一种基于DQN的改进NSGA-Ⅱ算法  

An Improved NSGA-Ⅱ Algorithm for Bullet Distribution

作　　者：吴亚楠 张剑[1] WU Yanan;ZHANG Jian(Wuhan Digital Engineering Institute,Wuhan 430205)

机构地区：[1]武汉数字工程研究所,武汉430205

出　　处：《舰船电子工程》2023年第4期29-33,共5页Ship Electronic Engineering

摘　　要：采用传统单种群NSGA-Ⅱ算法求解武器目标分配多目标优化数学模型,解算过程中容易陷入局部最优,且存在分布性不足、求解时间长等缺陷。为了改善算法性能,获得更好的Pareto最优解集结果,可结合深度强化学习和双种群“迁徙”思想对传统单种群NSGA-Ⅱ算法进行改进。采用深度强化学习DQN算法来对双种群“迁徙”操作中涉及到的各项迁徙参数进行调整优化,对深度强化学习要素进行设计,并通过实验验证改进后的NSGA-Ⅱ算法具有更好的算法性能,且算法耗时更少,表明论文改进NSGA-Ⅱ算法在求解武器目标分配问题上的有效性。The traditional single-population NSGA-Ⅱ algorithm is used to solve the mathematical model of multi-objective op⁃timization of weapon target allocation.It is easy to fall into local optimal,and has defects such as insufficient distribution and long solving time.In order to improve the algorithm performance and obtain better Pareto optimal set results,the traditional single-popula⁃tion NSGA-Ⅱ algorithm can be improved by combining deep reinforcement learning and two-population"migration"idea.The DQN algorithm of deep reinforcement learning is used to adjust and optimize the migration parameters involved in the"migration"opera⁃tion of the two-population,and the elements of deep reinforcement learning are designed.The experiment shows that the improved NSGA-II algorithm has better algorithm performance and less algorithm time.It shows the effectiveness of the improved NSGA-II al⁃gorithm in solving the weapon target assignment problem.

关 键 词：弹目分配 改进NSGA-Ⅱ算法 双种群迁徙 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]