【作者】 张涛；

【导师】 李超勇；

【作者基本信息】 浙江大学 ， 电气工程， 2023， 硕士

【摘要】 众所周知,战斗场景中的军事行动是非常复杂的,其中集群对抗的决策和控制问题对于军事冲突的自主管理至关重要,然而当前对于集群对抗决策的研究方法通常无法同时满足解的最优性和实时性,并且会将集群单元的运动规划和目标分配问题解耦,牺牲了决策整体的一体性。本文以集群对抗场景为研究对象,具体为导弹集群制导问题和异构集群智能体对抗决策问题,针对决策计算和目标分配问题展开了研究。首先,本文概述了集群对抗决策问题的研究动态,分析了需要解决的主要问题:由于当前对于集群对抗决策的研究主要将运动控制和目标资源分配问题解耦解决,这会导致解的最优性和可靠性问题,并且集群对抗过程通常对智能体的实时决策有较高的要求,有必要研究集群对抗决策的解的有效性和实时性问题。因此,本文以动态博弈策略为基本框架研究集群对抗场景的决策计算,使用快速的纳什均衡点搜索算法(Action-Reaction Search算法),并将目标任务分配问题抽象为图论问题,使用改进的Kuhn-Munkres算法进行求解,对集群同构智能体对抗问题(导弹拦截问题)和异构智能体对抗问题进行了建模和仿真。相比于传统算法,以计算博弈为基本框架的集群决策具有良好的适应性,能够很好地完成本文讨论的导弹集群拦截制导问题,并且能够自适应目标切换,满足制导过程中对于实时决策的要求。最后,面对大规模的集群对抗场景时,本文进一步从并行算法的角度高效快速解决目标分配问题,提出了Parallel Heavy Weight Matching算法,从而更好地应对大规模集群对抗的目标分配问题,并辅以纳什均衡点快速搜索算法能够实时计算出解,并且解能够保证ε纳什均衡。仿真结果验证了所提出的基本框架和搜索算法的有效性和快速性。更多还原

【Abstract】 It is well known that military operations in combat scenarios are extremely complex,and decision-making and control problems in clustered confrontations are crucial for autonomous management of military conflicts.However,current research methods for clustered confrontation decision-making often cannot simultaneously satisfy optimality and real-time performance,and decouple the problem of unit motion planning and target allocation,sacrificing the integrity of decision-making as a whole.This article focuses on the research of missile cluster guidance and heterogeneous cluster intelligence agent confrontation decision-making problems,and conducts research on decision-making and target allocation.Firstly,this article provides an overview of the research progress of clustered confrontation decision-making,and analyzes the main problems that need to be solved: because current research on clustered confrontation decision-making mainly decouples the problem of motion control and target resource allocation,this will lead to problems of optimality and reliability,and the clustered confrontation process usually has high requirements for intelligent agents’ real-time decision-making.Therefore,it is necessary to study the effectiveness and real-time performance of clustered confrontation decision-making.Therefore,this article uses dynamic game strategy as the basic framework to study decisionmaking in clustered confrontation scenarios,uses fast Nash equilibrium point search algorithm(Action-Reaction Search algorithm),abstracts target task allocation problem as a graph theory problem,uses improved Kuhn-Munkres algorithm for solution,and models and simulates homogeneous cluster intelligent agent confrontation problem(missile interception problem)and heterogeneous intelligent agent confrontation problem.Compared with traditional algorithms,clustering decision-making based on game calculation has good adaptability,can well complete the missile cluster interception guidance problem discussed in this article,and can adaptively switch targets to meet the requirements of real-time decision-making in the guidance process.Finally,when facing large-scale clustered confrontation scenarios,this article further proposes a parallel heavy weight matching algorithm to efficiently and quickly solve the target allocation problem from the perspective of parallel algorithms,thus better coping with the target allocation problem of large-scale clustered confrontation,and supporting Nash equilibrium point fast search algorithm to calculate the solution in real time,and the solution can ensureε Nash equilibrium.Simulation results verify the effectiveness and rapidness of the proposed basic framework and search algorithm.更多还原