Dynamic multi-swarm pigeon-inspired optimisation

Abstract

Pigeon-inspired optimisation (PIO) has shown favourable performance on global optimisation problems. However, it lacks the part of individual experience, which makes it prone to premature convergence when solving multimodal problems. Moreover, the landmark operator model in PIO may cause the population size to decrease too quickly, which is harmful for exploration. To overcome the shortcomings, a dynamic multi-swarm pigeon-inspired optimisation (DMS-PIO) is proposed in this research. In PIO, the entire population is divided into multiple swarms. During the evolutionary process, the size of each swarm can be dynamically adjusted, and the multiple swarms can be randomly regrouped. Relying on the dynamic adjustment of swarms' sized, exploration and exploitation are balanced in the initial evolutionary stage and last stage. Furthermore, the randomly regrouping schedule is used to keep the population diversity. To enhance the comprehensive performance of PIO, the map and compass operator and the landmark operator in it are conducted alternately in each generation. Experimental results between DMS-PIO and other five PIO algorithms demonstrate that our proposed DMS-PIO can avoid the premature convergence problem when solving multimodal problems, and yields more effective performance in complex continuous optimisation problems.