摘要
针对粒子群优化算法存在的全局探索与局部开发不平衡、易陷入局部最优等问题,本文提出一种基于强化学习与三种档案的粒子群优化算法(particle swarmoptimiz ationalgorithm based on reinforcement learning and triple archives,PSO-RLTA).该算法通过构建三类外部档案,分别用于提升种群多样性和收敛性,并在此基础上设计了四种学习策略,以增强算法在搜索过程中的自适应能力;进一步结合强化学习机制,提出一种学习策略的智能选择方法,从而加快收敛速度;此外,引入基于档案的扰动搜索策略,提升算法跳出局部最优的能力.在CEC2017基准测试函数和工程设计优化问题上的实验结果表明,PSO-RLTA在收敛速度与寻优精度方面均优于对比算法.
Aiming at the issues of imbalance between global exploration and local exploitation,and the ease of falling into local optima in particle swarm optimization algorithm,a particle swarm optimization algorithm based on reinforcement learning and triple archives(PSO-RLTA)is proposed.The algorithm is designed with three external archives based on the convergence and diversity metrics of the population,and four learning strategies are proposed based on these archives to better balance the exploration and exploitation of the algorithm.A learning strategy selection method is developed in conjunction with reinforcement learning to accelerate the convergence speed of the algorithm.A perturbation search strategy is proposed based on the external archives to enhance the population's ability to escape from local optima.Experimental results on the CEC 2017 benchmark suite and engineering design optimization problems show that PSO-RLTA delivers superior convergence speed and solution accuracy.
作者
郝晓曦
黄东娟
曾志强
HAO Xiaoxi;HUANG Dongjuan;ZENG Zhiqiang(School of Mechanical and Automation Engineering,Wuyi University,Jiangmen 529020,China;School of Electronic and Information Engineering,Wuyi University,Jiangmen 529020,China;School of Mechanical Engineering,Dongguan University of Technology,Dongguan 523808,China)
出处
《五邑大学学报(自然科学版)》
2026年第1期9-16,共8页
Journal of Wuyi University(Natural Science Edition)
基金
2022年度广东省普通高校重点领域专项项目(2022ZDZX3034)
江门市重大科技计划项目(江科(2023)117号2320002000601)。
关键词
粒子群优化
强化学习
学习策略
扰动策略
外部档案
Particle swarm optimization
Reinforcement learning
Learning strategies
Perturbation strategy
External archives
