A coevolutionary technique based on multi-swarm particle swarm optimization for dynamic multi-objective optimization

•A coevolutionary multi-swarm particle swarm optimizer is proposed.•All swarms utilize an information sharing strategy to evolve cooperatively.•A velocity update mechanism and a new boundary constraints technique are adopted.•A similarity detection operator is used to detect the environment change.•...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:European journal of operational research Ročník 261; číslo 3; s. 1028 - 1051
Hlavní autoři: Liu, Ruochen, Li, Jianxia, fan, Jing, Mu, Caihong, Jiao, Licheng
Médium: Journal Article
Jazyk:angličtina
Vydáno: Elsevier B.V 16.09.2017
Témata:
Coevolution
Dynamic multi-objective optimization
Multi-swarm Particle swarm optimization
Multiple objective programming
Similarity detective
Multi-swarm Particle swarm optimization
Dynamic multi-objective optimization
Coevolution
Multiple objective programming
Similarity detective
ISSN:0377-2217, 1872-6860
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:•A coevolutionary multi-swarm particle swarm optimizer is proposed.•All swarms utilize an information sharing strategy to evolve cooperatively.•A velocity update mechanism and a new boundary constraints technique are adopted.•A similarity detection operator is used to detect the environment change.•It is applied to solve eight benchmark problems, with a good performance obtained. In real-world applications, there are many fields involving dynamic multi-objective optimization problems (DMOPs), in which objectives are in conflict with each other and change over time or environments. In this paper, a modified coevolutionary multi-swarm particle swarm optimizer is proposed to solve DMOPs in the rapidly changing environments (denoted as CMPSODMO). A frame of multi-swarm based particle swarm optimization is adopted to optimize the problem in dynamic environments. In CMPSODMO, the number of swarms (PSO) is determined by the number of the objective functions, and all of these swarms utilize an information sharing strategy to evolve cooperatively. Moreover, a new velocity update equation and an effective boundary constraint technique are developed during evolution of each swarm. Then, a similarity detection operator is used to detect whether a change has occurred, followed by a memory based dynamic mechanism to response to the change. The proposed CMPSODMO has been extensively compared with five state-of-the-art algorithms over a test suit of benchmark problems. Experimental results indicate that the proposed algorithm is promising for dealing with the DMOPs in the rapidly changing environments. The flowchart of the proposed CMPSODMO. [Display omitted]
ISSN:0377-2217
1872-6860
DOI:10.1016/j.ejor.2017.03.048