Performance of infeasibility empowered memetic algorithm for CEC 2010 constrained optimization problems

Real life optimization problems often involve one or more constraints, and there is a significant interest among the research community to develop efficient algorithms to solve such constrained optimization problems. This paper presents a memetic algorithm combining the strengths of an evolutionary algorithm and a local search strategy. Since solutions of constrained optimization problems are expected to lie on constraint boundaries for most problems, the algorithm explicitly preserves marginally infeasible solutions to intensify search around the constraint boundaries. Furthermore, local search is done from solutions within the population to yield good quality solutions in early generations. The concepts of injecting high quality solutions in earlier generations and preservation of marginally infeasible solutions are both known to improve the efficiency of evolutionary algorithms for constrained optimization problems. The performance of the algorithm is presented on the newly proposed set of test functions (C01-C18)for 10 and 30 dimensions.