Dual-archive-based particle swarm optimization for dynamic optimization

Abstract

In dynamic optimization problems, although the problem environments keep changing, a new environment is usually related to its previous environments. Based on the relevance, the search experience in the previous environments can be reused to accelerate the optimization of the current new environment. Thus, in this paper, we propose a dual-archive-based particle swarm optimization to utilize the useful information accumulated in past environments as well as to explore the emerging information of each new environment. Specifically, in the proposed algorithm, the good solutions found in past environments are stored in two different archives, i.e., a fine-grained archive and a coarse-grained archive, so as to preserve both detailed information and systemic information, respectively. Once the environment is changed, the solutions in the two archives will be used for guidance to quickly find high-quality solutions in the new environment. The proposed algorithm is evaluated on the famous moving peaks benchmark in terms of two performance measures. The experimental results show that the proposed algorithm is competitive with state-of-the-art algorithms for dynamic optimization problems.