【作者】 苏菲；

【导师】 沈林成；

【作者基本信息】 国防科学技术大学 ， 控制科学与工程， 2013， 博士

【摘要】 多无人作战飞机（Unmanned Combat Aerial Vehicle, UCAV）协同任务规划技术是实现多UCAV在复杂战场环境中协同作战，提高多UCAV协同作战效能的关键技术之一。论文以多UCAV协同执行压制敌防空系统（Suppression of EnemyAnti-air Defense, SEAD）任务为背景，主要围绕有人机管理多UCAV协同执行SEAD和多UCAV自主协同执行SEAD两种典型任务模式，重点研究了多UCAV在线协同任务规划问题的建模及优化方法，并开展了仿真与实验验证。主要工作及创新点如下：（1）分析了动态环境下多UCAV协同执行SEAD任务的作战过程，建立了作战任务执行过程及相应规划过程中涉及的相关要素模型，提出了两种典型SEAD任务模式下多UCAV协同系统的分布式体系结构。从多UCAV协同SEAD任务的执行过程出发，在深入分析多UCAV协同SEAD任务独特性的基础上，对其执行及规划过程中涉及的各种任务要素进行形式化描述，建立了包括UCAV平台、任务载荷、战场威胁实体在内的多要素形式化模型。针对多UCAV协同SEAD的两种典型任务模式，提出了面向有人机管理下多UCAV协同SEAD的分布式体系结构以及面向多UCAV自主协同SEAD的分布式体系结构。基于该体系结构构建的多UCAV协同系统具有良好的伸缩性、鲁棒性和高可靠性，具备对环境的快速反应能力和灵活的系统重构能力，能够更好地适应未来高度动态化战场环境下的多UCAV协同作战的需求。（2）基于并行计算技术和协同进化思想，在异质多种群蚁群算法的框架下，面向多主体协同问题，提出了协进化多种群蚁群算法（Co-EvolutionaryMulti-Ant-Colony-Algorithm, COE-MACA）。在全面分析蚁群优化方法的基本原理、算法模型、优化特性、研究现状以及同质/异质多种群蚁群算法运行机制的基础上，根据多主体协同问题的特点，以异质多种群蚁群算法框架为基础，应用并行计算思想，将自然界多生物群落间的协同进化策略引入多个蚂蚁种群的迭代演化过程中，提出了协进化多种群蚁群算法。研究并设计了算法框架中的多种群系统结构、并行推进及信息交换策略、基于协进化机制的问题解性能评价。从提高算法优化能力的角度，引入了基于扩散机制的信息素更新机制。以多UCAV协同航迹规划问题为例，通过仿真实验对算法有效性进行验证，实验结果表明：经过算法要素的具体设计，协进化多种群蚁群算法可有效描述多主体之间的各类协同关系，通过种群演化，实现问题的快速迭代优化，且与进化算法相比，在实验想定条件下，其优化收敛速度提高了10%以上。（3）针对有人机管理多UCAV协同执行SEAD任务过程中的分布式在线协同航迹规划问题，基于分布式自适应模型预测控制（Distributed Adaptive ModelPredictive Control, DA-MPC）框架，建立了多UCAV在线协同航迹规划局部优化模型，提出了求解该模型的多UCAV分布式在线协同航迹规划算法。对多UCAV分布式协同航迹规划中的多平台协同约束进行了分析，建立了基于协同系数的多UCAV协同航迹代价模型；基于分布式模型预测控制（DistributedModel Predictive Control, DMPC）思想，结合任务自身特点，提出了基于代价势场的UCAV航迹终端罚函数及相应的代价势场生成算法，从理论上证明了所构建代价势场不存在局部极小值的特性，有效解决了传统势场构建方法中存在的“死锁”问题；以此为基础，建立了多UCAV在线协同航迹规划的分布式局部优化模型，将多UCAV协同系统大规模复杂优化问题分解为对应于各UCAV子系统的较小规模局部优化问题，从系统层面缩减问题空间；针对该局部优化模型的求解问题，提出了基于COE-MACA和滚动优化机制的局部优化算法，将多UCAV协同航迹规划问题在系统层和时间层上进行分解和优化，实现了多UCAV的分布式在线航迹规划。仿真实验结果表明：所提出的分布式局部优化模型及其求解算法能够实现原问题在不同层面的分解，降低多UCAV在线协同航迹规划的计算复杂度，减少规划时间，在所进行的实验统计中，论文提出的算法规划时间约为分布式条件下全局优化方法的10%，为集中式条件下局部优化方法的40%；同时，该算法能根据规划的具体需要，在时间性能和解的最优性之间进行权衡，对诸如突发威胁规避、突发威胁压制、威胁状态变化、任务目标变化等动态环境因素具有较好的适应能力。（4）面向多UCAV自主协同SEAD任务，建立了基于DMPC的多UCAV任务协调局部优化模型，提出了基于COE-MACA和滚动优化机制的多UCAV分布式在线任务协调局部优化算法。基于分层递阶思想，将多UCAV自主协同SEAD任务中的任务规划分解为任务协调层和航迹规划层两部分。针对多UCAV协同系统的在线任务协调问题，建立了基于灰色理论的任务目标评估模型和基于加权有向图的任务转移代价估计模型，设计了各UCAV子系统的规划窗口构造与更新机制，建立了基于DMPC的多UCAV任务协调局部优化模型，综合COE-MACA和滚动优化机制提出了多UCAV分布式在线任务协调局部优化算法。仿真实验结果表明：分布式在线任务协调局部优化算法能够有效地处理多UCAV多任务之间的协同关系，通过滚动优化机制的引入实现了复杂问题的分解，能够对新增目标、目标取消、UCAV故障或战损等动态因素进行合理响应，满足动态环境下多UCAV在线协同任务规划的要求。更多还原

【Abstract】 Cooperative mission planning for multiple Unmanned Combat Aerial Vehicles(UCAVs) is one of the key technologies to realize the cooperative operation for multipleUCAVs in complicated battlefield environment and enhance the efficiency ofmulti-UCAVs cooperation. This dissertation focuses on the online mission planningproblem of typical multi-UCAVs SEAD mission in dynamic battle fields, in terms ofmulti-UCAVs cooperative SEAD under the management of manned fighters andmulti-UCAVs independent cooperative SEAD, builds mathematical model, researchesand designs optimization algorithm, and carries out simulation and experiments. Themain work and the creative contribution are as follows:(1) The processes of multi-UCAVs cooperative SEAD mission are analyzed, themodels of elements which are related to the mission executing processes and planningprocesses are built, and the distributed structure of multi-UCAVs cooperative systemsfor two typical SEAD mission modes are presented.According to the processes of multi-UCAVs cooperative SEAD mission, therelated elements are described, the formalized model including UCAV platforms,mission payloads, menace entities are presented based on the characteristics ofmulti-UCAVs cooperative SEAD. Aiming at the two typical modes of multi-UCAVsSEAD, the distributed structures for multi-UCAVs cooperative SEAD under themanagement of manned fighters and multi-UCAVs independent cooperative SEAD arepresented. The multi-UCAVs system based on the presented structures is highly robustand credible, and has sensitive reaction to the environments, flexible ability of systemrebuilding; therefore, it can satisfy the requirement of multi-UCAVs cooperativeoperation in highly dynamic battle fields.(2) Based on the parallel computing technology and the idea of co-evolutionary,the Co-Evolutionary Multi-Ant-Colony-Algorithm (COE-MACA) is proposed under theframework of multi-difference-ant-colony algorithm to solve the cooperation problemswith multiple behavior units.With considerations on the basic principle, algorithm model, characteristics ofoptimization, current situation of ant colony optimization research, and functionalmechanism of multi-ant-colony algorithm, the Co-EvolutionaryMulti-Ant-Colony-Algorithm is proposed under the framework ofmulti-difference-ant-colony algorithm. The algorithm is designed according to thecharacteristics of cooperation problem with multiple behavior units, the parallelcomputing technology is used to guide the design of COE-MACA, and theco-evolutionary strategy is introduced to the evolvement of multiple ant colonies. Themultiple colonies structure, strategies of parallel progress and information exchange, results evaluation based on co-evolutionary mechanism is studied, and the pheromonediffusion mechanism is used to improve the optimization capability of COE-MACA.The simulation of multi-UCAVs cooperative path planning using COE-MACA iscarried out to confirm the efficiency of the proposed algorithm, the experimental resultsindicate that the COE-MACA has good performance in describing the cooperativerelations between multiple behavior units by designing the elements of algorithm, andthe solution can be optimized rapidly through the evolution of ant colony. Compared tothe evolutionary algorithm, the convergence speed of COE-MACA increases at least10%in specific experimental scenario.(3) Based on the framework of Distributed Adaptive Model Predictive Control(DA-MPC), the local optimization model of multi-UCAVs online cooperative pathplanning is presented to solve the online cooperative path planning of multi-UCAVscooperative SEAD under the management of manned fighters, and the distributed onlinecooperative path planning algorithm to solve this model is proposed.The cooperative constraints in multi-UCAVs distributed cooperative online pathplanning are analyzed, and the cost model of multi-UCAVs cooperative path based oncooperative coefficient is presented; based on the distributed model predictive control,the UCAV path terminal penalty function based on the cost field and its generatingalgorithm is proposed. It is proofed that the cost field has only one local minimal pointat the specific target point. Then the local optimization model of multi-UCAVs onlinecooperative path planning is proposed. In this model, the complex optimization problemof multi-UCAVs cooperative system is decomposed into several local optimizationproblems corresponding to each UCAV, which can reduce the complexity of originalplanning problem on the system level. To solve the proposed model, the localoptimization algorithm based on COE-MACA and RHC is presented. The algorithmseparates and optimizes the original planning problem on the levels of system as well astime, which makes multi-UCAVs distributed online path planning easier to realize.Simulation results indicates that the proposed model and algorithm can separates theoriginal problem on different level, reduces the complexity and time cost ofmulti-UCAVs online cooperative path planning, in statistics of the simulation results,the planning time of proposed algorithm is10%of the distributed global optimizationalgorithm, and40%of the centralized local optimization algorithm, and furthermore, theproposed algorithm can balance the weight between time cost and optimized solution,and has good performance in response to dynamic environment such as avoiding suddenmenaces, suppressing sudden menaces, menaces states change, targets change and soon.(4) To solve the online cooperative mission planning problem of multi-UCAVsindependent cooperative SEAD, a local optimization model for multi-UCAVs missioncoordination is proposed based on DMPC, and a distributed local optimization algorithm to solve the local optimization model is presented based on COE-MACA andRHC.Based on the hierarchical control, multi-UCAVs independent cooperative SEADmission planning is divided into mission coordination and path planning. To solvemulti-UCAVs mission coordination problem, the mathematical models of mission/targetevaluation and transfer cost estimation between different missions are studied based onthe grey system theory and weighted oriented graph. The strategies to construct andupdate UCAV planning windows are designed, an mathematical local optimizationmodel is proposed and the distributed local optimization algorithm for multi-UCAVsmission coordination is presented under the framework of COE-MACA and RHC.Simulation results shows that the mentioned algorithm can manage the cooperativeconstraints between different UCAVs and missions effectively, the introduction of RHCdecomposes the original problem to reduce the complexity, and the algorithm caneffectively response to the dynamic event such as target increasing, target cancelled,UCAV failure or damage quickly, so that the mentioned model and algorithm cansatisfy the requirement of multi-UCAVs online mission coordination in dynamicenvironment.更多还原