【作者】 陈曦；

【导师】 老松杨； 汤俊；

【作者基本信息】 国防科技大学 ， 控制科学与工程， 2021， 硕士

【摘要】 近年来,无人机及多无人机协同技术迅猛发展,应用场景已经不限于早期的军事领域,众多民用领域的应用中都显现了无人机的身影。多无人机协同任务规划问题直接决定无人机任务完成的效率,因此已经成为领域内一个重要的研究方向。基于此,本文对动态目标分配、任务路径规划和任务动态重分配等多无人机协同任务规划关键技术开展研究,针对多无人机协同工作的经典场景,设计了约束条件考虑充分、评价函数设置合理的问题模型,并基于群体智能算法实现了模型求解方法。面向重点区域动态目标侦察这一重要的多无人机协同任务场景,本文从任务背景要求及任务目标的动态移动特性等方面出发,建立了多无人机动态目标侦察问题模型。基于对k-means聚类和动态最短路的改进设计,提出了多无人机动态目标侦察任务分配算法,遵循分层递解的思路求解问题。将面向过程仿真中基于事件的时间推进机制加入算法,用于动态目标轨迹预测和关键节点位置计算。仿真结果表明,本文提出的任务分配算法能够有效解决多无人机对重点区域动态目标协同侦察的任务分配问题。面向多无人机协同执行地面目标毁伤打击任务,本文研究了任务开始之前的协同路径规划问题,基于图论和多旅行商模型建立了多无人机任务路径规划问题模型。随后针对任务特性,创新性地提出了混合分组粒子群算法(HGPSO),将粒子种群划分为大小相同的子群来代表可行解,子群内设置与无人机数量相同的粒子,同时将粒子编码为路径节点。随后本文基于遗传算法思想提出了包含四种交叉变异方式的粒子群速度更新规则。仿真对比实验显示,本文提出的混合分组粒子群算法优于其他几种算法,能够高效解决多无人机任务路径规划问题。除此之外,针对任务执行过程中的各类突发情况,本文还研究了多无人机任务动态重分配问题,总结出三类态势变化场景,并将表征毁伤打击能力的无人机型号和表征防空火力能力的目标评级加入模型,使更贴合实际作战任务。随后本文以应对突发场景为导向,基于对模糊C均值聚类和蚁群算法的改进设计,提出了多场景多无人机协同任务重分配算法,包括完全再分配、局部调整和分组基础上再分配三种核心策略。最后通过仿真实验对模型参数做出优化,验证了多场景下多无人机任务动态重分配模型的功能完备性和算法可靠性。更多还原

【Abstract】 In recent years,the technology of multiple Unmanned Aerial Vehicles(UAVs)collaboration has developed rapidly.The application scenarios are no longer limited to the early military fields.UAVs have appeared in many civilian and commercial applications as well.The problem of multi-UAV collaboration mission planning directly determines the efficiency of multi-UAV mission complete,and has become a core research topic.Based on this,this paper conducts research on the key technologies of multi-UAV collaborative mission planning including dynamic target allocation,task route planning,and task dynamic re-allocation.Aiming at the classic mission scenarios of multi-UAV collaboration,this paper designs problem models with sufficient consideration of constraints and reasonable evaluation function.Solving methods based on swarm intelligence algorithms are also implemented in this paper.For dynamic target reconnaissance in key areas,which is an important multi-UAV cooperative mission scenario,this paper establishes a multi-UAV dynamic target reconnaissance problem model based on the background requirements of the mission and the dynamic movement characteristics of the target.Based on the improved design of kmeans clustering and dynamic shortest path algorithm,a task allocation algorithm for multi-UAV dynamic target reconnaissance is proposed,which solves the problem following the idea of hierarchical recursion.The event-based time advance mechanism in process-oriented simulation is added to the algorithm for dynamic target trajectory prediction and key node position calculation.The simulation results show that the algorithm proposed in this paper can effectively provide the task allocation plan in multiUAV dynamic target reconnaissance problem.For the ground target strike mission by multiple UAVs,this paper studies the cooperative path planning problem before the beginning of the mission,and establishes a multi-UAV mission path planning problem model based on graph theory and multiple travelling salesman model.Then,this paper innovatively proposes a hybrid grouped particle swarm algorithm based on the task characteristics.The algorithm divides the particle population into subgroups of the same size to represent feasible solutions.The number of particles contained in the subgroups is the same as the number of drones,and the particles are coded as path nodes.Subsequently,based on the idea of genetic algorithm,this paper proposes a particle swarm velocity update rule containing four cross-mutation methods.Simulation comparison experiments show that the hybrid grouped particle swarm algorithm proposed in this paper is better than several other improved particle swarm algorithms,and can efficiently solve the problem of multi-UAV mission path planning.In addition,in response to various emergencies during mission execution,this paper also studies the dynamic re-allocation of multi-UAV missions.Scenarios of situation change are summarized into three categories.The UAV type that characterizes strike capability and the target grade that characterizes air defense firepower are added to the model to make it more suitable for actual missions.Subsequently,oriented to respond to emergencies,this paper proposes a multi-scenario multi-UAV cooperative task reallocation algorithm based on the improved design of fuzzy C-means clustering and ant colony algorithm.The proposed algorithm includes three core strategies: complete reallocation,partial adjustment and re-allocation on subgroups basis.Finally,the model parameters are optimized through simulation experiments.The experimental results verify the functional completeness and algorithm reliability of the multi-scenario multiUAV cooperative task re-allocation model.更多还原