Dynamic bee colony algorithm based on multi-species co-evolution

摘要

Most versions of the current Artificial Bee Colony (ABC) algorithm are based on single-species model. It is insufficient to extend the diversity of solutions and may be trapped into local extreme points at the end of the evolutionary process. In this paper, an enhanced version (Co-ABC) of the original ABC algorithm is proposed based on multi-species co-evolution. There are three versions in the Co-ABC algorithm, including the CABC, TABC and IABC algorithm. These three algorithms combine evolutionary strategies including the dynamic dividing of the swarm and the cooperative evolution with the original ABC algorithm. The CABC and TABC algorithm form the basis of the IABC algorithm. The CABC algorithm uses a global communication method to accelerate convergence, while the TABC algorithm uses a local communication method to improve accuracy. The IABC algorithm combines the convergence superiority in the CABC algorithm with the accuracy superiority in the TABC algorithm, obtaining complementary advantages. In the optimization process of the Co-ABC algorithm, the solution space shrinks dynamically according to the employed norm, while the subgroups adjust to environment adaptively to select the superior and eliminate the inferior. It can reduce computational complexity in optimization problems. The experimental results show that the Co-ABC algorithm can exhibit good performance on convergence, accuracy and robustness in most optimization problems. The CABC algorithm is suitable for the unimodal optimization problems with small dimensions, while the IABC algorithm exhibits good performance in multimodal problems and unimodal problems with large dimensions.