Carnegie Mellon University's East End traffic effort turns to pedestrian safety

Share with others:

Print Email Read Later

The East Liberty intersection that prompted a recent rally for pedestrian safety -- at the Target store at Penn Circle and Penn Avenue -- is one of 18 that Carnegie Mellon University researchers have been monitoring using a traffic control system they developed.

Making use of city cameras already in place, CMU's Traffic 21 project piloted its Scalable Urban Traffic Control system -- Surtrac -- at nine East Liberty intersections last year. The city installed the system at nine other intersections, from Penn to Fifth Avenue, last month.

Officials say the system has reduced vehicular emissions by 21 percent and travel times by 25 percent overall. The midday wait time for a green light was almost cut in half. The catch, and potential hazard, is that green arrows for cars light up at the same time the walking figure lights up for pedestrians.

CMU researchers are working on perfecting a pedestrian detection model and have a test mechanism already installed on Penn Circle at Collins Street and at Eastside.

A bus detection model will follow next month at Collins, said Stephen Smith, director of the Intelligent Coordination and Logistics Laboratory at CMU's Robotics Institute.

Components that recognize pedestrians, bicycles and buses are the next step in multi-modal systems for urban transportation planning. So far, foundations have granted Traffic 21 $435,000, starting with seed money from the Hillman Foundation. The Heinz Endowment made it part of its Breathe campaign for air quality, and additional money has come from the Richard King Mellon Foundation.

PG graphic: Smart traffic signals
(Click image for larger version)

The researchers also are working on plans to deploy Surtrac farther east and west, possibly networking with state Department of Transportation radar that's being installed on Penn from Fifth to Braddock avenues.

There are traffic cameras that can detect pedestrian presence, but they can't count the number of people waiting to cross the street. The CMU researchers hope to achieve that kind of precision with their system.

"The state of pedestrian modeling has come a long way," said project scientist Greg Barlow, "but pedestrians are hard to model. Cars are restricted in where they can go but pedestrians can start anywhere and end up anywhere."

On a lab simulator, vehicle units move in two sizes. One size represents cars, the other any large vehicle. One of the students in Mr. Smith's lab is working on a model that would distinguish a bus from a truck, the goal ultimately being to give priority to mass transit.

The system has to be taught to read what it sees. The researchers train the system the way you might a new puppy, giving it positive information so that over time it recognizes a bus the way a puppy learns the difference between "toy" and "ball" by subtlety, not size. Over time, the system recognizes that a bus pulls aside to discharge people. A big rig doesn't do that. It might recognize that every time the researcher told the system "this is a bus," the vehicle was lined with windows.

The pedestrian model is hard to refine because people are unpredictable. If a person jaywalks or suddenly drops to the ground to do push-ups while waiting for a light, the system might not know how to read those forms.

Surtrac at each intersection relays traffic information to the next intersection. This communication enables each one to respond with light changes. The biggest problems are related to weather. Heavy rains and glaring sun can distort its decisions.

"The sun during the morning rush blinds the camera at Penn and Fifth," Mr. Smith said. "That's where radar would solve the problem."

Intersection cameras and underground traffic detection systems have been used for years to regulate lights, most notably on suburban arteries where traffic takes awhile to build up on side streets and exit strip malls.

"Our niche is the urban environment," said Mr. Smith. "One reason we want to continue this work is to get density information."

CMU researchers hope to make Surtrac the state-of-the-art tool for efficient movement of multiple transportation modes in cities and spin it off as a commercial venture, he said.

The Target intersection is problematic because it has five lanes, including turn lanes. Also, Penn Circle South does not meet Penn at a 90-degree angle like streets on a grid. It swoops around, sending cars accelerating onto Penn before drivers can see the crosswalk.

The configuration of lights, including green turn arrows, was engineered by the city. Those who protested for safe crosswalks want a restricted red light to stop all traffic for pedestrians to cross in any direction.

The city has been working with CMU on the project and expects to improve the Target and other intersections "to develop a different product to pay attention to pedestrians, bikes and buses," said the city's transportation planner, Patrick Roberts. "Where we are now is to collect data on how often people are grouped together" to cross the street "and adjust the signals to create the smartest intersection possible."

He said there is consideration of a restricted red at the Target intersection.

"There is a need for pedestrian safety improvements," said Loralyn Fabian, planning coordinator at East Liberty Development Inc. "This is one of our priorities."

She said East Liberty Development is working on a walkability and accessibility study.

"We want to develop pedestrian detection," Mr. Smith said. "It's an extension of the same idea [as for vehicles] -- to move people."

Diana Nelson Jones: or 412-263-1626. Read her blog City Walkabout at

Join the conversation:

Commenting policy | How to report abuse
To report inappropriate comments, abuse and/or repeat offenders, please send an email to and include a link to the article and a copy of the comment. Your report will be reviewed in a timely manner. Thank you.
Commenting policy | How to report abuse


Create a free PG account.
Already have an account?