In Canada’s biggest cities, drivers spend 40, 50 or even 60 hours a year in traffic delays, according to the INRIX 2023 Global Traffic Scorecard. But the impacts go beyond frustrations behind the wheel.
In Toronto — ranked one of the worst cities in North America for traffic — congestion costs billions of dollars annually in wasted time and fuel. It drives up the costs of goods and causes more collisions. Stop-and-go traffic also churns out 200 percent more greenhouse gas emissions compared to free-flowing conditions. And it generates more air pollution, contributing to everything from asthma to heart disease.
As the co-director of McMaster University’s AI-Enhanced Mobility Lab, Hao Yang is turning to artificial intelligence to tackle the issue. “We want to resolve those congested points and make the whole road be more fluid,” says the transportation engineer.
AI forecasting that anticipates bottlenecks before they happen
Currently, apps like Google Maps use data from smartphone users on the road to advise drivers about real-time traffic conditions. But Yang wants to predict future traffic jams.
That’s why his team is developing an AI-powered platform that can use data from a variety of sources — traffic cameras, connected vehicles and third-party providers — to anticipate conditions. That includes predicting normal traffic flows over the course of a day. It also means detecting collisions, construction and other issues and predicting their impact.
The researchers began by modelling the transportation network of Hamilton, Ont. They then fed in minute-by-minute traffic data from third-party providers and simulated various road events to train their machine-learning models.
Producing those robust, city-wide simulations requires serious computing power to calculate the effects of everything from weather to pedestrians to the randomness of human driving habits. That’s where a CFI-funded AI server and advanced traffic simulator proved absolutely crucial for Yang’s team, allowing them to collect and crunch huge amounts of complex data.
“The transportation system is very complicated,” Yang explains. “The impact of the events at one location to the nearby regions, to the entire network, is really hard for us to model.”
Smarter, safer, smoother transportation systems
Although every city has its own unique conditions, Yang aims to develop a core set of AI algorithms that can be applied anywhere in the world to optimize traffic flow. For example, they’ve already extended the platform to Melbourne, Australia.
The resulting insights can then be used to improve traffic management in a variety of ways — for instance, by adjusting traffic signals at intersections or updating digital message boards on the highway. Or it could provide alerts and suggested routes to drivers of connected vehicles, helping them avoid jams up ahead.
But Yang predicts the biggest impact will happen when a critical mass of vehicles has autonomous capabilities to use the AI-powered data. Reaching that point will allow vehicles to make split-second adjustments to their routes, speed or distance between other drivers to keep traffic flowing smoothly.
“I believe this will be one of the most efficient ways to improve the transportation services,” he says. “By controlling or regulating the behaviours of those vehicles, we can improve the mobility of the entire network.”
It takes a lot of computing power to model transportation systems. Without CFI funding, we couldn’t run these city-level studies.
– Hao Yang, McMaster University
