Improved particle swarm optimization algorithm with random mutation and perception

带随机变异及感知因子的粒子群优化算法

¹ Fundamentals Department, Air Force Engineering University, Xi'an 710051, China
² School of Air and Missile Defense, Air Force Engineering University, Xi'an 710051, China

Received: 22 June 2022

Abstract

Since traditional particle swarm optimization(PSO) is prone to premature phenomenon when solving complex functions in high-dimensional space, a particle swarm optimization algorithm with random variation and dynamic perception factors in terms of the movement laws and dispersion characteristics of particles in space is proposed. In order to encourage individual particles to explore their own neighborhoods and reduce the premature phenomenon of particles due to over-reliance on individual optimality and global optimality, a random mutation factor with a questioning strategy for neighborhoods is added to the basic algorithm to improve the speed update. At the same time, a perception factor is added to the particle position update, so that the particle can dynamically and adaptively control the spatial distance between itself and other particles in the same dimension, so as to avoid falling into local optimum. The algorithm has obvious superiority and robustness in solving complex functions in high-dimensional space through test function experiments, algorithm comparison analysis experiments, random parameter influence experiments and algorithm complexity experiments.

摘要

针对传统粒子群算法(PSO)在求解高维空间中复杂函数时容易发生"早熟"现象, 根据粒子在空间中的运动规律和散布特点, 提出带随机变异因子和动态感知因子的粒子群优化算法。算法通过引入对邻域具有质疑策略的随机变异因子, 促使个体粒子对自身邻域进行探索, 降低粒子因过于信赖个体最优和全局最优而发生的"早熟"现象, 从而改进速度更新策略; 同时, 为粒子位置更新引入感知因子, 使粒子在同一维度上动态自适应控制自身与其他粒子的空间距离, 从而避免陷入局部最优。通过测试函数实验、算法对比分析实验、随机参数影响实验和算法复杂性实验, 验证了该算法在求解高维空间中的复杂函数等问题时, 具有明显的优越性和鲁棒性。