DIDSS环境下信息agent系统任务规范的分解  

Tasks specification decomposition on information agent systems under environment of DIDSS

作　　者：杨善林[1] 胡小建[2] 马溪骏[1] 

机构地区：[1]合肥工业大学计算机网络研究所,安徽合肥230009 [2]合肥工业大学材料成型与控制工程系,安徽合肥230009

出　　处：《系统工程学报》2004年第5期489-495,共7页Journal of Systems Engineering

基　　金：国家自然科学基金资助项目(70171033);教育部人文社科十五规划基金资助项目(01JA630061).

摘　　要：提出一个基于GDF(扩展决策函数)的多agent任务规范分解判据,提出多agent的任务规范分解性质;将多agent任务规范分解问题等价为建立一个与信息agent任务规范属性相关的MSBN(多模块贝叶斯网)问题,并证明了MSBN的d cutset(d-割集)和分解判据之间的关系,提出了与多agent任务规范对应的MSBN的分解方法.最后通过示例验证了所提出的方法的可行性.GDF( generalized decision function)_based decomposition criterion in multi-agent systems and multi-agent task specification decomposition features are put forward in this paper. It is pointed out that problem of tasks decomposition in multi-agent systems is equivalent to that forming a MSBN (multiply sectioned Bayesian network) related to information agent. The relation between d-cutset of MSBN and decomposition criterion is proved, and decomposition method of MSBN related to multi-agent task specification is put forward. Feasibility on the method put forward is verified with an example in the end.

关 键 词：信息AGENT DIDSS环境 扩展决策函数 GDF 任务规范分解 贝叶斯网 MSBN 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]