自治分层规范系统模型研究

规范多agent系统通过规范控制agent个体行为实现系统宏观目标。在自然语言描述的抽象规范和操作语言描述的具体规范之间存在巨大的语义鸿沟,采用分层规范模型将高层抽象规范逐级转换成底层具体规范是解决这一问题的有效手段。但是目前的分层规范模型存在着层次过多,层间界线模糊的问题,同时规范细化和层间规范转换完全依赖手工完成,不仅降低了系统可靠性和动态适应性,同时提高了规范系统开发成本。本文提出并实现了一种自治分层规范框架模型,该模型将规范系统分成三个层次,并且在形式化规范层建立概念语义蕴涵关系和行为责任关系实现了规范的自动细化,建立行为-操作映射模型实现不同层次的形式规范间的自动转化,提高了规范系统自治性,降低了规范系统开发成本,提高了系统可靠性和动态适应性。

Research of the Autonomous Layered Normative System Framework

Normative multiagent system realizes high-level macro objective through control of agent behavior with norms.There is a great semantic gap between abstract norms described by natural language and the concrete norms described by operational language.The layered normative system model which translates the high level abstract norms into low level concrete norms level by level is the effective mechanism to this problem.But the layered models to date all have too many layers with unclear borderline between layers,and the translation is fully dependent on manual.This not only reduces the system reliability and dynamic adaptability,but also increases the system development cost.This paper introduced an autonomous layered normative system framework which divided the normative system into three different layers,and set up semantic entailment relation of concepts and behavior responsibility relation to realize automatic norm refinement,behavior-operation mapping model to realize automatic inter-level norm translation.With these mechanisms,the model effectively increased the system autonomy,with reduced normative system development cost and increased system reliability and dynamic adaptability.