求解动态优化问题的分叉PSO算法

近些年来,求解动态环境中的优化问题已经逐渐成为进化计算领域的一个新的研究热点。为了改善一般PSO算法求解这种动态优化问题的能力,现提出了一种采用分叉策略的多粒子群PSO算法。该算法能够利用一个较大的主粒子群不断搜索问题适值曲线上新的峰,而利用从主粒子群中分离出来的若干个较小的子粒子群去跟踪已经发现的峰的变化。通过对一组标准动态测试函数的实验,能够证明所提出的算法在动态环境中具有较强的鲁棒性和适应性。

Abstract：

Recently,there has been increased interest in evolutionary computation algorithms applied into dynamic environments since many real-world optimization problems are time-varying.Inspired by a forking mechanism,a new multi-swarm optimization algorithm （Forking PSO,FPSO） was proposed to enhance simple PSO’s search in dynamic landscape.In FPSO,a larger main swarm is continuously searching for new peaks and a number of smaller child swarm,divided from main swarm,are used for tracking the achieved peaks over time.Experimental study over a benchmark dynamic problem suggests that the proposed algorithm has much stronger robustness and adaptability in dynamic environments.

Forking PSO Algorithm for Dynamic Optimization Problems

Recently,there has been increased interest in evolutionary computation algorithms applied into dynamic environments since many real-world optimization problems are time-varying.Inspired by a forking mechanism,a new multi-swarm optimization algorithm （Forking PSO,FPSO） was proposed to enhance simple PSO’s search in dynamic landscape.In FPSO,a larger main swarm is continuously searching for new peaks and a number of smaller child swarm,divided from main swarm,are used for tracking the achieved peaks over time.Experimental study over a benchmark dynamic problem suggests that the proposed algorithm has much stronger robustness and adaptability in dynamic environments.