【作者】 李佳；

【导师】 李智；

【作者基本信息】 四川大学 ， 电路与系统， 2021， 硕士

【摘要】 在过去的十年中,无人机被广泛的应用在各种军事任务中,如侦察、打击、评估任务。当前执行任务的无人机更多的是单一机种,而在未来,将会是种类多样、能力多样的异构多无人机协同执行任务。异构多无人机协同面临着许多挑战,但其中最重要的是异构多无人机协同任务分配问题。通常,任务分配的主要目的在于寻找合理和可行的分配方案,以便将不同的任务分配给可用的无人机。协同多任务分配问题(CMTAP)是一个NP难组合优化问题。有效的协同任务分配算法可以在满足作战需求的情况下制定多无人机系统的任务计划。目前,针对单机种的无人机任务分配算法有很多,但是针对异构多无人机协同任务分配的研究还比较少。另外,随着无人机的广泛使用,无人机执行任务的环境也是多种多样的,当前还没有任何一种算法能够解决任何环境下异构多无人机任务分配问题。本文为了解决这个问题,提供了另外的一种思路:任务分配算法不能是一成不变的,应该根据不同的任务环境来应用不同的分配算法。因此本文着重研究了不同任务环境下的异构多无人机协同任务分配问题,其主要工作创新研究成果如下:(1)针对静止环境下的异构多无人机的CMTAP,提出了一种自适应遗传算法(AGA)。虽然遗传算法已被广泛应用于求解CMTAP,但现有的方法不能在具有任务耦合约束的异构多无人机协同任务分配条件下直接应用。为此,引入多型基因建立无死锁的染色体编码策略,所提出的染色体编码策略保证了生成的染色体满足无人机异构性和任务耦合约束。根据染色体编码策略,改进了交叉和变异操作,以保证遗传算法种群的可行性。此外,该算法还根据迭代时间动态调整交叉和变异概率。(2)针对动态环境下的异构多无人机的CMTAP,提出一种改进的基于一致性束算法。基本的基于一致性束算法是一种只考虑同构代理和独立任务的分散任务分配方法。为了解决无人机的异构性和任务耦合约束,在内部共识阶段开始时提出一个Can-do列表,记录每个无人机满足任务耦合约束的任务,每个无人机只会对在“Can-do”列表上的任务出价。为了解决不同的任务重要性不同的问题,改进了得分函数,确保重要的任务能够被优先分配。为了处理任务的时间约束,采用任务执行时间列表存储无人机分配任务的执行时间。然后使用改进的共识策略来检查耦合任务的优先约束。内部共识达成后,外部共识阶段通过引入插入位置可行性指数来处理约束违背。最后,通过实验验证了所提出ICBBA具有更好的收敛时间和收敛性能。(3)针对在复杂对抗环境下无人机可能遭遇破坏,导致分配的任务执行失败的情况,提出一种对抗环境下无人机机群的损伤容错分配算法。目前,对具有容错机制的无人机任务分配算法的研究文献还比较少。因此,本文首先建立了无人机和任务的数学模型,提出了一种新的无人机损伤模型。与现有的分配算法不同,所提出的算法不仅分配任务,而且在需要时对初始任务进行备份,然后分配给无人机,以提高任务被成功执行的可能性。然后,提出了两种任务备份机制,备份-主重叠和备份-备份重叠的重叠机制,并对其属性进行了详细分析。在容错和重叠机制的基础上,提出了一种新的容错任务分配算法,旨在提高任务成功概率。针对一些重要的任务,提出动态调整备份阈值的思想,以保证重要任务在任何情况下能够被备份。最后通过实验证明所提出无人机损伤容限分配算法的可以有效的提高任务分配成功率。更多还原

【Abstract】 Over the past decade,UAVs have been widely used in a variety of military missions,such as reconnaissance,strike,and assessment.At present,the task of unmanned aerial vehicles is more of a single type,but in the future,there will be a variety of types and capabilities of heterogeneous multi-unmanned aerial vehicles to carry out the task.Heterogeneous multi-unmanned aerial vehicles(UAVs)cooperation faces many challenges,but the most important one is task assignment.Usually,the main purpose of task assignment is to find a reasonable and feasible assignment scheme so that different tasks can be assigned to the available UAVs.Cooperative multitasking assignment problem is a NP difficult combinatorial optimization problem.Efficient collaborative task assignment algorithm can make mission plan of multiunmanned aerial vehicle system under the condition of meeting operational requirements.At present,there are many task assignment algorithms for single UAV,but there are few researches on heterogeneous multi-UAV cooperative task assignment.In addition,with the wide use of UAVs,UAVs perform tasks in a variety of environments.At present,there is no algorithm that can solve the problem of heterogeneous multiUAV task allocation in any environment.In order to solve this problem,this article provides another way of thinking: task assignment algorithm can not be invariable,we should apply different assignment algorithm according to different task environment.Therefore,this paper focuses on the problem of heterogeneous multi-UAV cooperative task assignment under different task environments,and its main innovative research results are as follows:(1)An adaptive genetic algorithm(AGA)was proposed to solve the task assignment problem of heterogeneous multi-UAV in static environment.Although genetic algorithm has been widely used to solve CMTAP,the existing methods cannot be directly applied in heterogeneous multi-UAV cooperative task assignment with task coupling constraints.Therefore,multi-type genes were introduced to establish a deadlock-free chromosome coding strategy.The proposed chromosome coding strategy ensured that the generated chromosomes could meet the constraints of heterogenicity and mission coupling of UAV.According to the chromosome coding strategy,the crossover and mutation operators are improved to ensure the feasibility of genetic algorithm population.In addition,the crossover and mutation probabilities are dynamically adjusted according to iteration time.(2)Aiming at CMTAP of heterogeneous multi-UAV in dynamic environment,an improved the consensus-based bundle algorithm is proposed.The basic congruencebeam based algorithm is a decentralized task assignment method which only considers isomorphic agents and independent tasks.In order to solve the heterogeneity and taskcoupling constraints of UAVs,a can-do list was proposed at the beginning of the internal consensus stage to record the tasks of each UAV meeting the task-coupling constraints.Each UAV would only bid for the tasks on the "can-do" list.In order to solve the problem that different tasks have different importance,the scoring function is improved to ensure that important tasks can be assigned in priority.In order to deal with the time constraints of the task,the task execution time list is used to store the execution time of the task assigned by the UAV.The improved consensus strategy is then used to check the precedence constraints of the coupled tasks.After the internal consensus is reached,the external consensus phase deals with the constraint violation by introducing the feasibility index of the insertion position.Finally,the proposed ICBBA has better convergence time and performance through experiments.(3)Aiming at the situation that the UAVs may suffer damage in the complex countermeasures environment,which leads to the failure of the assignment task,a damage fault-tolerant assignment algorithm of UAVs in the countermeasures environment is proposed.At present,there are few researches on the task assignment algorithm of UAV with fault tolerance mechanism.Therefore,this paper first establishes the mathematical model of UAV and mission,and puts forward a new damage model of UAV.Different from existing assignment algorithms,the proposed algorithm not only distributes tasks,but also makes a backup of the initial tasks when needed and then distributes them to the UAVs,in order to improve the likelihood that the tasks will be successfully executed.Then,two task backup mechanisms,backup-primary overlap and backup-backup overlap,are proposed,and their properties are analyzed in detail.Based on the fault-tolerant and overlapping mechanism,a new faulttolerant task assignment algorithm is proposed to improve the probability of success.Aiming at some important tasks,the idea of dynamically adjusting the backup threshold is put forward to ensure that the important can be backed up under any circumstances.Finally,the experiment proves that the proposed UAV damage tolerance assignment algorithm can effectively improve the success rate of task assignment.更多还原