【作者】 周鋆；

【导师】 张维明；

【作者基本信息】 国防科学技术大学 ， 管理科学与工程， 2011， 硕士

【摘要】 现代战争是参战双方作战力量间的高度对抗,对敌目标进行分析,对关键目标实施打击,从而在信息、火力和机动性等领域形成压倒性的对抗优势是军事强国进攻作战的最优先考虑。因此对战场多目标网络特性进行综合分析,总结其中目标遭受打击后的演化规则,探索其重心以及弱点具有十分重要的意义。本文以战场多目标网络打击问题为研究对象,对问题重心进行了分析,并研究了打击策略生成方法。主要包含以下几个方面的内容。首先,本文对战场多目标网络自身开展了相关的理论研究。详细描述了战场目标、目标影响关联、以及目标演化规则,即备份规则和自修复规则。在图论模型的基础上,构建了较为抽象的战场多目标网络模型,为重心分析和打击策略生成研究提供了基础。其次,本文提出了基于多实体贝叶斯网络的重心分析框架。该框架利用多实体贝叶斯网络来描述打击对象的不确定性,以及目标演化等不确定信息,在此基础上分析了战场多目标网络中各个目标对网络重心的影响程度。其中,对多实体贝叶斯网络的基本理论以及推理算法进行了分析,并给出了战场多目标网络重心选择方法和实体片断组合算法。再次,本文提出了基于马尔科夫决策过程的打击策略生成方法。该方法借鉴马尔科夫决策过程在军事计划领域的成功应用,构建了问题抽象后的马尔科夫决策过程模型,包括利用目标的状态构成状态集,利用对目标的打击构成行动集,利用目标状态的变化确定状态转移函数,以及利用之前的重心分析模型构建效益值函数。最后给出了问题的求解算法。最后,在案例实验中,实现了重心分析的全部过程,使得方法本身的可行性得到了检验,另外还得到了各个目标对网络重心影响的具体数值,并将该数值转化为马尔科夫决策过程模型中的奖励值。在重心分析的基础上,接着给出了案例对应的马尔科夫决策过程模型,通过策略迭代算法,对模型进行了求解,得到了包含在不同状态下开展不同行动的最优打击策略。战场多目标网络打击问题是军事领域的一个难题,本文在战场多目标网络重心分析的基础上探索了打击策略生成的问题,研究成果可以作为军事指挥人员选择打击目标或生成打击策略时的参考。更多还原

【Abstract】 The morden warfare could be regarded as the counterwork between the military forces in both sides. To analyze the contrary targets, attack the critical target, and get the superiority in information, firepower, and flexibility are in the first consideration of power nation to carry out the attack operation. So it is meaningful to analyze characteristics of multiple targets network in battlefield, summarize the transformation rules of its elements, and explore its center of gravity (COG) and vulnerability. This paper focused on the attacking problem of multiple targets network in battlefield, analyzed the COG of this multiple targets network, and studied the method of attack strategy generation. This paper contains these works listed below.Firstly, this paper carried out the theoretical study on multiple targets network in battlefield. It analyzed targets in battlefield, relationships between targets, and transformation rules of targets. Then two transformation rules (backup rule and self-repair rule) were outlined that based on the former analysis. Then, the mode l of multiple targets network in battlefield was built that based on the graph theory model. This model afforded supports for later studies on the COG analysis and the attack strategy generation process.Secondly, this paper proposed the COG analysis framework, which based on the Multi-Entity Bayesian Networks (MEBNs) theory. This framework used MEBNs to describle the uncertain information such as the uncertainty in target choosing and the evolution of targets, and analyzed the effects of every target on the COG in a multiple targets network. During this study, we introduced the basic theory and inference algorithm of MEBNs, and proposed the COG selection method and the combination algorithm of MFrags.Thirdly, this paper proposed the method of attack strategy generation, which based on the MDP. As there are so many successful applications that applied MDP in military planning domain, this paper also applied this method in attack strategy generation process of multiple targets network in battlefield. We built the stochastic system of this problem, the state set composed of targets’states, the action set composed of attacks on every target, the transition function based on the transition of targets’states, and value function based on the former results of COG analysis. Then we outlined the algorithm to slove this problem.Finally, in the case study, we implemented all parts of this COG analysis process, which tested the feasibility of this method. What’s more, we also got the specific impact values of every element on COG from it. These values were transformed into the rewards in Markov Decision Process (MDP), which was the input of attack strategy generation. Then, this MDP mode l of this problem was built that based on the former COG analysis. With the policy-iteration algorithm, we got the optimization attack strategy, which contained different actions refer to different states.Attacking multiple targets network in battlefield is really a difficult problem in campaign. This paper explored the attack strategy generation problem which based on the COG analysis. The results could afford supports for staff to choose appropriate target or generate optimum attack strategy.更多还原