基于多种群自适应遗传的异构无人机集群动态任务分配算法  

Multi-population adaptive genetic algorithm for dynamic task allocation in heterogeneous UAV swarms

作　　者：王传云[1] 胡安琪 高骞 刘琼[2] 王田 Wang Chuanyun;Hu Anqi;Gao Qian;Liu Qiong;Wang Tian(College of Artificial Intelligence,Shenyang Aerospace University,Shenyang 110136,China;School of Automation,Beijing Information Science and Technology University,Beijing 100192,China;Institute of Artificial Intelligence,Beihang University,Beijing 100191,China)

机构地区：[1]沈阳航空航天大学人工智能学院,沈阳110136 [2]北京信息科技大学自动化学院,北京100192 [3]北京航空航天大学人工智能研究院,北京100191

出　　处：《战术导弹技术》2024年第6期45-56,共12页Tactical Missile Technology

基　　金：国家自然科学基金(61703287,62302051,61972016);辽宁省自然科学基金计划(2024-MS-137);辽宁省教育厅科研计划(LJKZ0218,LJKMZ20020556)。

摘　　要：异构无人机集群在复杂任务环境中具备多层次协同作业的优势,现有的任务分配方法仍面临如易陷入局部最优解、对任务属性和数量变化的适应性不足以及多目标优化效率低等诸多挑战。为此,提出一种基于多种群自适应遗传的动态任务分配算法,该算法根据染色体的初始目标函数特性将总种群划分为若干子种群。各子种群在迭代过程中独立进行自适应遗传优化,智能调整变异概率及选择交叉点。每次迭代结束前,算法采用优胜劣汰策略保持染色体多样性及较优秀染色体结构。若任务属性或数量发生变化,算法将灵活根据变化情况重新划分子种群或增加染色体数量。与基于遗传优化的变体算法相比,本算法在不同类型基准函数上表现出较高的稳定性。在处理复杂场景时,本算法的目标函数值提升44.15%,收敛速度提升59.85%,求解稳定性提升53.99%。该算法具有卓越稳定的求解能力,且能够在更短的计算时间内更有效地避免陷入局部最优解。Heterogeneous UAV swarms are demonstrated to have advantages in multi-level collaborative operations within complex task environments.However,significant challenges are still faced by existing task allocation methods,such as susceptibility to local optima,poor adaptability to changes in task attributes and quantities,and low efficiency in multi-objective optimization.To address these issues,a dynamic task allocation algorithm based on multi-population adaptive genetic optimization is proposed.The total population is divided into several sub-populations based on the initial objective function characteristics of the chromosomes.Independent adaptive genetic optimization is performed on each sub-population during the iterations,with mutation probability and crossover points intelligently adjusted.Before the end of each iteration,a survival-of-the-fittest strategy is applied to maintain chromosome diversity and preserve the structure of superior chromosomes.In response to changes in task attributes or quantities,the sub-populations are flexibly re-divided or the number of chromosomes is increased based on the new conditions.Compared to variant algorithms based on genetic optimization,higher stability is exhibited by this algorithm across different types of benchmark functions.In complex scenarios,the proposed algorithm results in a 44.15%improvement in the objective function value,a 59.85%increase in convergence speed,and a 53.99%enhancement in solution stability.Excellent and stable problem-solving capabilities are demonstrated by the algorithm,effectively avoiding local optima within shorter computation times.

关 键 词：异构无人机集群 动态任务分配 协同作业 多目标优化 遗传算法 参数自适应 收敛性能 局部最优解 

分 类 号：V279[航空宇航科学与技术—飞行器设计]