摘要
组合可能冲突的信息是智能推理中重要的研究内容,其中,证据理论是一种典型的不确定性智能推理方法。在证据理论中,用以融合和更新证据信息的推理组合规则是其理论基石。Dempster组合规则通过合取运算来构造信息结构,得到了广泛的应用。然而,当证据冲突比较大、证据源不完全可靠或辨识框架不完备时,Dempster组合规则得到的结果并不符合常理。为此,改进组合规则和处理证据冲突,成为证据理论中一个重要的研究课题。将全面回顾和比较国内外现有的主要推理组合方法,基于不同的冲突处理机制,建立相应的理论框架,涵盖绝大多数推理方法,从理论上对证据推理方法的研究,进行一个全面的总结,并剖析各种组合方法的优缺点和相互关系,指出最新的研究趋势。
The combination of the possibly conflicting information is an important issue in intelligent reasoning. Evidence theory is one of classical intelligent reasoning theories under uncertainty. The reasoning combination rule for combining and updating evidence information is an important component in this theory. Dempster rule of combination that reconstructs information structure by filtering conflict is widely applied. Unfortunately, when the conflict is severe, the combined result obtained by Dempster combination rule is illogical. Therefore, to modify the combination rule and manage the conflict is the focus in the research of reasoning with uncertainty and multi-source fusion. A review and comparison of currently principal reasoning methods are presented. We construct the corresponding theory frameworks based on different methods and analyze their merits, drawbacks and mutual relationships. We have tried our best to predict the research trend.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2007年第6期890-898,共9页
Systems Engineering and Electronics
基金
国家自然科学基金(30400067)
国防重点实验室基金(51476040105JW0301和51476040103JW13)
航空支撑基金(2004JD-4)
中国博士后基金(200436273)
图像处理与智能控制教育部重点实验室基金(TKLJ0410)
上海市自然科学基金(03ZR14065)资助课题
关键词
证据理论
冲突
组合规则
不确定性
evidence theory
conflict
evidential combination rule
uncertainty
