Single-objective optimization; constrained problems;differential evolution;benchmark tests

A variant of competitive differential evolution modified for the solution of single objective constrained real-parameter optimization is presented in this paper. The modification consists in alternation of minimization of the constrains' violation and minimization of objective function value. The variant was tested on the family of 18 problems developed for the Special Session S000 of CEC2010. Using competitive setting avoids tuning of most control-parameter values because only size of population must be set by the user of the algorithm. The experimental results showed acceptable reliability of the search in most of test problems.