A communication choreography for discrete step multi-agent social simulations

Considerable research has been done on agent communications, yet in discrete step social agent simulations there is no standardized work done to facilitate reactive agent-to-agent communication. We propose an agent-to-agent interaction framework that preserves the integrity of the communication process in an artificial society in a 'time-stepped' discrete event simulator. We introduce the modeling language called Agent Choreography Description Language (ACDL) in order to model the communication. It serves in describing the common and collaborative observable behaviour of multiple agents that need to interact in a peer to peer manner to achieve some goal. ACDL further adopts the parallel and interaction activities to model proper communication in an artificial society. The ACDL communication framework is implemented and tested in REPAST. It employs a communication manager to generate and execute ACDL specification according to agent's communication needs.