基于对话的多Agent协作交互模型

针对传统协作设计过程中任务分解、任务分配的随意性和盲目性,提出了一种基于对话的多智能体协作交互模型.初始承担者首先通过信息搜索对话,依次询问设计辅助智能体的专业领域及参与完成设计任务的能力、愿望和机会,以确定其能否作为潜在的协作组成员,然后选用劝说对话,依次劝说潜在协作组成员把实现设计任务作为其意图,最后再次以劝说对话的形式,使潜在协作组成员意识到组内其他成员也都拥有承担设计任务的独立意图和相关意图,以联合信念的形式予以确认,并广播到每个成员.给出了一个达成的对话过程,总结了对话模式的效用与代价公式.实际应用表明,所提对话交互模型在没有固定消息顺序的情况下,可限制每个点上的可能响应,同时其Agent可以表达特定言语行为的态度,并产生这一态度的逻辑推理过程.

Dialogue-Based Model for Multi-Agent Cooperative Interaction

Due to the randomness and blindness in the task decomposition and task allocation of the cooperative design,a dialogue-based multi-agent interaction framework was proposed.Firstly,the initiator opened an information-seeking dialogue with each agent in turn,inquired the specialty domain,ability,opportunity and willingness for joining in a task,and judged whether an agent was suitable.Secondly,the initiator opened a persuasion dialogue for asking each agent to take same intention,namely regarding-realizing design task as its own intention.Thirdly,the initiator might ask each agent to accept that the agents in the potential team have same individual intention and mutual intention which would be confirmed in joint belief and broadcasted to each agent.The utility and cost of dialogues based on an example were analyzed and further research issues were proposed.