A Robust Archived Differential Evolution Algorithm for Global Optimization Problems

Abstract

A robust archived differential evolution algorithm is put forward by means of embedding a flexibility processing operator and an efficiency processing operator based on original DE and ADE. A special constraint-handling mechanism based on dynamic penalty functions and fitness calculation of individuals is adopted in the proposed method to deal with various constraints effectively, which is further extended by means of a flexibility processing operator so as to make it suitable for different type problems, including those with or without constraint(s) and those with continuous, discrete or mixed discrete-continuous variables. Furthermore, an archive of solutions is maintained during the evolutionary process so as to keep the useful information of previous solutions and local optima for the estimation of new solutions. Based on the archive of solutions, an iterative control operator and an efficiency processing operator are designed in the algorithm. The former guides the evolutionary process towards a promising search space, avoiding unnecessary and worthless search. The latter improves the local searching efficiency and the final searching quality. Experimental results based on a suite of six well-known optimization problems reveal that the proposed algorithm is robust, effective, efficient and suitable for different type global optimization problems.