|PI||Tsatsoulis, Costas||Org||The University of Kansas|
|Award #||AIST-05-0029||Managing Center||GRC||Status||Open|
|Title||An Adaptive, Negotiating Multi-Agent System for Sensor Webs|
|The Department of Electrical Engineering and Computer Science of the University of Kansas proposes to perform research under NRA NNH05ZDA001N-AIST. The proposed research develops and tests the technology that allows nodes (pods) in a Sensor Web to collaborate in a rational manner, thus achieving improved sensing through intelligent, informed changes to the behavior of parts of the Sensor Web.|
Our work treats pods as agents in a multi-agent environment, and uses the observations of a pod or of a group of pods to guide future data collection activities of the Sensor Web or of large pieces of it. We develop techniques to identify significant events in the sensed data, that trigger the need to adaptively form pod coalitions and to collaborate for more effective sensing and processing. We also develop task planning behavior, such that pods not only react to the world they sense, but use this information to plan the execution of their behavior now and in the future, and prepare the appropriate pod coalitions. Rational behavior is achieved through negotiation for sensing and processing resources, assuring that pods agree to collaborate only when it improves the utility of the whole Web.
The proposed research involves the areas of multi-agent systems, event monitoring, coalition formation, and negotiation between autonomous agents that leads to maximizing the group utility.
The proposed work is of three year duration (August 16, 2006-August 15, 2009). The entry TRL is 2, and the exit TRL is expected to be 5.