Bellevue Using AI to Identify Unsafe Roads

Depiction by Emily M. Eng, Seattle Times.

Depiction by Emily M. Eng, Seattle Times.

Why wait for people to get hurt to make roads safer? Technology can identify dangerous roads before car accidents happen.

The right to safe roads

Here in Washington, drivers, bikers and pedestrians enjoy the right to reasonably safe roads. That includes not just roads free of dangerous potholes or other defects, but roads that are not improperly designed to create danger.

Especially old roads are sometimes designed with dangers: narrow lanes, confusing or barely visible signage, no safe pedestrian area, or no median barrier. In other situations, roads designed to support a smaller traffic flow become dangerous with significantly more traffic or higher traveling speeds.

Cities, counties, and the state have a legal duty to keep roads reasonably safe for the traveling public. Part of how they do that is by tracking car accidents through collision reports. They use car accident data to determine which roadways warrant greater or different safety measures such as a sign, crosswalk, or speed limit.

The problem with this method is that they only consider reported car accidents, and not unreported accidents and near misses. Bellevue is attempting to solve this problem, and identify unsafe roads before people are hurt, using artificial intelligence.

Modern technology to make roads safer

Per this article by Michelle Baruchman of the Seattle Times, Bellevue is using AI to monitor and process traffic camera footage around the city. As the article explains:

“Currently, jurisdictions rely on police-reported events to inform where they make improvements. That’s problematic. Those crashes are rare and random, and it often takes five-plus years for a pattern to emerge. During that time, you have all of these people being injured,” said Franz Loewenherz, principal transportation planner with the city of Bellevue.

Bellevue’s one week study found 20,000 “critical conflict interactions,” or near misses. “The analysis of 5,000 hours of traffic-camera video and millions of trips by car, bike and on foot found that near misses in which crashes nearly occur are a reliable predictor of where future collisions will happen.”

Bellevue is conducting this research in partnership with Transoft Solutions and the nonprofit Together for Safer Roads. Transoft is a global company. This technology is available not just to Bellevue, but any sizeable city.

Because near misses are so frequent and are a reliable indicator of future collisions, moderate to large cities should be adopting it, rather than waiting for people to get hurt to identify unsafe roads.