The award supports research into a methodology under which autonomous agents can develop commonly shared rules of interaction and commonly shared communication languages, while interacting in a multi-agent environment. Interaction and communication are viewed as agents' attempts to maximize their performance, measured as expected utility of outcomes. The research aims to enable agents to benefit from rules of behavior that are shared and that could emerge among interacting agents without the necessity to recompute currently optimal courses of action. The research also seeks to enrich the agents' communicative abilities by formalizing the role and usefulness of shared communication languages. The continuation of this thread is the precise formalization of a mechanism of language formation. The most significant impact of this research will appear in multi-agent scenarios characterized by unanticipated change, such as those in hazardous environments, distributed management of communications networks, and flexible manufacturing systems.
