Every time you board a plane, turn on a light, or chat with a neighbor, you become part of a network: the air traffic system, the power grid, the pool of possible victims of a virus.
To Sandip Roy, an assistant professor of electrical engineering and computer science at Washington State University, and his graduate student, Yan Wan (’08 Ph.D.), such networks have a lot in common. They’re all composed of distinct points, with every point connected, directly or indirectly, to every other point. Like a spider web, if you pluck one strand of the network, the whole web jiggles.
By devising mathematical equations that describe the points and routes in a network, Roy, Wan and senior faculty member Ali Saberi aim to predict the “jiggles”–how a network will behave under certain conditions, and how those who manage it might change that behavior. Can we reduce flight delays throughout the air traffic system if we hold planes on route A rather than route B? What’s the most efficient way to distribute power during peak-use times? Is vaccination or quarantine more likely to prevent an epidemic?
“From our perspective, there’s no difference between working on virus spread and control versus working on computer networks,” says Roy. “We’re achieving the same thing [in both cases]. We’re giving engineering insights, system insights, into these domains.”
In the case of the power grid or air traffic system, Roy and Wan look for ways to ease the flow (of electricity or planes) through the network. In the case of a disease spreading, they look for ways to shut down flow, to prevent a virus from, well, “going viral.”
Wan’s model of the 2003 SARS epidemic in Hong Kong found that traffic among the region’s 18 political districts varied widely. So did infection rate. Within each district, people mingled a lot, allowing the virus to spread quickly once it gained a foothold in the population. Between districts there was much less contact. Two districts were of special concern because the high rate of travel between them and other districts gave them enormous influence on the spread of the disease. Wan’s model showed that if public health officials had stopped the disease in those two districts, they probably would have stopped it from becoming a region-wide epidemic.
Wan says her model let her compare control efforts that focused on the points, such as vaccinating everyone in the key districts, with those focusing on the connecting routes, such as quarantining the hotbed districts. Either way, says Roy, a targeted approach would be at least as effective as a blanket approach that doesn’t take the network structure into account–and it costs 20 percent less.
Roy says professionals in public health and in air traffic control have been receptive to his and Wan’s findings. Still, getting established control systems to change is a slow process. Many existing networks include features that cannot be changed, for financial or political reasons as often as for engineering reasons.
“You don’t really want to be modifying [air traffic systems] very quickly,” he says. “It may be 20 years before something like this is fully used, but the concepts are slowly being put forth.”
Roy has only been working on network control and design for a few years.
“My work as a Ph.D. student was modeling networks, just understanding how they behave and the uncertainties in them,” he recalls. Soon after he arrived at WSU in 2003, Saberi offered him some gentle but firm advice.
“He was nice about it. He just said, ‘you’re not really attacking the important problems. You’ve got to look at control.’”
At about the same time, Roy began sparring with Wan, who was a student in several of his classes. “I could not get her to miss an exam problem,” he says. After several tries he finally came up with a problem she couldn’t solve–and she was hooked.
When Wan joined his lab, her interest in biological systems opened a whole new range of projects to Roy. “I feel more comfortable working in that direction with her expertise,” he says. The two have already co-authored more than a dozen journal articles.
Roy says the image of electrical engineering held by the public (and even by some of his collaborators) hasn’t caught up with how the field has changed in the past 20 years or so.
“Some [people] have the perspective, ‘an electrical engineer is someone who builds circuits and [who] solders,’” he says. “I can solder. I have that background. But the heart of electrical engineering nowadays is systems engineering, the idea that we develop methods that work for a range of systems.
“We build toaster ovens, we do things like that, but we also hope that our way of building toaster ovens helps in other types of systems. Networks are one critical family of systems that badly needs new modeling and control techniques.”