Constrained optimization based on improved teaching–learning-based optimization algorithm

• Published in: Information sciences Vol. 352-353; pp. 61 - 78
• Main Authors: Yu, Kunjie, Wang, Xin, Wang, Zhenlei
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 20.07.2016
• Subjects: Algorithms; Benchmarking; Constrained optimization; Constraint handling; Constraints; Convergence; Evolution; Learning strategy; Optimization; Strategy; Teachers; Teaching–learning-based optimization; Constraint handling; Teaching–learning-based optimization; Learning strategy; Constrained optimization
• ISSN: 0020-0255; 1872-6291
• DOI: 10.1016/j.ins.2016.02.054

•An improved constrained teaching-learning-based optimization (ICTLBO) method is proposed.•The subpopulation-based strategy and information exchange mechanism are developed for teacher phase of ICTLBO.•A new learning method is proposed for learner phase of ICTLBO.•Three different constraint handling methods are adopted for three situations in the process of search.•ICTLBO is evaluated through benchmark functions from CEC2006 and CEC2010 and engineering problems. This paper proposes an improved constrained teaching–learning-based optimization (ICTLBO) method to efficiently solve constrained optimization problems (COPs). In the teacher phase of ICTLBO, the population is partitioned into several subpopulations, and the direction information between the mean position of each subpopulation and the best position of population guide the corresponding subpopulation to the promising region promptly. Information exchange between different subpopulations is used to discourage premature convergence of each subpopulation. Furthermore, in the learner phase, a new learning strategy is introduced to improve the population diversity and enhance the global search ability. Three different constraint handling methods are adopted for three situations, which are infeasible, semi-feasible, and feasible situations, during the evolution process. To evaluate the performance of ICTLBO, 22 benchmark functions presented in CEC2006 and 18 benchmark functions introduced in CEC2010 are chosen as the test suite. Moreover, four widely used engineering design problems are selected to test the performance of ICTLBO for real-world problems. Experimental results indicate that ICTLBO can obtain a highly competitive performance compared with other state-of-the-art algorithms.