Cognitive behavior optimization algorithm for solving optimization problems

• Published in: Applied soft computing Vol. 39; pp. 199 - 222
• Main Authors: Li, Mudong, Zhao, Hui, Weng, Xingwei, Han, Tong
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.02.2016
• Subjects: Cognitive behavior model; Exploitation; Exploration; Global optimization; Nature-based algorithms; Cognitive behavior model; Nature-based algorithms; Exploration; Global optimization; Exploitation
• ISSN: 1568-4946; 1872-9681
• DOI: 10.1016/j.asoc.2015.11.015

•A new swarm intelligence algorithm, COA, is developed for the optimization problems.•The novel behavior model in COA makes the algorithm more effective and intelligent.•Performance on 53 different benchmark problems is considered.•The problem solving success of COA is compared with 8 state-of-the-art algorithms. Nature-based algorithms have become popular in recent fifteen years and have been widely applied in various fields of science and engineering, such as robot control, cluster analysis, controller design, dynamic optimization and image processing. In this paper, a new swarm intelligence algorithm named cognitive behavior optimization algorithm (COA) is introduced, which is used to solve the real-valued numerical optimization problems. COA has a detailed cognitive behavior model. In the model of COA, the common phenomenon of foraging food source for population is summarized as the process of exploration–communication–adjustment. Matching with the process, three main behaviors and two groups in COA are introduced. Firstly, cognitive population uses Gaussian and Levy flight random walk methods to explore the search space in the rough search behavior. Secondly, the improved crossover and mutation operator are used in the information exchange and share behavior between the two groups: cognitive population and memory population. Finally, the intelligent adjustment behavior is used to enhance the exploitation of the population for cognitive population. To verify the performance of our approach, both the classic and modern complex benchmark functions considered as the unconstrained functions are employed. Meanwhile, some well-known engineering design optimization problems are used as the constrained functions in the literature. The experimental results, considering both convergence and accuracy simultaneously, demonstrate the effectiveness of COA for global numerical and engineering optimization problems.