Monday, 25. October 2010 25.10.10 16:40 Age: 7 Jahre

Solving Problems together

By: Katharina Bätz

[Translate to Englisch:] Ihr Spezialgebiet sind Agenten: (von links) Thomas Kemmerich, Michael Baumann und Markus Eberling. (Foto: Nadija Pejic)

Paderborn. Agents are known to computer scientists as units operating autonomously within a system. That they can do much more than work on their own is what Markus Eberling, Thomas Kemmerich and Michael Baumann demonstrate. The three are part of the research group Knowledge-Based Systems at the Computer Science Institute at the University of Paderborn. Their research focuses on coordinating individual agents and getting them to act together.

"The increasing interconnectedness of computers allows us to dissect an assignment, divide it among various computers, and ask them to execute the assignment cooperatively", reports Markus Eberling. "In this way, a single computer can be seen as an independent agent, interacting with other agents." He goes on to mention that robots can also be modeled as agents.

Normally, autonomous agents within a system have a limited local view, which means that they can only perceive those agents and their views that are in their direct vicinity. "An individual agent possibly doesn’t know that the system can consist of many thousands of agents", says Eberling. His goal is to get agents to cooperate in solving a common task, even though they basically act individually and selfishly.

Thomas Kemmerich, doctoral candidate at the International Graduate School of Dynamic Intelligent Systems, is working on the problem of how local knowledge, that agents gather while working, can be effectively used by other agents in the system to develop a solution for a common problem. He is also trying to find out whether agents can learn effective cooperative strategies based on this knowledge.

In complement to this research, Michael Baumann (also a doctoral candidate at the International Graduate School) is developing team strategies for agents. Agents do operate autonomously, but must coordinate their movements in space in order to execute a common task and avoid collisions amongst themselves. '

The practical application of the cooperation and coordination of agents is due to be studied by a computer science students’ project group in the coming two semesters. There will be a simulation of a faraway planet environment that is inhospitable to humans, where robots must mine resources and transport them back to Earth. Because remote controlling the robots from Earth would be too involved, the robots must learn to make autonomous decisions and act accordingly. They must also learn to specialize in certain tasks so that they can solve the bigger issue cooperatively and effectively. The project team will study this agent system using this simulation.

Research group Knowledge-based Systems
05251 60-3360