引用本文:王峰,付青坡,韩孟臣,邢立宁,吴虎胜.LeCMPSO算法求解异构无人机协同多任务重分配问题[J].控制理论与应用,2024,41(6):1009~1017.[点击复制]
WANG Feng,FU Qing-po,HAN Meng-chen,XING Li-ning,WU Hu-sheng.Learning-guided coevolution multi-objective particle swarm optimization for heterogeneous UAV cooperative multi-task reallocation problem[J].Control Theory and Technology,2024,41(6):1009~1017.[点击复制]
LeCMPSO算法求解异构无人机协同多任务重分配问题
Learning-guided coevolution multi-objective particle swarm optimization for heterogeneous UAV cooperative multi-task reallocation problem
摘要点击 704  全文点击 177  投稿时间：2022-07-25  修订日期：2023-12-28
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  DOI: 10.7641/CTA.2023.20665
  2024,41(6):1009-1017
中文关键词  无人机多任务重分配  粒子群优化算法  多目标优化  协同进化
英文关键词  UAV multi-task reallocation  particle swarm optimization  multi-objective optimization  coevolution
基金项目  国家自然科学基金项目(62173258)资助.
作者单位E-mail
王峰* 武汉大学 计算机学院 fengwang@whu.edu.cn 
付青坡 武汉大学 计算机学院  
韩孟臣 武汉大学 计算机学院  
邢立宁 国防科技大学 系统工程学院  
吴虎胜 武警工程大学 装备管理与保障学院  
中文摘要
      无人机系统在军事领域有着广泛应用, 由于战场环境复杂多变, 无人机遭遇突发状况后需进行任务重分配.异构无人机是指多种类型的无人机, 可完成单一无人机无法完成的多类型复杂任务, 异构无人机协同多任务重分配问题约束条件复杂且包含混合变量, 现有多目标优化算法不能有效处理此类问题. 为高效求解上述问题, 本文构建多约束异构无人机协同多任务重分配问题模型, 提出一种学习引导的协同多目标粒子群优化算法(LeCMPSO), 该算法引入基于先验知识的初始化策略和基于历史信息学习的粒子更新策略, 能有效避免不可行解的产生并提升算法的搜索效率. 通过在4组实例上的仿真实验表明, 与其他典型的协同进化多目标优化算法相比, 所提算法在解集的多样性、收敛性及搜索时间方面均具有较好的性能.
英文摘要
      UAV system has been widely used in military field. Due to the complex and changeable battlefield environment, UAV tasks need to be reassigned after an emergency. Heterogeneous UAVs refer to multiple types of UAVs, which can accomplish multiple types of complex tasks that a single UAV can not. The heterogeneous UAV cooperative multi-task reallocation problem has complex constraints and mixed variables, and the existing multi-objective optimization algorithms can not deal with this kind of problems effectively. In order to solve the above problems efficiently, a multi-constraint heterogeneous UAVs cooperative multi-task reallocation model is constructed at first in this paper, and a learning-guided cooperative multi-objective particle swarm optimization algorithm (LeCMPSO) is proposed to solve that. In LeCMPSO, a prior knowledge based initialization strategy as well as a history information learning based particle update strategy are introduced to avoid the generation of infeasible solutions and improve the search efficiency of the algorithm. The simulation results on 4 sets of examples show that the proposed algorithm outperforms the other typical coevolutionary multi-objective optimization algorithms on diversity of solution sets, convergence, and search time.