Predicting Where the Potholes Will Be

Over the past three years, City of Edmonton inspectors and work crews have identified and filled about 1.7 million potholes.

That’s right — 1.7 million. That’s more than one pothole per Edmontonian. That’s a heck of a lot of complaints that inspectors have to investigate — and what seems to be endless work for the crews. But, unless we find a new way to pave our streets, the pothole problem is not going to go away. We have a lot of freeze-thaw cycles in Edmonton — which is why we have potholes. Water seeps in through the cracks, expands when it freezes, and destabilizes the pavement.

In fact, if anything, the pothole problem will get worse. With our rapidly increasing population, more streets are being mapped out. New neighbourhoods are being built. And, that means, more pavement that will eventually need to be fixed.

Meanwhile, the City’s resources are tapped out. The City has to look at fiscal restraint.

So, here’s the story of how a co-op student named Dawu Liu helped the City’s Data Science and Research Team change the way the City responds to pothole complaints — and to incorporate artificial intelligence to predict where the next pothole outbreaks will be found.

“I think it’s really amazing to highlight the fact that all of this return comes from the result of a $25,000 investment in a student,” says Kris Andreychuk, an Edify Top 40 Under 40 alumnus and head of the Data and Research Team. “Because of that student, we have this product.”

So, how does it work? Well, before the use of this program, inspectors would have to respond to complaints on a case-by-case basis. A citizen called with a pothole complaint, and someone from the City would come out, have a look. The pothole information would be entered into lists and tables, and out of that, supervisors and crews would try to figure out which problems needed to be dealt with first.

And how long does it take to fix a pothole? It depends.

“It depends on size, it depends on the severity, it depends on the location,” says Andrea Belous, the general supervisor of the City’s Infrastructure Maintenance, Planning and Monitoring department. “But, there’s a lot of them. The biggest factor is that we’re a winter city. If we lived in a warm climate where the temperature was stable, we wouldn’t run into these issues. But we have extreme temperature fluctuations.”

So, the app that was created — and one that City staff can access — has three major functions.

The first one is simple: And that’s to collect all the pothole data and place it in one easy-to-access spot. It also helps the inspectors visualize where the complaints are. Instead of going through a number of hotspots on a list, they’re visualized on a map.

The second is the removal of duplicated complaints. Belous said that, of a sample size of 7,500 complaints, staff found that 1,500 were duplicates. So, inspectors were being sent out to look at potholes that other inspectors had already visited. By simply cutting out the duplicates, efficiency would be improved by 20 per cent.

The third bit is where it gets really interesting. Seventy-eight weeks worth of pothole data was used to teach artificial intelligence on how the next pothole hotspots can be predicted.

“We’re predicting one week ahead where we’re likely going to see future complaints,” says Andreychuk. “The idea is that the resources are stretched, but if and when the time allows, we can think that we’ve got some capacity, and we really should be looking at this area, because we’re likely going to see extra potholes here.

“We’re just using the past to predict the future.”

Both Belous and Andreychuk believe that we’ve just begun to see what this app can do to help make the City more efficient when it comes to infrastructure fixes. They think it can be used to eventually prioritize things like sidewalk fixes.

There are millions of reasons to get this right. It’s a problem with lots of depth to it.