A Hybrid Biomimetic Genetic Algorithm Using a Local Fuzzy Simplex Search

This paper presents a hybrid genetic algorithm that expands upon the previously successful approach of twinkling genetic algorithm (TGA) by incorporating a highly efficient local fuzzy-simplex search within the algorithm. The TGA was in principle a bio-mimetic algorithm that introduced a controlled deviation from a typical GA method, by not requiring that every genevariable of an offspring be the result of a crossover. Instead, twinkling allowed the genetic information of the randomly chosen gene locations to be directly passed on from one parent, which was shown to increase the likelihood of survival of a successful gene value within the offspring, rather than requiring it to be blended. The twinkling genetic algorithms proved highly effective at locating exact global optimum with a competitive rate of convergence for a wide variety of benchmark problems. In this work, it is proposed to couple the TGA with a fuzzy simplex local search to increase the rate of convergence of the algorithm. The proposed algorithm is tested using common mathematical and engineering design benchmark problems. Comparison of the results of this algorithm with earlier algorithms is presented.