By Emily Jankauski
Natural disasters happen when you least expect them. Their impact is sporadic, seemingly affecting individuals at random. But what if you could predict the future?
That is the feat Hadi Meidani is attempting.
The University of Illinois at Urbana-Champaign (Illinois) Department of Civil and Environmental Engineering assistant professor is hoping machine learning ― the ability to analyze data using algorithms — and artificial intelligence (AI) ― a computer’s ability to perform human-like tasks, such as speech recognition or translation of languages ― are just the ticket for solving traffic prediction in those life-changing moments.
“The better we can predict and estimate the complex behavior of infrastructure systems, the better we can manage them, optimize them and be prepared for things that could happen in the case of abnormalities like natural disasters,” Meidani said.
The inspiration for his vision?
Meidani credits his upbringing in Tehran, Iran, where he saw new tower construction “on a daily basis.”
Well, that and his interdisciplinary approach to research, which he attributes to his studies at the University of Southern California. Here he earned master’s and doctoral degrees in civil engineering as well as a master’s in electrical engineering. But Meidani also holds a master’s in structural engineering from Sharif University of Technology (Tehran, Iran).
“The degree (in electrical engineering) allowed me to have the right foundation to tackle problems from an interdisciplinary angle and also have the confidence to communicate with other disciplines and be exposed to research in other domains,” he said.
Meidani’s doing just that by taking those technologies, such as machine learning and AI, that are traditionally housed in computer science and applying them to civil engineering problems.
“A lot of those problems (in computer science) were computer vision, image classification, language processing,” Meidani said. “(They’re) very different applications than what we are dealing with in civil engineering, which are physical systems like traffic flow.”
The biggest hurdle for Meidani?
The “governing physics” behind how these systems behave and operate.
“We are now trying to see how we can adjust the AI and machine-learning tools that are out there for systems that are physics-based or governed by physical laws,” he said.
“We have used machine-learning tools in order to speed up the computer simulation of mechanical systems,” he added. “And we have also used machine learning to predict the traffic conditions and how you can, for example in the case of a natural disaster, use these techniques to estimate missing or unavailable data.”
So if sensors along the highway or sensors in an autonomous vehicle fail to communicate information, Meidani hopes machine-learning will step in to estimate the necessary data and help predict the flow of traffic.
“There could be wild fires, there could be flooding and you may not have access to those sensors,” he said. “But nonetheless, since in the future smart cities will be increasingly dependent on sensor-based management of infrastructures, we can still use some approximation for missing data in optimizing our decisions.”
While preparation is key, these days, Meidani’s hopping into action working on a proposal to optimize COVID-19 community screening by attempting pool testing.
“(It requires) mixing samples from several people and testing the mix once for faster overall assessment of the (virus’) spread,” he said.
But in the meantime, Meidani hopes his natural disaster preventative measures will save resources and allow civil engineers to be at ease and not be “caught by surprise” if something goes awry.
“If we have a rigorous, accurate way of quantifying the uncertainty and (we’re) accurately predicting the complex behaviors of civil systems, then we can better manage their operations,” he said.
That drive and passion fuel Meidani to better the future, especially in his current role at Illinois.
Photo: Hadi Meidani, right, was all smiles during the 2019 Transportation Research Board Conference in Washington, D.C.