An SIUC computer science professor and his students are developing cutting-edge approaches aimed at helping machines behave in similar, cooperative ways.

Since the 1980s, Henry Hexmoor has looked for ways to translate the complex social and psychological mechanics of human behavior into the binary language that runs computers, and apply these techniques to artificial intelligence and robotics.

The work involves studying social theory and assigning numerical values to what humans would consider "emotions" or "values." Assigning values allows Hexmoor to build mathematical models that relate such concepts to machines.

"There are rules that govern human society. We mimic those in computers," Hexmoor says.

He and graduate students Kishore Thakur and Gaurav Tuli work as a team to program their robots to deal with complex problems by using human ways of thinking. They want to find ways to make multiple machines—called "agents"—work independently but cooperatively to achieve a single objective.

One such experiment, for example, replicates the sidewalk dilemma, where several robotic agents are trying to pass through roughly the same space at roughly the same time. A video of the experiment shows the robots heading toward one another from different parts of a room, but in the last seconds they manage to miss one another.

"We engineered them so that thousands of times per second they look at the situation and make changes," Hexmoor says. "In the end, they come up with a solution that's good for everyone."

Not surprisingly, the U.S. Department of Defense has long held an interest in Hexmoor's work, as it readily applies to the complex issues at work in many theaters of combat where multiple "agents" seek to identify and engage multiple "targets." It also lends itself to military operations that emphasize off-site control of remote equipment, such as Predator unmanned aircraft.

Hexmoor has received about $2 million in research grants over the years. Most recently, he received a $35,000 subcontract from the U.S. Air Force Research Laboratory to examine human control of a "community" of five robots. His students will assist him in writing the software for the project.

His work also is applicable to many peacetime pursuits. For example, he was part of a team working for NASA on "personal satellite assistants," or PSAs. The small machines are designed to fly around a spacecraft, such as the International Space Station, and "sniff" for leaks. The machines work together, communicating in order to triangulate the location of any detected leak. NASA's Ames Research Center is testing the system.

—by Tim Crosby, Univ. Communications