Daniel Olejarz, Transportation Planner at IBI Group and graduate of U of T’s MASc program, presented “AI in Transportation – Industry Trends and Opportunities” on February 5, 2021.
Through data and machine learning, people can train AI to perform tasks in the transportation industry. However, AI is only as good as the data powering the learning process.
Conventional data collection for traffic operations has one person manually counting. This produces low sample size and high variance data which is not well suited for AI applications. With technology development and emerging data sources, there is improved data collection through video detection, infrared counting, and bluetooth detectors. These methods rely on sensors, which can be affected by weather and wear from the environment.
As companies produce more data, the need for people who know how to work with data grows. Several government departments already have big data groups, such as the City of Toronto’s Big Data Innovation Team. However, the burden of proof for AI effectiveness is high for government agencies, where they need to prove accuracy and understand bias. Government officials want to be early adopters, but need to feel comfortable on technologies in a council report, or when talking in front of the city council and mayor.
Olejarz explains that attitudes toward AI in transportation differ in industry vs academia. Some examples are shown below.
| Attitudes toward AI in transportation | |
|---|---|
| Industry | Academia |
| Risk averse | Risk tolerant |
| Cost sensitive | Less cost sensitive |
| Established processes | Ad-hoc processes |
| Competing interests | Single objective |
Despite these differences, both industry and academia include technically competent people who desire to experiment and innovate.
Examples of current AI application in transportation include inclement weather response, and incident detection and response. For weather response, hyperlocal forecast, real-time traffic data, and telematics data are used for road condition predictions, road treatment suggestions, and snow plow dispatch. In addition, video cams on highways feed data to the AI for incident detection and prioritization to quickly restore the network to full capacity. This vigilant monitoring, especially in remote areas where resources are limited, can be a vital safety measure.
Although people have been continually improving AI capabilities, Olejarz notes that there are many jobs, such as decision-making in traffic operations, that still depend on people. Nevertheless, the demand for transportation data skills is growing.
The use of AI in transportation started in big cities and is now being adopted by smaller regions. With this increasing focus on technology, other players besides established transportation organizations now entering the market and shaking up the industry.
Daniel Olejarz is a Transportation Planner at IBI Group, a graduate of U of T’s MASc program, and former President of U of T ITE. Daniel’s graduate research focused on last-mile parcel delivery using autonomous ground vehicles. At IBI, he provides consulting services related to emerging mobility technologies, active transportation, and transportation data analytics.
Presented by University of Toronto ITE Student Chapter.