【作者】 贺勇军；

【导师】 戴金海；

【作者基本信息】 国防科学技术大学 ， 航空宇航科学与技术， 2004， 博士

【摘要】 随着信息时代的来临,构建综合性的空间信息获取系统对于社会发展和国家安全具有重大战略意义。而智能化的多卫星分布式自主协同工作,将是此系统的主要运行模式。如何分析、评估和优化其总体效能,以寻求最满意设计方案,是迫切需要研究的重要课题。 论文分析指出,唯有采用普适性的顶层分析和优化设计框架,才能够有效解决面临的种种复杂性问题。而基于进化计算的优化方法,采用自适应随机搜索全局优化技术,具有广泛的适应性和包容性;基于多Agent的建模与仿真方法,则能够胜任复杂智能系统的综合效能仿真分析任务。为此,提出采用基于进化计算的仿真优化设计框架,并以对地观测多卫星系统为主要研究对象,重点对其效能分析方法、效能测度的确定及其求解模型、综合建模与仿真方法及相应软件的研制等方面展开了研究与实践。 首先,研究了多卫星系统具有的诸多复杂性特征;分析和归纳了多种典型构成形态;从时空、物理和功能特性等方面,研究了多颗卫星之间的分布与协同特性;对智能化自主协同运行多卫星系统的总体功能、特性、实现机制及其组织结构进行了分析。 然后,对一般性系统效能分析的几种主要方法进行了简要述评,提出了基于仿真的多卫星系统效能分析方法的总体框架。遵循此框架,界定了多卫星系统的层次结构;就复杂大系统中子系统效能的分析方法进行了讨论,给出两种基本方法:任务要求分解法和全系统效能敏感性分析法;提出了系统信息获取的四元组效能测度:内容、范围、质量和时效性,并将其具体化为遥感信息获取的空间(Space)、频谱(Spectrum)、能量(Energy)、时间(Time)(简称SSET)四维空间的相应测度。 接着,提出了离散化SSET空间的信息粒模型,给出了单个遥感器信息获取容量的计算模型,并对多卫星系统的时间、空间、频谱和能量维信息容量进行了分析;面向效能分析,提出了“物理覆盖”、“有效覆盖”、“充分信息量”等概念,并通过一典型示例系统,给出了基于仿真的效能测度求解方法及其基本模型;针对系统覆盖区域的非规则性,设计和实现了“池中投石法”、“油环点火法”和“逐步吸收法”等基于仿真的覆盖区域通用求解算法。 分析了复杂系统可靠性模型和可靠度综合计算的特点,提出了复合逻辑树(Composite Logic Tree Model,简称CLTM)综合建模方法及其可靠度递归综合算法,并设计和实现了相应的软件系统,论文对其设计思想、总体结构、主要功能及其实现机制进行了阐述。 此后,主要针对复杂多卫星系统的综合建模与仿真方法展开了研究。首先分析指出,所建模型应为分解结构描述水平上的多学科混合异构层次化系统整体同构模型。分析了面向对象的建模与仿真方法对此提供的支持机制及其局限性,提出了基于多Agent/Object的整体建模与仿真方法,并给出其总体框架。然后,分别就Object与Agent共存机制、实体属性和行为的确定原则、基于进化计算的多学科优化对系统整体建模的要求、Agent/Object模型粒度确定等方面进行了讨论,分别给出了面向Object和面向Agent的解构与重构方法,更多还原

【Abstract】 With the coming of information era, it is of great significance for society development and nation security to construct integrated space information acquisition system, which will operate in intelligent, distributed, collaborative, autonomous, multi-satellite mode. How to analyze, evaluate and optimize the integrated effectiveness of the complex systems to get the most satisfying one is a crucial problem which must be answered as early as possible.We made a detailed analysis and concluded that only if a general system analysis and optimization framework were adopted can we solve the multitudinous complex problems. The Evolutionary Computation (EC) based optimization methods, adopting self-adaptive stochastic searching technology for global optimization, possess comprehensive adaptability and compatibility, moreover, the Multi-Agent Based Modeling and Simulation (MABMS) methods are competent for analyzing the integrated effectiveness of complex intelligent systems. So, we presented an EC based simulation optimization design framework for Multi-Satellite Systems (MSSs). Under the framework, emphasizing on earth observation MSSs, we have done some research mainly to propose effective methods for system effectiveness analysis, establish Measures of Effectiveness (MoEs) and their solving models, set forth integrated Modeling and Simulation (M&S) methods, and construct corresponding M&S software environment.First, we investigated the various complexity characteristics existing in MSSs in detail, analyzed and induced typical composing configuration of MSSs, explored their distribution and collaboration relationships from the aspects of space, time, physics and function, and analyzed the overall functions, main characteristics, implementation mechanisms and organization structures of MSSs operating in the intelligent, autonomous, collaborative mode.Then, we reviewed briefly several main general methods for effectiveness analysis, and put forward the overall framework of effectiveness analysis methods based on M&S. According to the framework, we defined the hiberarchy of MSSs, discussed how to analyze the effectiveness of the subsystems in System of Systems and gave two basic methods: mission requirements decomposing and effectiveness sensitivity analysis of holistic system, set forth the four-element group of MoEs for information acquisition: Content, Area, Quality, Response time, and concretized them into corresponding MoEs in four-dimension space (Space, Spectrum, Energy, Time, SSET for short) of information acquisition by remote sensing.And then, we put forward an information grid model in discrete SSET space, deduced the models for computing total information capability of one single remote sensor, analyzed the information capability of MSSs in respective dimension of the SSET space, and set forth the following concepts: physical coverage, effective coverage, sufficient information quantity.Through a typical example, we provided simulation based approaches and models to solve the MoEs for the system in detail. In order to search the points in or on the edge of the irregular coverage regions, we designed and implemented several general algorithms and named them throwing a stone into a pond, igniting an oiled hoop, absorbing step by step. The main ideas and contents of them are expounded.By analyzing the characteristics of the models and integrated computing for reliability of complex large systems, we set forth a new model named Composite Logic Tree Model (CLTM) and its modeling principles, and designed a recursive algorithm based on CLTM, and developed the corresponding software. The design ideas, architecture and main functions, implementation mechanisms of the software are introduced.And then, we pointed that, the models built for holistic system optimization should be integrated multidisciplinary isomorphic Hybrid Heterogeneous Hierarchical Models (HHHMs) in the description level of decomposing structure. By analyzing the supporting mechanisms and limits of the Object-Oriented M&S (OOMS) methodology, and the excellent characteristics of the MABMS methodology, we presented the Multi-Agent/Object (MAO) based M&S and optimization method framework for complex MSSs. Then, we discussed the mechanisms of Agent/Object co-existence, the principles of choosing the granularity of the Agents/Objects and defining their attributes and behaviors, the requirements to integrated system modeling by EC based Multidisciplinary Design Optimization (MDO), and presented the methods of Agent/Object oriented structure decomposition and re-construction, constructed class diagrams of objects in MSSs, set forth the rules for hierarchical entity decomposition and aggregation. Taking example for satellites, we presented their various Agent models. At last, the main organization models of Multi-Agent Systems (MAS) in MSSs are presented.Adopting above ideas, methods and models, we have designed and implemented the Modeling and Simulation Software for Integrated Effectiveness of Multi-Satellite Systems (MSSE for short). With powerful visualization and interactive capabilities, it can be used to build HHHMs of complex MSSs, configure the missions flexibly, monitor the whole simulation process, present all-around reports of data analysis and display vivid 2-D and 3-D animations, and eventually provide powerful supports for the design, evaluation and optimization of the MSSs. Finally, the design concepts, architecture, functions, and implementation mechanisms of the software are expounded.更多还原