A two-stage coevolution approach for constrained optimization

Abstract

In this paper, a coevolution approach with two stages is proposed for constrained optimization problems (COPs). At the first stage, the approach enters the feasible region rapidly by utilizing the feasibility rule with incorporation of objective function information (FROFI), which is an effective method for the balance between constraints and objective function. At the second stage, the population of the first stage coevolves with an additional population to locate the global optimum. The additional population is generated when a feasible solution is found. Penalty function as a constraint-handling technique is employed on the additional population. By means of coevolution, elite individuals from the original population and the newly generated population are exchanged to promote each other for the global optimum. The performance of our approach is evaluated on a suite of benchmark functions from IEEE CEC 2010. Experimental results have shown that the proposed approach generally outperforms four other state-of-the-art constrained optimization algorithms on most of the benchmark functions.