Particle swarm optimisation based on self-organisation topology driven by different fitness rank

To explore the relations between the population structure and the performance of the particle swarm optimisation (PSO), the self-organisation population structure driven by fitness rank (SOTDFR) is developed. Meanwhile, to improve the performance of PSO, under invariable network size, the SOTDFR evolution involves two kinds of operations: adding and removing link. Moreover, due to the particles' fitness rank impacting heavily on the SOTDFR evolution, two kinds of fitness rank are designed and also SOTDFR according to different fitness rank designs is referred to as VSOTDFR and UVSOTDFR respectively. To make a deep insight, VSOTDFR-based PSO and UVSOTDFR-based PSO are used to solve two types of benchmarks: unimodal and multimodal functions. Simulation results demonstrate that UVSOTDFR-based PSO can generally obtain the better solution than VSOTDFR-based PSO within the allowed iterations. In addition, the performances of the UVSOTDFR-based PSO and some variants of PSO are compared. The simulation results show that UVSOTDFR-based PSO is competitive.