Hybrid Real-Coded Genetic Algorithm with Quasi-Simplex Technique

Abstract

†Summary This paper proposes a new real-value mutation operator and a hybrid real-coded genetic algorithm with quasi-simplex technique using this new mutation operator (RCGAQS). Compared with the classical GA (CGA), RCGAQS has the following distinguish features: (1) A new real-value mutation mechanism was used to increase the capability of global search (exploration); (2) The modified simplex technique, so called the quasi-simplex technique, was employed to generate prospective offspring to increase the capability of local search (exploitation); and (3) The dynamic subpopulation strategy, in which the entire generation is subdivided into a number of subgroups in each evolution step, was adopted to enhance the abilities in both exploration and exploitation. RCGAQS algorithm has been implemented and tested on typical benchmark functions along with CGA. The experimental study has shown that RCGAQS is impressive in finding the near global optimal solutions cross all the selected benchmark functions and is substantially robust.