【作者】 陈彬；

【导师】 黄柯棣；

【作者基本信息】 国防科学技术大学 ， 控制科学与工程， 2010， 博士

【摘要】 随着军事领域联合作战、工程领域载人航天和社会领域应急管理等复杂系统的规模的日益扩大,传统建模与仿真方法难以满足其对多领域和仿真性能等方面更高的要求。而多范式建模与仿真方法,这一新的复杂系统建模与仿真支撑技术对于适应多体系联合作战等复杂系统仿真具有重要意义,故对此关键技术的研究显得非常重要。论文针对多体系联合作战,从理论基础、模型变换方法和仿真优化方法三个方面对多范式建模与仿真方法进行系统的研究。多范式建模方法的理论基石是一般系统理论和模型变换方法。复杂系统的多领域性由一般系统理论衍生出的大量基本形式化方法支持;而仿真一致性则由模型变换和仿真互联方法保证。对离散事件系统规范(Discrete EVent System specification,DEVS)的研究表明,DEVS可以等价描述其他连续和离散的形式化方法。因此,在基于元模型的模型变换理论指导下,论文选择DEVS作为通用仿真形式化方法(General Property Simulation Formalism,GPSF),研究将多体系联合作战系统中多领域模型变换为DEVS模型的一系列方法:(1)对描述宏观行为的StateCharts模型变换为DEVS模型的方法进行研究,以解决高低层次模型之间协同仿真的问题。变换方法建立了包括语法、状态和事件三个方面的变换规则,实现了迁移、使能条件和历史状态的等价以保证规则的完备性。(2)对组件化的BOM模型变换为DEVS模型的方法进行了研究,以重用BOM模型。变换方法采用将组件内核映射到DEVS模型的方式实现从BOM到DEVS的变换,证明了采用相同组件内核的DEVS-CK模型与BOM组件模型是等价的。(3)基于DEVS耦合模型的耦合和封闭特性,对DEVS耦合模型的扁平化方法进行研究。通过剥离模型层次化和构造扁平化原子模型两个步骤完成耦合模型的扁平化,从而解决高耦合度模型通信效率低的问题。(4)对多体系联合作战系统中连续-离散混合系统的模型描述问题进行研究,提出利用Modelica语言描述混合系统DEVS模型的方法。实现了Modelica-DEVS编译器,将Modelica模型编译为可以执行的仿真模型。在模型变换方法的基础上,为了提高仿真运行效率,论文分别从分布式仿真和模型活跃度两个角度对DEVS仿真的优化方法进行了研究:(1)提出了基于MPI的分布式并行仿真框架,设计了支持乐观时间策略的DEVS仿真引擎以提高仿真性能。同时,在仿真框架中实现了错误探测、状态存储和系统恢复的算法使其具备了容错能力。(2)基于模型活跃度提出了活跃度增强式建模方法,在领域模型中嵌入融合活跃度元模型(Activity Combined Meta Model,ACMM)实现活跃度操作。在追踪活跃度和资源占用量化方法的支持下,资源敏感(Resource-aware)仿真框架通过计算活跃度进行资源的重分配,达到最优配置。此外,论文还提出了基于代理模型的仿真互联方法,解决DEVS和高层体系结构(High Level Arichitecture,HLA)仿真系统的互联问题;讨论了基于Web的仿真过程管理方法,支持对异构仿真系统的一体化管理工作,通过合理配置仿真资源优化系统运行。基于以上研究工作,论文设计和实现了面向DEVS的多范式建模与仿真原理样机系统。在建模阶段,系统引用多种建模工具进行多领域建模,通过模型变换获得模型的一致性描述;在仿真阶段,系统利用基于Web的仿真实验管理工具和态势统计工具实现了B/S结构的仿真运行架构。最后通过多体系联合作战用例的测试,验证了论文所研究的多范式建模与仿真方法的可行性。更多还原

【Abstract】 With the development of complex systems such as Joint Operations in military field, Manned Space Flight in engineering field, Emergency Response Management in social field, the traditional Modeling and Simulation theory cannot satisfy the requirements from complex systems such as multi-field and simulation performance now. Multi-Paradigm Modeling and Simulation which is a novel support technique can solve problems on Modeling and Simulation for Complex Systems. The research on the key techniques in Multi-Paradigm Modeling and Simulation means a lot to the Multi-System Joint Operations simulation. We study the Multi-Paradigm Modeling and Simualtion on three aspects: Basic Theory, Key Model Transformation methods and Simulation Optimization Algorithms.General System Theory and Model Transformation are the basement of Multi-Paradigm Modeling and Simulation. The Multi-Field of Complex System is supported by the formalisms derived of General System Theory. The simulation consistence of Complex System is maintained by Model Transformation and Co-Simulation. The research on Discrete EVent System specification (DEVS) shows that DEVS could be used to equivalently describe other Continuous or Discrete formalisms. We select Discrete EVent System specification (DEVS) to be the General Property Simulation Formalism (GPSF) in Complex Systems. As a result, we study a series of Model Transformation for the Multi-Filed Models in Multi-System Joint Operation under the instruction of Meta-Model based Model Transformation.(1) StateCharts is a typical high level formalism to describe the macroscopical behavior of system. The transformation from StateCharts to DEVS is proposed to sovle the interoperability between high level and low level models. Transform rules in gramar, state and event are constructed. The equivance of transition, guard conditions and history state is implemented to accquire the completeness of transformation.(2) Base Object Model (BOM) is a model description in BOM-based Distributed Simulation. We reuse the Compenent Kernel in BOM model by developping the transformation from BOM to DEVS. We prove that DEVS-CK is equal to the BOM component if they use the identical Component Kernel.(3)The Flattening algorithm of coupled DEVS model is proposed according to the closure under coupling of DEVS. It improves the efficiency in High Coupled Low Computation models. Flattening is realized by direct connection and flattened atomic model construction.(4)As the continious-discrete systems are used a lot in Multi-Systems Joint operations, we give the DEVS description of Hybrid System in Modelica. Modelica-DEVS compilor is implmented to generate simulation models. Based on the Model Transformation, we study the optimization algorithms for DEVS simulation in the view of distributed simulation and model activity:(1) We propose the MPI-based distributed framework which supports parallel simulation. DEVS engine is designed to support optimistical time algorithm. The Fault Detection, State Storage and System Recovery are implemented in framework to do the Fault Tolerance.(2) We present the Activity Enhancing Modeling based on model activity. Activity Combined Meta Model (ACMM) is designed to integrate the activity into model. With the help of Activity Tracing and Quantization of Resource Usage, the Resource-aware Modeling and Simulation framework can re-allocate the resource by the activity calculation.In addition, we present an Agent-based Co-Simulation method, which works well in the Co-Simulation between DEVS and HLA simulation system. Meanwhile, the Web-based Simulation Management is given to support the integrated management for heterogeneous simulation.Based on the research on the Multi-Paradigm Modeling and Simulation, we design and implement the prototype system. In the modeling phase, many kinds of modeling tools are reused in the system to do the Multi-Field Modeling. Model Transformation is used to maintain the consistence in model description. In the simulation phase, the web-based management and statistical tools are used to implement the B/S system structure. The Multi-System Joint Operation cases testify the research on key techniques for Multi-Paradigm Modeling and Simulation.更多还原