A series of ill-timed accidents could snarl traffic. A bridge could collapse. Construction could shut down a major artery.
The number of things that could go wrong with a city’s road system is so vast, it literally exceeds the number of electrons in the planet.
Decision-makers can take their best educated guesses about how to prepare, but it’s impossible to enumerate the risk for each potential scenario, even with supercomputers.
New research at Clemson University is laying the the mathematical groundwork for a new way that could give decision-makers hard data to consult and help take some of the guesswork out of moving forward, said Cole Smith, chair of the industrial engineering department.
The work could apply to a wide range of military and civilian systems, ranging from the nation’s roads and electric grid to logistics and supply chain on the battlefield.
A team led by Smith has received $384,182 from the Office of Naval Research for the three-year project.
“You want to use modern mathematics to tackle these really difficult problems,” Smith said. “In recent experiments, we’re finding that problems that used to take days and days of computational time to solve we can now solve in about 20 minutes. We did not do that by using a faster computer. In fact, we used a slower computer. We used a laptop. The mathematics are the key, not the computing power.”
What makes the research cutting-edge is that Smith and his team will be looking at the impact of information asymmetry on the problems they are solving. Information asymmetry could include lack of information or misleading information.
“We’re going to be looking at mathematical properties of the problems,” Smith said. “We’re going to try to find anything that allows us to implicitly search all of these solutions. Implicitly means we don’t explicitly look for every single solution. We look at a few possibilities, which we can then prove rules out tons and tons of solutions so we don’t have spend computational time searching for them.”
Smith said the idea for the research came from when he was working for a military organization and he heard a general wonder aloud about why it was always preparing for worst-case scenarios but ignoring all the other things that could happen.
“One of the things I took from that is, ‘Why don’t we look at a portfolio of decisions we can make?’” Smith said. “On one side, if I make a decision, what’s the worst-case scenario? And what’s the average-case scenario on the other side?”
The research the Smith team is doing could help prepare for an enemy attack and other adversarial situations. It could also help decision-makers with design options, such where to put new roads.
“For example, one option may leave you with less risk, but you would tend to get more delays than the average case,” Smith said. “Another would give you more risk in the worst-case scenario, but you might see fewer delays in the vast majority of cases.”
Ultimately, the decisions would be up to the human beings involved.
“We don’t advocate making practical system design based on a computer program,” Smith said. “We advocate giving human decision-makers — teams of people — a portfolio of options. But without the information they’re guessing. They’re making decisions based on preconceived notions. They can fall into one of a million different cognitive traps. We’re giving them the data so they can make intelligent decisions.”
Clemson’s personnel, students and computing power makes the university particularly well-suited for the research. The Palmetto Cluster supercomputer and other high-performance computing resources will help enable the project.
Smith said the mathematical strength he began building as an undergraduate at Clemson helped put him in a position to do the research. He received his Bachelor of Science in mathematical sciences and went on to get his Ph.D. in industrial and systems engineering at Virginia Tech.
Smith was on the faculties at the universities of Arizona and Florida before becoming department chair at Clemson.
For the past decade, Smith’s research thrust has been on interdiction and game theory within optimization, which encompasses the topic of his new research project.
At least two Clemson Ph.D. students will be employed by the project, along with Maj. Timothy Holzmann, a military Ph.D. student who began dissertation work with Smith last fall.
Anand Gramopadhye, dean of the College of Engineering, Computing and Applied Sciences, congratulated Smith and his team on the grant.
“The award is highly deserved and serves as a testament to Smith’s innovative approach to solving complex problems,” Gramopadhye said. “His work is providing excellent research experiences to students and helping keep the department of industrial engineering on the cutting-edge.”
*This article originally posted on the Clemson Newsstand. Read it here.*