TRUCE: Agent Coordination Through Concurrent Interpretation of Role-Based Protocols

Established protocols for coordination are essential for implementing joint-action activities among collaborating software agent. Most existing agents, however, are designed only to support static protocols, which limit their interaction domain to specific sets of agents. We develop an agent collaboration framework for open systems that enables an agent to expand its acquaintance set and to adapt to various coordination protocols dynamically. This is achieved through writing coordination scripts that are interpreted at collaboration time. We developed a role-based coordination language for writing these scripts, where the coordination mechanism used is the concurrent interpretation of a single script by the participants of a given collaboration. This interpretation defines the behavior of every agent. Thus. their individual interpretation may differ depending on the roles that were assigned to them. This new coordination language provides various coordination primitives in which the basic synchronization is achieved via distributed rendezvous points. In this paper, we present and demonstrate the elements of the TRUCE language