摘要
在多 Agent系统中协调是系统效率的瓶颈 ,目前常用的协调机制有组织结构、交换元级信息、多 Agent规划、集体目标等 ,而集体目标具有最低协调成本的功效 .在没有一个标准逻辑描述集体目标的今天 ,在某些约束下 ,给出集体目标的形式规范是可行的 .首先给出了动态可能世界语义学 ,在此基础上证明 Sandu给出的集体目标约束条件过强 ,并回答了 Sandu没有解决的问题 :集体成员如何相互相信各自同时实现集体目标 .
Coordination is the bottle-neck of system efficiency in multi agent systems. At present, the mechanisms of coordination which are often used are group structure, exchange meta information, multi agent plan, collective goals etc, but collective goals possess the character of the lowest coordination cost. Since collective goals can't be described with norm logic, it is feasible to present the formal norm of collective goals in some restriction. In this paper, semantics of dynamic probable world is first presented and it is proved on this basis that the conditions of collective goals presented by Sandu are too broad and the unsettled problem is solved: how collective remembers believe each other and realize collective goals at the same time.
出处
《计算机研究与发展》
EI
CSCD
北大核心
2001年第11期1310-1314,共5页
Journal of Computer Research and Development
基金
国家自然科学基金资助 ( 699730 2 3)