基于动态贝叶斯网络的空战决策方法

摘要

针对无人机高强度空战对抗问题,开展智能决策方法研究.分析了目前无人机空战决策方法的研究进展,总结了影响空战决策方法适用性的主要原因.针对空战决策的实时性和不依赖飞机运动学模型的需求,提出了基于DBN的无人机空战决策模型.综合分析无人机可以获得的机载传感器和情报信息,从中提取出影响无人机机动决策结果的态势因素,依据各因素间的因果关系,建立了无人机空战决策模型的网络拓扑结构.阐述了网络模型有向弧条件概率表、决策节点状态转移概率的设计思路.设计了网络决策模型的前向递归推理算法,并进行了算法复杂度分析,分析表明算法时间复杂度在可接受的范围内,能够满足工程应用的要求.最后,开展典型战场态势下的单机对抗仿真实验.仿真结果验证了决策网络的适用性以及推理算法的收敛性和实时性,表明动态贝叶斯网络空战决策模型提高了无人机机动决策的有效性和灵活性.%The decision-making method is studied for high-intensity UAV air combat in this paper.The present research progress of the UAV air combat decision method is analyzed.And the main reasons that influence the applicability of the decision method are summarized.For real-time, fault-tolerance, and not relying accurate mathematical model of aircraft requirements, a decision-making model based on DBN is proposed for one-to-one UAV air combat.By synthetically analyzing the information that can be obtained from the airborne sensors and intelligence, the situational factors affecting UAV`s maneuver decisions are extracted and the DBN topology structure is constructed.The design ideas of the directed-arc conditional probability table and decision node state transition probability table of the network model are illustrated.The forward recursive reasoning algorithm for decision-making network model is designed and the algorithm complexity is analyzed.The analysis result indicates that the time and special complexity of the algorithm is acceptable, which could satisfy the requirement of engineering application.Finally, the simulation results verify the practicability of the decision network, and prove the convergence, real-time performance of the reasoning algorithm, the adaptability of network model.The results indicate that Air combat decision model improving the flexibility and validity of the UAV`s maneuver decision.