ICE: self-configuration of information processing in heterogeneous agent teams.

Teams of agents, solving complex tasks in dynamic environments, require high-quality information about the current situation. One way of achieving high-quality information is reliable information processing, that is suitable for the application domain. However, the characteristics of some domains such as disaster scenarios are partially unknown at design-time. Therefore, specifying information processing at design-time becomes nearly impossible and leads to unreliable information. We tackle this problem with the ICE middleware which supports adaptive information processing for teams of autonomous agents. It provides a decentralized self-configuration and dynamic integration of information sources. A configuration is created with respect to required information, available sources, and resource constraints. Our evaluation shows that ICE is sufficiently efficient to be operated in highly dynamic domains.