Could artificial intelligence catch distracted drivers in Edmonton? U of A piloting program

A new pilot project in Edmonton is using artificial intelligence to detect drivers who are using their cellphone while behind the wheel.

The three-week pilot, led by the University of Alberta, aims to determine the prevalence of distracted driving in the city and test whether artificial intelligence is effective at catching distracted drivers.

The City of Edmonton and the Edmonton Police Service are also involved in the pilot.

The automated technology will be tested at a few Edmonton locations, using sensors to capture high-resolution images through vehicles' windshields.

Researchers said that in the pilot, someone with a phone on their person — whether in their hands, pocket, or lap — would be counted as a distracted driver.

"[The technology] looks for a sign of the cellphone within the view of the windshield," explained Karim El-Basyouny, urban traffic safety research chair at the University of Alberta.

"If there's a cellphone there and it's moving, that is sort of what it's trying to detect."

El-Basyouny said it can be hard for police to know how prevalent distracted driving is in Edmonton. He hopes the pilot can give researchers a better idea of the size of the issue.

"Right now the only way is to just use manned enforcement," he said.

"It's not very easy for you to detect and … the police need to detect that there is distraction and they need to then pull the vehicle [over] and then issue that ticket."

According to data from the Ottawa-based Traffic Injury Research Foundation, distracted driving fatalities surpassed those caused by impaired driving in Canada in 2019. 

Distracted driving as deadly as impaired driving:

Australian-based company Acusensus developed the AI the university is testing. General manager Tony Parrino said it's been used in Australia and in several locations in the United States.

How does it work?

So how does artificial intelligence catch drivers?

Parrino said his company shows the program examples of distracted driving, and not-distracted driving, and the technology uses an algorithm to learn.

The technology takes a picture of the potential incident on the road, which is then reviewed by a human. If the human decides the driver was distracted, that information would be turned over to police within 48 hours to issue a citation. 

"With distracted driving, you can see the evidence in the photo," said Parrino.

"There's no dispute, it's not like a speeding ticket where you have to determine whether or not the system is working properly. It's pretty obvious that the system is working properly for distracted driving when you could see somebody's phone in their hand."

At the moment, the university is just testing the technology, and those photographed will not be ticketed.