November 24, 1999
When it comes to heavy Seattle traffic, ‘fuzzy logic’ smooths the flow
Sure, Seattle was ranked third earlier this month among U.S. cities for the amount of time drivers spend stuck in traffic. And just this week, the intersection of Interstate 5 and I-90 was named one of the worst in the nation. But next time you’re suffering oh so slowly through the rush-hour mess, ponder this: It would likely be a lot worse if a new “fuzzy logic” regulating system weren’t in the driver’s seat.
Although it sounds imprecise, fuzzy logic can deliver exacting answers about the ever-changing status of area freeways to help move traffic more efficiently, according to researchers at the University of Washington. After several months of testing at selected on-ramps last spring, the system performed so well that it now dictates traffic flow on all 126 of the metered freeway on-ramps in the Seattle area. And UW researcher Cynthia Taylor said she’s getting calls from traffic engineers around the country interested in learning more about her method.
“There seems to be a lot of interest,” said Taylor, a research engineer with the UW’s Department of Electrical Engineering. “We are very pleased with how well it seems to be working.”
Taylor, working with Deirdre Meldrum, UW associate professor of electrical engineering, designed the new fuzzy logic algorithm that has replaced the old “threshold” method that regulated the metered freeway on-ramps. The metered ramps feature traffic lights to control how quickly vehicles leave the ramps to merge with mainline traffic. The old method worked on a sort of “yes-no” scenario, according to Taylor. It generally wouldn’t take action until after traffic had reached a threshold limit. By that time, a problem had often already developed and would persist for hours before it could be worked out.
Fuzzy logic, on the other hand, uses smooth, continuous control to prevent or delay congestion.
“It uses linguistic variables and rule-based logic,” Taylor said. “Because it is similar to the way we talk and think, it is easier for us to adjust.”
One key fuzzy logic feature is its ability to balance conflicting objectives. That’s a primary issue in managing traffic. An effective system must be able to strike the precarious balance between keeping freeway traffic flowing and avoiding long lines of vehicles on the ramps, waiting for a chance to merge.
Fuzzy logic can also operate with incomplete data, making decisions based on what information is available. That’s critical, says Taylor, because up-to-the-minute traffic information can be spotty.
“We may not get data because of construction, communication problems and hardware failures, and data that we do get may be innaccurate,” she said.
In addition, the fuzzy logic algorithm is forward-looking. Information from each on-ramp site is considered with information gathered from known trouble spots farther down the road. As problems begin to develop, fuzzy logic anticipates them by taking action upstream to mitigate the trouble.
Such characteristics set the stage for an impressive showing when the fuzzy logic method was tested beginning in March. Researchers alternately used fuzzy logic and the old system at two sites – Interstate 90 westbound from State Road 900 in Issaquah to Eastgate in Bellevue, and Interstate 405 southbound from N.E. 160th Street in Bothell to N.E. 72nd Place in Kirkland – to make a comparison.
On I-90, the fuzzy logic system produced an 8.2 percent reduction in freeway congestion – significant enough to be noticeable on a day-to-day basis. It also prevented a bottleneck near the Eastgate on-ramp, a task at which the old algorithm failed. Results show that traffic flow was 4.9 percent better with fuzzy logic.
On I-405, fuzzy logic produced slightly higher freeway congestion and slightly higher flow. The trade-off was that lines at the on-ramps were much shorter using fuzzy logic. According to test data, unacceptably long on-ramp lines occurred nearly twice as often using the old system as they did with fuzzy logic. So fuzzy logic significantly decreased the ramp queues while maintaining mainline efficiency.
Numerous factors, including weather, accidents and daily variations in traffic, make an exact comparison difficult, Taylor said. But the test results indicate that, as a whole, fuzzy logic reduces total travel time and increases flow. It’s also very easy to adjust, so traffic engineers can readily fine-tune it to meet current needs – a big plus in a freeway system as large and variable as Seattle’s.
As a result, transportation officials opted to begin implementing the fuzzy logic algorithm systemwide after test results were available at the end of the summer.
Taylor said she believes the key to fuzzy logic’s success lies in its flexibility – its ability to cope with imperfect input and adapt as the situation changes.
The freeway system is chaotic and non-linear, she said. Accidents, special events and bad weather make sudden changes in traffic more the rule than the exception. That underlines the need for a system that can respond to a wide variety of conditions.
“It’s really more complex to be vague,” Taylor said. “This algorithm is able to use that vagueness to achieve a precise answer. By considering shades of gray and all factors simultaneously, you get a better answer, one that is more suited to the situation.”
###
For more information, contact Taylor at (206) 440-4459, (206) 543-6829 or taylorc@isdl.ee.washington.edu.
