基于寿命的粒子群算法研究

RESEARCH ON PARTICLE SWARM OPTIMISATION BASED ON LIFESPAN

针对粒子群算法易陷入局部最优的缺陷,提出了一种具有寿命的PSO(LS-PSO),算法赋予gbest有限的寿命,并且根据其引导能力对寿命进行自适应调整。当gbest耗尽其寿命时,它将失去领导能力,并被一个新产生并经测试具有足够引导能力的粒子所代替,继续引导群体搜索解空间的不同区域,并在两个单峰标准测试函数和六个多峰标准测试函数上对算法进行了测试。结果表明,LS-PSO比传统PSO及改进算法CLPSO有更好的求解精度和收敛速度。

For the problem that the particle swarm optimisation（PSO） algorithm suffers from being trapped in local optimal,a PSO with lifespan（LS-PSO） is proposed,the algorithm endues gbest with limited life and adaptively adjust the life according to its inducing ability.When gbest uses up its life,it will lose its leading ability and will be replaced by a new generated particle with sufficient inducing ability after being tested so as to continue to lead the swarm searching different spatial regions.Tests of the algorithm are conducted on two unimodal and six multimodal benchmark testing functions.The result shows that,the proposed LS-PSO outperforms the traditional PSO and the improved comprehensive learning PSO（CLPSO） with respect to the solution quality and the convergence speed.