How artificial intelligence is helping urban planners

From housing demand to traffic congestion, city planners have long relied on imperfect models to predict how people move and live. Now, artificial intelligence promises to make those forecasts more accurate.

Mohamad Khalil, a transportation engineering researcher who is currently a postdoctoral fellow at the University of Alberta, has been interested in machine learning long before it was trendy.

“I actually started working with machine learning in 2014,” he says. “To me, at the time, it seemed very appealing and very futuristic and an important next step to [urban planning] modelling.” Mr. Khalil says that most urban planning models, which were developed 50 or 60 years ago, are overly simplistic.

“It assumes that humans are 100-per-cent rational and will choose the best option for their own sake,” he says. “For example, you will choose the best route to go to work based on travel time. However, this is not 100-per-cent true. Sometimes, for some reason, you might choose another route.” In comparison, machine learning models are able to use countless data points collected from devices such as mobile phones and vehicle GPS systems to create more complex models, faster and with greater accuracy.

“Machine learning excels with complicated behaviour,” he says.

Mr. Khalil, who conducted his PhD thesis in transportation engineering at the University of British Columbia, built a “modelling suite” – a collection of tools that help simulate and visualize different land-use scenarios using AI to make better predictions that factor in changes across urban, transportation and demographic systems.

“If we’re implementing a policy, how is that policy going to affect a city?” Mr. Khalil says. “If 20 per cent of people are working online compared to 100-per-cent remote, maybe we’ll see less traffic on our roads, maybe people need bigger homes if both partners are working from home and maybe they don’t need two vehicles.”

He envisions his research and modelling tools being adopted by city planners and decision-makers.

“We present to them the different scenarios that could happen,” he says. In turn, planners can make decisions and recommendations to elected officials about infrastructure development such as zoning, building new transit lines and housing.

By considering multiple scenarios, which can be produced quickly and accurately, urban planners can take a more creative and flexible approach to their work by experimenting with different parameters and possibilities.

That ability to test real-world scenarios before making costly infrastructure decisions is already taking hold in Canadian cities.

Ryan Smith, divisional director of planning and development services for the City of Kelowna in the southern interior of British Columbia, has been using predictive modelling to make more informed decisions and recommendations for nearly a decade, although he says the technology has improved recently to analyze larger data sets.

“We’re flying less blind now,” he says. Kelowna has been an early adopter of such technologies. One example is an AI-enabled predictive modelling tool Mr. Smith uses to see what neighbourhoods are likely to be redeveloped soon. These are typically areas with older homes and buildings that might be demolished soon and rebuilt with additional housing density.

“We know what year a house was built, whether or not the owner lives in the house, the improvement value on the property and how much the building is worth,” he says. “We can create a probability of redevelopment with that data and make smarter infrastructure decisions.”

That might look like curb, gutter, sidewalk and street tree improvements, replacing and upsizing sewer and watermains or improving electrical infrastructure in neighbourhoods that are likely to see a higher rate of redevelopment and therefore an increase in residents. These tools allow planners like Mr. Smith to “get ahead” of risks, such as ensuring neighbourhoods have sufficient infrastructure to support more residents.

But while some planners see clear benefits to integrating AI into city planning, others caution against letting the technology steer too much of the process.

Pamela Robinson, a professor at Toronto Metropolitan University’s School of Urban and Regional Planning, cautions against becoming too reliant on it.

“I would argue that AI could be an input into research and decision support, but it shouldn’t be making the decisions,” Ms. Robinson says. “Planners need to stay in charge and be the humans in the loop around the sound professional advice they offer.”

Ms. Robinson sees the potential for AI to improve city building in several different ways, from expediting the approvals process for issuing building permits to platforms for public engagement and consultation and design decisions, such as what types of cladding on a building have lower greenhouse gas emissions. However, she encourages urban planners, decision makers and elected officials to use these technologies with care.

“I think planners are appropriately curious and cautious, and I think that’s a good thing,” she says. “The planners that we’ve worked with want to deliver good outcomes for their residents and they’re committed to their work and the communities where they’re planners. There’s a lot of hype around these tools. It’s early days and I think this kind of curiosity and caution will serve Canadian cities well.”

*

*