在基于Agent的计算中,现有研究对于动机的定义和动机影响Agent行为决策的分析比较片面,缺少对动机的清晰考虑和相应目标产生的描述.为此建立了具有动机的Agent思维状态模型,将动机分为社交行为准则(N)、策略(P)、契约准则(c)和内部愿望(D).形式化地描述了动机扩展的Agent思维状态模型(NPCD-Agent)和BNPCDGI逻辑,阐述了逻辑的语法、语义以及相应的公理系统.给出了动机算子基于Kripke可能世界的新的语义解释,验证了动机算子不存在逻辑全知问题.通过计算0-1规划和设定优先顺序,建立了动机冲突消解模型,改进了现有包含动机的Agent思维状态模型实现方法.实例分析表明,改进的Agent思维状态模型使得Agent行为决策过程更加理性.
In agent-based computing, recent studies on the definition of motivation and its effects on agent's decision-making are not complete. There are no clear descriptions of motivation and relative process about goal production. An agent mental state model with motivation was constituted, in which motivation is divided into four classes, i.e. social behavior norm (N), policy (P), contract norm (C) and desire (D). The extended mental state model (NPCD-Agent) and BNPCDGI logic of agent were formally described, and the syntax, semantics and axiom system of the logic were discussed. Especially for motivation operators, a new semantic interpretation based on Kripke’s possible worlds was presented. The avoidance of logical omniscience problem on motivation operator was also proved. Through the computation of 0-1 programming and configuration of preference, a conflict resolving model of motivation was described. The previous implementing approach to agent mental state model with motivations was improved. The instance proved that with the extended agent mental state model, the agent's decision-making process becomes more rational.