This research investigates how cooperating agents in real-world domains can select the most useful information to communicate to others. The research focuses on communication between agents in distributed systems performing complex, problem-solving tasks. Here, only partial knowledge of other agents is available, and agents have limited computational resources for communication. To address these constraints, the researchers will examine how an agent's knowledge of its own use of information can indicate the value of that information to other agents. Heuristics will be developed to evaluate the usefulness of a given piece of information based on how that information is used or inferred as an agent performs its problem-solving task. These heuristics will be integrated into the communication system for AUV's that is being developed at the University of New Hampshire. The research will contribute to the ability of AUV's to work together, as well as providing insights into how other distributed systems and natural language interfaces can select useful information to communicate.//
