City, Microsoft team up to predict bike accidents

The City of Bellevue, Microsoft and the University of Washington are partnering to create a safer city for cyclists and pedestrians.

The City of Bellevue, Microsoft and the University of Washington are partnering to create a safer city for cyclists and pedestrians.

How?

By predicting the future, of course.

The city’s transportation department, researchers for Microsoft and the university, and engineers from Seattle’s Toole Design Group are using existing infrastructure in Bellevue to see if there are patterns or data which can be used from collisions and “close-calls” to help pinpoint where they are most likely to occur. From this data, the entities are looking to prevent future crashes and maybe save some lives.

Franz Loewenherz, senior transportation planner for Bellevue’s transportation department, is taking the city’s lead on the project. He has likened the project to the 2002 science-fiction film “Minority Report,” in which police can predict crimes before they happen.

The start to this Philip K. Dick-like predictive future begins with collecting a significant amount of data.

“We have great data on auto and transit traffic,” he said. “But we have no good data on bike and pedestrian trails.”

The Bellevue Pedestrian and Bicycle Implementation Initiative looked for input through research and public comment to find where issues for cyclists and pedestrians exist in the city.

Despite a reputation as a vehicle-centered city, Bellevue did earn bronze for a bicycle friendly city from the League of American Bicyclists in 2015, and city engineers want to improve upon that with the predictive collision project.

The previous method used to check safety for cyclists is still in place, but it is not quite as effective as officials would like it to be.

“We started digging into the data but we are in reactive mode,” Loewenherz said. “We rely on the data of the past year. We compile all serious injuries and collisions, but these are the ones reported. A lot go unreported. And it’s all reactive data. We were asking “how can we get out ahead of these?””

The project is part of a larger city-wide and national project titled “Safer Streets for Safer People,” which Bellevue’s city government had signed onto previously.

Loewenherz found a similar project used by McGill University in Quebec using GoPro cameras mounted on poles at suspected dangerous intersections. The research team at Microsoft was looking to develop a new program to help in city planning. The groups found each other and a partnership was quickly struck up.

Dr. Victor Bahl, head of Microsoft’s research team, said the project was given an early boost by Bellevue’s infrastructure.

“There are lots of camera in the city of Bellevue,” he said. “Every once in a while someone would call up and say an intersection is very dangerous. Then they had to have someone watch that intersection. It was people watching these feeds and not very effective.”

Bellevue shares hundreds of hours of these traffic feeds with Microsoft, which hopes to use them to learn more about traffic in the city.

Bahl’s team is hoping to create a program or algorithm which can not only electronically acknowledge and separately identify a vehicle, cyclist or pedestrian, but also pinpoint peak traffic flows and collisions or near-collisions.

“We would hope to develop an automated system which can analyze the video and tell us who is using what paths and when,” he said. “We might look at whether a cyclist was using a pedestrian path, or what directions cars were going when they had to brake quickly.”

The project just began this summer, and Bahl remains cautiously optimistic until more data can be analyzed and a possible program unveiled.

“I think this project is extraordinarily intellectually challenging,” he said. “It stresses every component of our system. But it’s very exciting and it’s a worthy cause which could save lives.”

Theoretically, by using data gleaned from the city’s cameras, Bellevue could pinpoint where having bike lanes could be beneficial or where drivers have blind corners.

To Loewenherz’s knowledge, this was the first time such a project on this scale has been undertaken anywhere in the country. Dr. Yinhai Wang, of the University of Washington, was throwing his transportation engineering experience and the weight of PacTrans (a Pacific Northwest traffic consortium) behind the project, Loewenherz said.

“We need to have some knowledge on what exactly constitutes a near miss,” he said. “Dr. Wang brings that to the table.”

David Grant, a public information officer for the city, said although Bellevue was working with leading researchers at the university and Microsoft, the initial capital cost was not a big factor.

“This is a fixed system with cameras which have been around for years,” he said.

Bellevue’s traffic flow cameras can be viewed at http://bellevuewa.gov/trafficmap/. Only two of them are currently being analyzed by Microsoft. That number should increase as the project matures.

“Eventually we can use real-time data to flag near-misses and adjust accordingly,” Loewenherz said. “We won’t have to wait for someone to be injured to make changes.”

While traffic monitors like Raid Tirhi, a senior transportation engineer at Bellevue City Hall, will continue to monitor traffic flows on a daily basis, the new partnership hopes to make streets safer for all traveling parties. Microsoft would like to use the data to expand software to other cities.

“This is capturing imagination in the industry,” Loewenherz said. “And a lot of what we did comes from best practices elsewhere. Everyone is building off of everyone else. It’s an exciting time.”