Research On Key Technologies Of Military Analytic Simulation For High Level Decision-Making

Modeling and simulation is an important way to study complex system. Especially, under the condition of large-scale joint operation, high level decision support system, which is built upon military analytic simulation, can provide supports for the high level decision makers in their warfare command process through a combination of qualitative and quantitative analysis methods.There are a lot of characteristics of military analytic simulation, including: high speed computing, multi-samples running, human-computer interaction, and interoperability for running, attributes and I/Os. In order to predict the outcomes of the future warfare, it should also focus on the overall behaviors for the system which arise from the interactions between large-amounts of entities, due to the deep uncertainty and emergence of the military complex system. However, because they are designed for different purpose, the current military analytic simulations cannot meet these needs fully. Therefore, the study of Military Analytic Simulation for High Level Decision-Making(MASHLDM) is the basis for building military decision support system, and has great significance in theory and practice for our national defense and army construction.Under the guidance of “combination of qualitative and quantitative analysis” approach, the main objective of this dissertation is how to build a MASHLDM. Aiming at the characteristics of MASHLDM, the key problems of the research framework, the simulation engine, the interest management, and the building of simulation models for MASHLDM are thoroughly studied in the dissertation. The innovations and contributions are listed as follows:(1) Building the general research framework, and proposing the system theory of “3-Driven, and 1-Core” for MASHLDM。The basic concepts in MASHLDM are analyzed systematically. The principles for building MASHLDM is put forward based on the research on basic elements from the technical perspectives of simulation software, application in decision-making and military domain. After the system structure of MASHLDM is designed, the elements and their relationships are clearly pointed out, where the simulation engine is the core. The core of static elements and the dynamic features of the system, which are “C2-driven military model”, “Model-driven development of model”, and “data-driven running of models”, form the system theory of “3-Driven, and 1-Core”.(2) Proposing an Agent cycle model of “sense-think-lookahead-act”, designing a mediator-based hybrid simulation framework, and implementing a MASHLDM simulation engine based upon hybrid simulation. As a mixture of Agent-based(ABS) and Discrete Event Simulation(DES) techniques, hybrid simulation can reveal the complexity of military system, and skip over irrelevant computes and updates of Agents through timing mechanism of partial resynchronization. However, the cyclic dependencies between Agents and prediction of next resynchronization time are both issues need to be addressed in current hybrid simulation. To resolve these problems, the paper first presents a lookahead approach based on “time window”, and introduces the lookahead step into traditional “sense-think-act” cycle, which makes the interactions between Agents to be planned and the events are scheduled in a proper sequence. A formal proof shows that “lookahead” can eliminate the cyclic dependencies during resynchronization. Then, a MABS(Mediate-based and Agent-Based Simulation) framework is designed to set resynchronization intervals, and to provide common functions which are independent of behavior logic of Agents in problem domain.(3) Presenting a hybrid interest management for MABS in large-scale military analytic simulation. Considering that there are interaction data and message data exchange between MABS Agents, the spatial-based and predicate logic-based approaches are studied to implement interest filtering at physical sensation level and communication level. In order to improve the efficiency of sensation mediator, optimization models of moving speed and path are built, and a dynamic quadtree based spatial interest management is introduced, which greatly reduce the trigger conditions and amounts of sensation matches.(4) Presenting the modeling method of MABS-DES models based upon MDA(Model Driven Architecture) technique. Simulation model is the foundation of simulation experiment, and building MABS-DES models is an important step in building MASHLDM. When applying MDA approaches, the Agent-based model is chosen as PIM(Platform Independent Model), and the DEVS(Discrete Event System Specification) model is used as intermediate PSM(Platform Specific Model). After formal description of the PIM and PSM, the feasibility of model transformation between Agent-based model and DEVS model, and between DEVS model and MABS-DES model are validated in this paper. And the mapping procedures between these three models are studied. Finally, a MABS-DES model development tool is designed and implemented based on model-driven to support the rapid construction of MABS.(5) Designing and implementing a prototype military analytic evaluation system based on the research of this paper, and a hypothetical scenario is taken as an example to introduce the application of the system. The results of the example experiment show that the work in the thesis can improve the performance of large-scale military analytic system compared with traditional simulation which is based on KD-HLA.