协同编队自主空战战术决策方法研究

CASIA OpenIR > 毕业生 > 硕士学位论文

	协同编队自主空战战术决策方法研究
	贺雪梅
	2016-05-23
学位类型	工程硕士
中文摘要	多机编队协同空战近年来受到广泛关注，先进的人工智能技术为其提供了坚实的理论基础，飞速发展的空空导弹技术为其提供了武器支持，有人机和无人机之间的编队协同作战的相关研究被世界各国作为未来空战的主要研究方向。目前多机编队协同算法的研究大多数停留在数据信息融合、态势估计、协同目标分配、协同火力分配、协同航路规划、协同对地打击目标以及编队队形保持等问题上，鲜有涉及实时多机战术协同的研究。针对这种情况，本文在已有的单机战术决策系统架构的基础上研究适用于多机战术决策的系统架构，主要研究工作包括态势估计、战术决策以及机动决策。在已有的态势估计方法中，基于贝叶斯网络模型的方法应用最广泛，但是这种方法除了概率之间必须满足加性原则，还需要构建大量的条件概率表，降低了算法扩展性。除此之外，随着节点的增加，很难保证系统的实时性。针对这些问题，本文基于不受可加性限制的D-S证据理论，设计了多层次态势估计模型。为了降低人为设定证据初始不确定度对系统鲁棒性的影响，本文采用模糊映射计算证据初始不确定度。此外，为了保证系统实时性，采用了基于聚类的证据裁剪方法，试验证明在不降低态势估计有效性的同时保证系统实时性。空战过程难以用特定的数学模型表达，本文针对一架有人机和一架无人机的情况，基于大量的文献介绍，总结了巡航、接敌、进攻和撤退四个主要空战阶段的战术规则，并利用第三方软件CPN Tools实现了基于Petri网的战术决策推理过程。CNP Tools的可视化建模方式使得系统设计和推理过程更为直观，支持子图的功能降低了系统不同模块之间的耦合性，从而一定程度上保证了战术规则库的维护性和扩展性。空战环境复杂多变，无人机有可能遭遇突发威胁，在长机无法及时处理这类紧急状况时，无人机很可能因为无法规避威胁而坠毁。为了避免这种情况，本文根据飞行员训练经验建立了表征空战经验的数学模型，使无人机能够在得到自主权后实现自主机动决策。仿真结果表明，在与有人操作的模拟机的对抗中，无人机的规避敌机甚至击落敌机的概率超过90%，说明该机动决策方法符合无人机视距内自主空战要求。在上述研究工作的基础上，采用多四旋翼演示平台验证该多机编队协同战术决策系统的合理性和有效性。
英文摘要	Cooperative combat of multi-aircrafts formation raises worldwide concern, it gains solid theory foundation provided by advanced artificial and weapon support provided by the development of air to air missiles. Related researches of the cooperative combat among the manned aircrafts and the unmanned aircrafts are treated as the primary research direction of air combat form in the future by many countries. Many existing algorithms of cooperative combat of multi-aircrafts formation focus on areas such as data fusion, situation assessment, target assignment, firepower assignment, route planning, suppression of enemy air defense and maintaining formation all collaboratively, rarely involve real time tactics and maneuvering decision for multi-aircrafts. In view of this situation, herein an architecture is proposed based on tactics architecture of one-to-one air combat, the primary research work mainly includes three parts: situation assessment, tactics decision and maneuvering decision. In the existing situation assessment methods, the Bayes network model is mostly used, yet this method has to meet the additivity condition as well as constructing many conditional tables, which reduces the extensibility of the method. Besides, with the increasing of nodes, it is difficult to ensure a real-time system. To solve these problems, herein a situation model in multi-level based on D-S theory, which is free from additivity condition, is proposed. In order to reduce the influence of man-made evidences on system robustness, fuzzy functions are used to calculate the initial uncertainty of evidences. In addition, the use of clustering to cut evidences ensures the real-time system. Experiment results shows that this situation model is able to maintain real-time ability and effectiveness at the same time. Combat process is hard to be modeled mathematically, for the case of cooperation between one manned aircraft and one unmanned aircraft, herein rules of four phases in air combat are summarized, including cruising, approaching, attacking and retreating, and a third party software, CPN Tools, ensures the reasoning process of rules designed on Petri net. At the same time, the abilities of visualizing design and supporting sub graphs reduce the coupling among various parts and in some extent improves the extensibility and maintainability of the rules data base. Combat situation is complex and changeable, UAVS will meet pop-up threats at any time. When the leader is unable to deal with the emergencies, UAV is easy to be attacked and crash. To avoid this situation, herein a mathematical model of pilots’ training experiences is proposed, which ensures the automatic maneuvering decision of the unmanned wingmen. The experiment results show that in the combat with human handled simulation aircraft, the UAV wins more than 90% of the game, which indicates that this model is in accord with the requirements of air combat. Base on the existed research work, a multi quadcopters platform is used to test the effectivity and rationality of the cooperation formation air combat tactics decision system of multi aircrafts. The results show that, in the barrier free environment and slow speed, the platform is able to make reasonable tactics decision, indicating the reasonability and effectivity of the decision system.
关键词	态势估计战术决策机动决策 D-s Cpn Tools
文献类型	学位论文
条目标识符	http://ir.ia.ac.cn/handle/173211/11736
专题	毕业生_硕士学位论文
作者单位	中国科学院自动化研究所
第一作者单位	中国科学院自动化研究所
推荐引用方式 GB/T 7714	贺雪梅. 协同编队自主空战战术决策方法研究[D]. 北京. 中国科学院研究生院,2016.

条目包含的文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
协同编队自主空战战术决策方法研究.pdf（3985KB）	学位论文		限制开放	CC BY-NC-SA