【作者】 姜鑫；

【导师】 沙基昌；

【作者基本信息】 国防科学技术大学 ， 管理科学与工程， 2011， 博士

【摘要】 面向人因复杂性的军事对抗决策研究围绕军事复杂系统由“人”参与而产生的复杂性,以人因复杂性为特定视角系统深入地研究不确定性环境下军事对抗决策的分析、建模与应用。本文将军事对抗决策视为一类具体的人因复杂问题,引入多主体影响图、博弈论以及超博弈等理论工具分析军事对抗决策,并提出博弈网模型建模和求解被描述为超博弈的军事对抗决策问题,为军事对抗决策研究提供了全新的视角,具有重要的理论价值和现实意义。本文工作的主要贡献和创新总结如下:(1)研究了面向人因复杂性的军事对抗决策理论基础复杂性、人因复杂性和人因复杂问题是本文研究的概念基础,人因复杂问题求解框架是后文问题研究的需求分析、认识和解决途径的方向性指导,为后文研究奠定了方法论基础。在引入新的复杂性概念体系基础上,提出了复杂性概念的一般认识框架,为复杂性概念自身以及复杂问题求解方法等方面的学术研究和实践应用提供了认识论基准。在人因复杂性、人因复杂问题两个重要概念的基础上,逐步深入论述了“军事对抗决策是典型的人因复杂问题”这一核心命题,分析了军事对抗决策八个方面的复杂性特征以及它们同人因复杂性的联系。最后,建立了军事对抗决策等人因复杂问题的抽象求解框架,指出“从复杂到简单”须经过构建观念框架和变革观念框架两个步骤,并以建立数理模型方式为“变革观念框架”的效果提供了严谨的逻辑推理基础。(2)构建了面向人因复杂性的军事对抗决策分析框架面向人因复杂性的军事对抗决策观念分析框架,为描述和理解现实中的军事对抗决策行为提供了全新的视角。观念分析框架的构建分为两个阶段,首先将军事对抗决策置于博弈论框架下,将其描述为多主体影响图,根据多主体影响图结构诱导的关联图可以有效地识别不同决策者决策行为之间的依赖关系,能够直观地分析人因复杂性本质特征的决策行为循环依赖性;进一步,针对现实中不确定性环境下的军事对抗决策的具体复杂性特点,分析了经典博弈论处理现实军事对抗决策的局限性,对基于多主体影响图和博弈论的分析框架进行了反思,提出利用超博弈替代经典博弈论应对现实中共同知识和理性假设失效的困境,经过变革进一步建立了基于多主体影响图和超博弈的军事对抗决策观念分析框架。观念分析框架的研究遵循人因复杂问题求解框架的基本思路,研究结论进一步印证了该求解框架的指导性和有效性。(3)提出了博弈网模型并研究了基于博弈网的军事对抗决策建模及应用博弈网模型是对超博弈问题进行描述的图模型,能够恰当地反映超博弈的基本特征。系统讨论了博弈网的定义、解析表达,并用两个实例展示了博弈网模型的建模效果。在引入单方博弈网的基础上,定义了信念映射算子,证明了“单方博弈网是博弈网在某个局中人信念下的映射”,讨论了博弈网的信念结构。定义了博弈网的两种形式的解的概念,分别是理论意义上的均衡解和实际情况下的优化解。最后明确了采用博弈网建模超博弈问题的基本流程。基于博弈网的军事对抗决策方法引入了不确定性和有限理性的处理机制,将军事对抗决策过程等价为某个局中人作为决策者的单方博弈网的建立、修正和求解的过程。详细讨论了新方法的基本流程,并用“左勾拳”和“镰割”等两个战史上赫赫有名的军事计划说明了该方法的可行性、有效性和实用性。(4)提出了扩展型博弈网模型并研究了基于扩展型博弈网的军事对抗决策建模及应用在已有研究基础上,提出了三种扩展型博弈网模型,基于扩展型博弈网,研究了面向更加一般性问题的军事对抗决策的建模及应用。首先,对具有多维、动态和层次三种特征的超博弈问题进行了界定,进而分别研究了多维、动态和层次博弈网模型。进一步,将具有多领域和多阶段特征的军事对抗决策问题的求解抽象为扩展型博弈网模型的建立和求解过程。扩展型博弈网求解的基本思路是将其转化为简单博弈网模型,然后采用求解简单博弈网的方法讨论它们的均衡解和优化解。最后,构造算例展示了基于扩展型博弈网的多领域和多阶段军事对抗决策问题的建模与求解过程。将博弈网模型和基于博弈网的军事对抗决策方法进一步推向实用,使得作为本文核心内容的这两项研究兼具理论意义和应用价值。更多还原

【Abstract】 The research on Military Conflict Decision-making (MCD) oriented to Human- caused Complexity (HCC) is guided by the complexity of military complex systems caused by human involving. This dissertation studies thoroughly and systematically on the analyzing, modeling and applications of MCD under uncertainty environments from the view of HCC. The MCD can be seen as a particular kind of Human-caused Complex Problems (HCCP) and analyzed using Multi-Agent Influence Diagrams (MAID), Games and Hypergame. The proposed Network of Games (NoG) model is used to model and solve MCD which is described as Hypergame problems. The new research perspective of MCD is of great theoretical significance and application importance. The main results and contribution of this dissertation are as follows.(1) The theoretical foundation of MCD oriented to HCC is investigated.Complexity, HCC and HCCP are conceptual foundation of this dissertation. The solving framework of HCCP is the methodological foundation of the subsequential research. On the basis of a new conceptual system of complexity, the general epistemic framework of complexity is constructed for conciliating some famous contrary viewpoints. Hoping this conceptual system and the general epistemic framework can further researches and applications of the complexity notions and complexity-oriented problems. With the important concepts of HCC and HCCP as the foundation, it is elucidated that MCD is a representative kind of HCCP. Then the eight complexity features of MCD and the relationships between them and HCC are analyzed. Finally, the abstract solving framework of HCCP is established including two steps which are respectively building and transforming conceptual framework. A mathematical model provides rigorous inference basis for the second step of this framework.(2) The analytical framework of MCD oriented to HCC is constructed.The conceptual analytical framework of MCD oriented to HCC provides a new angle of view on describing and understanding MCD behavior in reality. There are two stages in the building of analytical framework of MCD oriented to HCC. Firstly, the framework is constructed based on MAID and Games. With the problems of MCD under game theory, the circular dependence decision behavior can be captured and handled using MAID. Secondly, considering most military conflicts in reality are under uncertainty environments and the hypotheses of common knowledge and rationality are no longer satisfied, the transforming framework is established based on MAID and Hypergame. The research of the conceptual analytical framework is conformed to the basic idea of the solving framework of HCCP whose guidance is verified again.(3) The Network of Games (NoG) model is proposed, based on which the modeling and applications of MCD are investigated. The NoG model which is defined as a graphical model of the Hypergame problem can capture fundamental futures of Hypergame. The definition and analytic expression of NoG are presented and two examples are used to show the modeling effect of the NoG model. On the basis of inducing Individual NoG (INoG), the mapping operator of belief is defined, the belief structure of NoG is discussed and the proposition of INoG is a map of NoG in one player’s belief is proved. Two kinds of solution concepts are developed. Equilibrium solutions represent agents’best strategies given their beliefs over others. Optimal solutions represent how agents may deviate from their equilibrium solutions. At the end the basic procedures to model Hypergame problems using NoG are created. The MCD method based on NoG induces mechanisms to deal with uncertainty and bounded rationality. The resolving of MCD can be regarded as the process of building, updating and resolving of the decision-maker’s INoG. Then the basic processes of the new proposed method are formulated. Finally, case study of two famous military plans“Left Hook”and“Sickle Cutting”imply the feasibility, validity and practicability of the proposed method.(4) The Extended-Network of Games (E-NoG) model is proposed, based onwhich the modeling and applications of MCD are investigated. Three E-NoG models are proposed on the basis of the above research. The modeling and applications of MCD oriented to more general problems are investigated based on E-NoG. Firstly, three kinds of Hypergame problems are defined and then Multidimensional NoG, Dynamic NoG and Hierarchical NoG are put forward. Secondly, the resolving of MCD problems characterized by multidimensional and dynamic is regarded as the process of modeling and resolving of the corresponding E-NoG models. The basic idea of resolving E-NoG model is to convert it to a simple NoG model. And using the resolving method of simple NoG models investigates the equilibrium and optimized solutions of these E-NoG models. Finally, the case study shows the modeling and resolving procedures of multidomain and multistage MCD problems based on E-NoG models. This part makes the NoG model and the MCD method based on NoG to be more practical and these two key contents of the dissertation to be of great theoretical significance and application importance.更多还原