AASHTO Journal - Knowledge Session: Examining Role of AI in Transportation

A variety of insights into the current and future role of artificial intelligence or “AI” in the transportation sector were shared during a knowledge session held during the American Association of State Highway and Transportation Officials 2024 Spring Meeting in Madison, WI.

[Above photo by AASHTO]

Moderator Bill Panos – senior director of engineering software provider Bentley Systems – noted during the session that a lot of this discussion about AI in transportation is focused on “human centrism” and understanding how AI focuses on people.

“You know, AI is not new to us,” noted Panos, who previously served as the director of the Wyoming Department of Transportation and North Dakota Department of Transportation.

“We’ve been using it for about a decade in our systems and it powers many of the things in your daily life, so there is familiarity with it – one has to point it out sometimes where AI is concerned,” he said.

“But to maximize the benefits of AI, we need both the private and public sectors to work together. It’s important to remember that it is a tool; a tool to help us solve very complex problems,” he said. “The development of AI has been accelerated by the availability of data, more complex algorithms, and greater computational power. Not in my lifetime have we seen all three of these things come together.”

Yet while AI systems can handle more extensive and complex data sets, as with any new technology, there’s also critical challenges that must be addressed, Panos noted.

“Issues related to data privacy, cybersecurity, ethical considerations need to be incorporated into AI solutions to ensure that the benefits of the technology are realized while minimizing potential risks,” he stressed.

Rob Cary with Deloitte Consulting explained that he is a “huge fan” of the potential for AI, especially in the state department of transportation community, because of its ability to turn data in actionable information.

“[AI] also has this huge potential to do several things, the number one being to eliminate dull, dirty, and dangerous jobs,” noted Cary, who spent 30 years with the Virginia Department of Transportation.

“If you are in accounting and you’re entering data repeatedly into one system or another, this is a perfect use-case for some automation to come into play,” he said. “Then think of dangerous jobs. Think of drones and robots that can enter confined spaces. Now you don’t have to worry about a human being in there anymore.”

But perhaps the biggest benefit is AI’s ability to generate actionable information to improve decision making and provide deeper insights to management.

“That also allows for proactive actions,” Cary said. “A good example that that we’ve been working on is predictive crash clearance times. Maybe you have 20 years’ worth of data about a particular highway area. Based on certain crash characteristics extracted from that data, are you looking at a two-hour event where maybe setting up a detour is not going to be possible? Or is it an eight to 10 hour where we can set up the detour? That’s an easy use case.”

Dara Wheeler – division chief for research, innovation, and system innovation for the California Department of Transportation said her agency has issued two contracts involving AI; one focused on improving safety for vulnerable road users or VRUs and the other for improving traffic mobility.

“We’re looking at AI to facilitate the anticipation of high risk areas and the generation of a roster of empirically validated safety measures to mitigate the potential for accidents,” she explained. “This was an opportunity to bring in a new tool to get new data sets when it comes to VRU safety and traffic mobility. Both of these use cases are now under contract and are being tested right now for six months. Because of the confidential nature of process, we will not know the results until September this year.”

Kristina Miller, AI and automation branch manager for the Texas Department of Transportation, discussed what she called the “democratization” of AI technology.

“What do I mean? It is about making the capabilities of those [AI] tools and technologies available to everyone, regardless of their skill level and knowledge of AI,” she explained. “There’s a lot that of responsibility entailed in doing that, so, within democratization, how will you accept those risks? How will you apply governance in your enterprise wide AI tools, while also trusting your workforce to be users and customize those AI tools that fit their needs? There’s a lot that goes into that.”

Miller believes one approach is by offering broad AI training and upskilling to the state DOT workforce.

“We developed a community of practice for our workforce, making them aware of where we’re going from a governance perspective and also from an operational perspective,” she said. “Also, we didn’t create a brand new governing body or governance framework. We applied AI considerations and risks to our existing frameworks.”

At the root of that and other AI-related issues, contends David Noyce – executive associate dean for the College of Engineering at the University of Wisconsin-Madison – is a single word: Trust.

“So trust is a field of science. I have a colleague at the University of Wisconsin who’s built a career on studying trust, such as trust and technology, trust and information, and so forth. And it’s extremely important,” he explained.

“So one of the things we must talk about is how do we build that trust? How do we get people comfortable with using AI? This is where partnerships between state DOTs and academic researchers comes in,” he said. “I know this sounds like a selfless plug. But it’s never been more important to have a state DOT-research university partnership as we move forward.”

That’s because Noyce stressed that the development in AI is happening rapidly and it is the job of academic researchers to help advance and develop that technology so state DOTs can use it to improve things such as traffic safety, workforce safety, etc.

“There’s many challenges that go along with it, such as cost,” Noyce noted. “So it’s about using AI in a way that’s beneficial, but in a way that’s also in a usable and correct format for you as a state DOT to move forward with.”