无人机集群协同空战对抗非线性动态模型研究

The key point of command and control problem for Unmanned Aerial Vehicles(UAV)swarm conflict is to establish effective nonolinear dynamic model. This project aims to investigate dynamic process and autonomous decision-making method for UAV swarm conflict. Based on analysis of the key elements related to the evolutioin process of UAV swarm, Multi-Agent Dynamic Influence Diagram approach is employed to construct information topology of swarm conflict and Dynamic Bayesian Network is used to update constantly changing situation in order to overcome the information uncertainty.Thus the nonlinear dynamic model for UAV swarm conflict is establised. Considering the time sequence of target trajectory, a kind of knowledge discovery method based on trajectory evolution with multi-layer space metric positioin feature is proposed to deduce the tactic intention of the hostile UAV swarms. Combining this with the proposed optimized Bayesian network based on bio-inspired swarm intelligence, situation assessment for UAV swarm conflict in the complex environment can be made. Then, from the cognition .aspect, autonomous decision-making of mission planning for UAV swarm conflict is proposed. Using artificial intelligence and bio-inspired method, simulated human adaptive combat behaviors and rules are deeply studied and proposed. Based on which, further deep investigation will be made to explore and unveal the mechanism behind the emergence of the aggregate behavior of UAV individuals in the swarm conflict process. Based on HLA/RTI frame, a distributed simulation platform will be established to perform UAV swarm vs.swarm conflict simulation and to verify the effectiveness of the proposed nonlinear dynamic model for UAV swarm conflict.

无人机集群对抗问题的关键在于建立有效的对抗过程非线性动态模型，课题旨在对无人机集群对抗动态过程和自主决策方法进行研究。在分析集群对抗演化过程影响要素基础上，采用多Agent动态影响图方法构建集群对抗信息拓扑，应用动态贝叶斯网络进行态势更新以克服信息的不确定性，从而建立集群对抗的非线性动态模型。针对目标轨迹时间序列，提出一种多级空间尺度位置特征轨迹演化的知识发现方法，据此推断敌方战术意图，结合提出的基于仿生群智能方法的贝叶斯学习网络对集群对抗复杂空战环境进行态势评估，在此基础上，从认知角度，提出集群对抗任务规划和自主空战决策方法。利用人工智能和仿生学方法，研究并提出无人机的拟人自适应对抗行为方式和规则，揭示无人机集群与环境的相互作用所涌现的集群对抗表象内在机理。建立基于HLA/RTI的集群对抗仿真系统，通过无人机集群对抗仿真实验验证模型的有效性。

无人机技术的长足发展以及无人机集群的出现必将使得现代空战模式发生根本的改变。面临敌方大规模无人机集群的威胁，必须采用大规模无人机集群进行拦截，着必将发生无人机集群之间的空战对抗。无人机集群协同控制、集群对抗已成为无人机应用领域研究的热点，特别是无人机集群对抗的研究还处于起步阶段。本项目以此为背景开展集无人机群对抗模型研究。. 本文主要研究内容包括：1）无人机集群协同对抗非线性动态过程构架；2）无人机集群对抗态势评估及任务规划自主决策；3）无人机拟人自适应对抗行为模拟；4）系统仿真验证。. 本项目研究中，研究了基于分布式一致性控制的无人机集群控制方法及稳定性证明，基于针对较小规模的无人机集群对抗，利用影响图理论建立了对抗的框架，采用贝叶斯网络推理论对空战态势判断及敌机意图推断，从而进行机动决策。对于大规模无人机集群对抗，采用多Agent理论建立集群对抗框架，并设计拟人对抗行为规则，实现对抗过程的智能化、自主化。对集群空战过程进行了系统仿真。. 本项目研究在Automatica, Chinese Journal of Aeronautics, SCIENCE CHINA Technological Sciences, Neurocomputing, JOURNAL OF THE FRANKLIN INSTITUTE 等权威期刊上发表期刊论文19篇，包括：SCI收录期刊论文13篇，EI收录期刊论文2篇，北大核心期刊论文5篇。发表会议论文并被EI收录5篇，包括控制领域顶级会议CDC2018，在ICGNC、ICCA等国际会议上组织专题讨论；专利授权3项，1项在审；授权软件著作权3项。培养博士生2名，硕士生6名。撰写项目技术总结报告1本。. 本项目在集群控制、集群对抗模型建立、集群对抗决策控制方法等方面提出了新的理论方法，取得的成果为无人机集群应用和集群空战理论的发展提供了理论支撑。