Multimodal function optimization using an improved swarm optimizer

In multimodal optimization, convergence of the basic particle swarm optimizer (BPSO) is relatively slow at the late evolution. And, particle with the best fitness may fluctuate around the globally-optimal solution, which decreases optimization precision. Therefore, an improved swarm optimizer with controllable velocity factor is proposed. On the basis of the definition of three strategies for velocity control of evolved particles, i.e. the completely random one, the partial controllable one and the completely controllable one, optimization precision and computation expense of the modified optimizers are researched comparatively by using several tracks for optimization with different velocity-changing features. Experiments show that performance of the BPSO algorithm is improved to some extent by these controllable modes for velocity-updating. Especially, those improved swarm optimizers using the completely controllable strategy are not only of high precision, but also of faster convergence, both of which imply their better overall performance in multimodal optimization.