A fully and discriminatorily informed particle swarm optimization with different sharing strategies for superior and inferior information

Abstract What kind of information is used and in which way the particles interact with each other have direct effects on the particle swarm optimization (PSO) algorithm efficiency. In order to use the information more fully, effectively and reasonably, the paper proposes a fully and discriminatorily informed PSO (FDIPSO) algorithm. Unlike the traditional PSO algorithm, the algorithm takes the positive effects of all the superior neighbors and the negative effects of the inferior particles into account and employs different mechanisms for attractive and repulsive effects to prevent confusion of swarm evolution caused by the different effects. The superior information is fully used to build high quality equilibrium points as attractors to guide particles while the repulsive effect from inferior information is embodied by introducing an ‘escape coefficient’ to help adjust the movement of particles. Experimental studies are conducted on a set of well-known benchmark functions including unimodal, multimodal and rotated problems. Computational results show positive effect and negative effect work collaboratively in FIDPSO, verify the relative superiority of this strategy over four other information sharing strategies and indicate that the approach outperforms several other state-of-art PSO variants on the test problems.