as the beginning of recurring congestion on a section of road. Their initial analysis showed the best indicator of traffic breakdown was a drop in speed—usually of 10 mph or more. Once they developed a breakdown probability model, they used a traffic simulator to see how the model worked at the Ontario and Minnesota sites.

Both sites presented challenges. In Toronto, congestion started at one ramp one day and at another the next day. At the Minnesota site (a stretch of 494 between Bass Lake Road and Rockford Road), researchers found that when they tried to reduce congestion at a downstream ramp, the traffic queue at an upstream ramp jumped.

“You’re always playing a balancing game between trying to avoid congestion on the freeway and not extending the queue too long on the ramp,” Elefteriadou said. Still, results showed the new model was able to postpone traffic breakdowns during peak periods. When congestion occurred, it didn’t last as long. And total travel time for drivers was reduced. At the Minnesota site, the model showed breakdown was postponed for several minutes.

Because the calculations are data-intensive, the model is more likely to work on sites that have historic bottlenecks occurring in the same location at the same time. Still, it is a step forward from one-size-fits-all ramp meter decisions, Elefteriadou said. Now that researchers have established a model, it should be easier in the future to apply it to new sites and even to look at underlying factors causing breakdowns.

“It shows some promise,” Elefteriadou said. “Particularly at locations with adequate local data for breakdowns.”