Research Of Multi-Agent Interactive System Based On Argumentation

In multi-agent systems, the information consistency problem and conflict of interest aredifficult to avoid because of the uncertain, inconsistency and incorrect information. In order to solvesuch problems, the interaction between agents are required to not only send messages but also whythere messages are appropriate in order to modify the agent’s belief, desire and intension andmaintain the consistency of the knowledge base of each agent. Such interaction is a kind ofargumentation-based interaction. This paper proposed a multi-agent argumentation framework bycombining the argumentation theory and the multi-agent systems together. The argumentation theoryof this framework is based on a model of argumentation. The basic unit of the argumentation modelis composed of a number of premises and a conclusion. It combines the Toulmin model and Dungframework together. And it not only represents the internal structure of the argument but alsorepresents the relation between the arguments. Firstly, we describe the multi-agent argumentationframework formally and we propose the multi-agent system is made up of argumentation agent, themediator agent\the expert system agent, the inference engine and the knowledge base. Then we useEclipse as the development environment and build the JADE platform in it. Next we create theargumentation agent,the expert system agent and the mediator agent. And implant the argument intothe FIPA ACL agent communication language. And establish the knowledge base with tuPrologengine. And implement the argumentation algorithm with Prolog programming language. Finally, weuse an example that a proponent agent and a opponent agent debate around a argumentation topic,and we get the acceptable argument set by the argumentation algorithm. The result show that thismulti-agent argumentation system are able to deal with conflict resolution and maintain theconsistency of the knowledge base.