【作者】 刘秀罗；

【导师】 黄柯棣；

【作者基本信息】 国防科学技术大学 ， 控制理论与控制工程， 2001， 博士

【摘要】 CGF行为建模作为先进分布仿真技术的重要研究方向，它涉及多学科的交叉和融合， 在军用仿真技术领域有着广泛的应用前景。论文主要是围绕CGF行为建模方法、CGF建 模相关技术以及指挥控制建模展开研究的，重点讨论了基于Agent的行为建模方法以及多 Agent系统理论在指挥控制建模中的应用，最后介绍了神经网络技术与指挥控制建模在实 际仿真系统中的应用。 首先，从CGF行为建模方法的研究出发，在介绍传统CGF行为建模方法的基础上， 分析了其存在的不足，进一步提出了基于软计算的行为建模方法，并对每种方法的工作原 理、适用环境进行了描述；依据认知处理的一般过程，提出了行为建模的集成框架，并具 体分析了各组成部分及其相互关系;文中重点讨论了两种具体的CGF行为建模方法：基于 有限状态机的行为建模方法和基于Agent的行为建模方法，分析了每种方法的特点、实现 机制以及有关的算法； CGF建模涉及到许多相关技术的研究，环境建模与多分辨率建模就是其中比较有代表 性的。论文首先对与CGF行为建模相关的环境建模问题进行了讨论，提出了基于SEDRIS 的CGF地形建模过程；接着，又从多分辨率建模的有关理论基础出发，在介绍有关概念的 基础上，讨论了多分辨率建模的一致性问题，并重点分析了其中基于位置的一致性求解算 法，对于多分辨率建模的实现，主要介绍了三种方法：聚合解聚法、视点选择法和多分辨 率实体法，其中对多分辨率实体法的描述模型和一致性维护方法进行了深入分析，并将它 与其它两种方法进行了比较，指出了其在多分辨率建模方面存在的优点。 指挥控制建模是CGF行为建模在作战仿真中的一个重要应用领域，根据指挥控制建模 的一般特点和对行为建模的要求，提出了利用多Agent系统理论来建立指挥控制中的协同 工作模型，其中重点分析了基于联合意愿理论和共享计划理论的协同工作模型，对其实现 机理进行了深入的分析，并给出了具体的实现算法；基于对行为建模和指挥控制方法的研 究，提出了作战仿真指挥控制建模的框架C~2MF，并具体分析了其中指挥控制行为的描述 方法、通讯机制的建模以及基于分层表示的规划模型。 最后，论文结合两个具体的仿真系统，讨论了CGF行为建模与指挥控制建模在实际系 统中的应用。首先介绍了离散Hopfield神经网络模型在火力分配中的应用，给出了具体的 求解算法和仿真结果；然后，结合所提出的对指挥控制建模框架C~2MF，对“陆军师联合 登岛作战仿真系统”中的指挥控制建模进行了深入的分析。更多还原

【Abstract】 As an important research direction for advanced distributed simulation, CGF behavior modeling is characterized as an intersection of multidisciplinary and holds great promising for the military simulation technology. Our research work in this dissertation focuses on CGF behavior modeling, relative technologies to CGF modeling and command and control modeling, with emphasis on Agent-based behavior modeling method and the application of Multi-Agent system theory in the command and control modeling. In the end, the application of neural network technology and C2 modeling into military simulation systems are introduced.The research of CGF behavior modeling methods is first considered. Upon the introduction of traditional behavior modeling method, its disadvantage is analyzed, and then new methods based on soft computing are presented, with the description of their work principle and adapted environment. According to the cognitive process, the integrative framework of behavior modeling is proposed, as well as the analysis of its components and their relationships. Two specific CGF behavior modeling, finite state machine based method and agent-based method are discussed in this dissertation, their characteristics, implementation mechanism and relative algorithms are also introduced.There are many technologies relative to CGF behavior modeling. Environment modeling and multi-resolution modeling are two among them. The author first discusses the environment modeling, and proposes SEDRIS-based terrain modeling process. Then, another technology, multi-resolution modeling is proposed. Based on its fundamental theory, some concepts are introduced, and then the consistency problem of Multi-resolution modeling is discussed, with emphasis on the position-based consistency resolving algorithm. There are three methods for implementing the multi-resolution modeling, including aggregation-disaggregation, select-view and multi-resolution entity, among them the descriptive model and consistency maintaining method of multi-resolution entity are analyzed in detail. After compared with the other two methods, the advantage of multi-resolution entity method is given.Command and control modeling is an important application area in CGF behavior modeling. According to its common characteristics and requirements for behavior modeling, the multi-agent system theory based teamwork model for command and control modeling is proposed, which is built upon the joint-intention theory and shared-plan theory. Its implementation mechanism is analyzed in detail, and the specific implementation algorithm is also given. As a result of the research of behavior modeling and command and control modeling, the author presents the framework of the command and control modeling, which is C2MF. And discusses its behavior description method, communication mechanism modeling and hierarchical representation based planning model.Finally, two specific simulation systems are discussed, which incorporate CGF behaviormodeling and command and control modeling. At first, the author introduces the application of discrete Hopfield neural network model in the fire-allocation problem, and gives the implementation algorithm and simulation results. Then based on the framework of command and control modeling: C2MF, a thorough analysis on command and control modeling of the "Infantry Division Landing Island Joint Warfare Simulation System" is made.更多还原