Muon: designing multiagent communication protocols from interaction scenarios

摘要

Designing a suitable communication protocol is a key challenge in engineering a multiagent system. This paper proposes Muon, an approach that begins from representative samples of interactions or scenarios. Muon identifies key semantic structures and patterns based on (social) commitments to formally analyze the scenarios and offers a methodology for designing protocols that would meet stakeholder needs. Interestingly, Muon applies its formal representations to suggest ways to identify additional scenarios needed to address exceptions arising in the interactions. This paper contributes (1) a conceptual model of message types and causal relationships among them as a foundation for developing commitment-based communication protocols; (2) a robust, reusable characterization of semantic structures reflecting the above model; (3) a mapping from an annotated scenario to causally related interactions; and (4) a methodology to synthesize specifications of communication protocols. This paper reports on an empirical evaluation involving developers creating protocols from two real-life cases.