Information fusion

UNO

A handful of remote-controlled, diesel-powered trucks about the size of Tonka toys; some rugged terrain and scaled-down buildings to hide them behind; a radar system; and several surveillance cameras. When Xiao-Rong Li talks about the new laboratory his team is building, it sounds a bit like a 10-year-old's dream sandbox.

The reality, of course, is much different: when the lab is complete about a year from now, it will become a proving ground for the latest advances in information fusion, a fast-growing field that Li, a University of New Orleans research professor and chair of the department of electrical engineering, has big ideas about-and big plans for.

On the most basic level, the term "information fusion" describes the process of combining multiple sources of data and making them work together toward solving a given problem. Human beings are pretty good at this: When you see an object in the distance, for example, your mind registers its color, speed, shape-even the noises it makes-and then fuses that information with data from your memory banks, all in order to make an identification.

Teaching computers how to make these sorts of decisions, however, is a more difficult challenge, and that is where Li comes in. While computers are terrific at crunching numbers and other known quantities, they aren't very good at grappling with situations in which exact values aren't known. If you don't know the exact speed of the object on the horizon in the example above, you can make an educated guess in order to arrive at an identification. But can a computer be taught how to estimate, and then fuse, information-especially information from disparate sources-to arrive at a conclusion?

"Uncertainty is the one part we are fighting," Li says. "Humans can arrive at conclusions that override uncertainty, but computers have trouble with things that are hard to quantify."

Li has devoted his research-and considerable energy-to this information fusion problem and a host of others, both practical and theoretical. A little more than a decade after earning his Ph.D. in electrical engineering from the University of Connecticut, he has published a book that colleagues in the field refer to as a "bible," developed a "seminal" theory of estimation fusion-the problem above-and still found time to produce several book chapters, two graduate texts, some 190 articles and papers, and co-found the International Society of Information Fusion.

Within the broader field of information fusion, one of Li's specialties is target tracking, a problem that the military is very much interested in. If a soldier sees a vehicle moving toward him, for example, how can he fuse together what he sees and hears with satellite images, intelligence reports-a chaotic flow of information-and use it to identify the vehicle as friend or foe, tank or supply truck? Testing solutions to problems like this one will take place in the lab that Li's team is building at UNO.

"The military has a lot of money, and (military applications) make up the biggest part of the field right now," Li says. But he is quick to note the potential that information fusion has in many other directions as well. Office automators and business schools stand to benefit from advances in the information fusion field, as do the makers of Geographic Information Systems (GIS) devices. The medical applications, meanwhile, are vast, from turning two-dimensional images of the body into 3D images, to helping doctors fuse a combination of medical experience, symptoms and lab results into more accurate diagnoses.

Even as Li works on practical problems like these, he is also searching for ways to separate his field from the others, like electrical and computer engineering, that it is often lumped in with. To that end, he spends a great deal of time thinking about theory, searching for a principle that will capture the gist of the information fusion problem, and thereby define the field.

At the same time, Li is working to raise UNO's profile in the field. UNO's electrical engineering department has devoted four faculty positions-one-third of a small department-to study information fusion. Three of Li's Ph.D. students are also researching fusion problems. That degree of focus has put UNO on the information fusion map.