A new approach to prevent the decrease of population diversity in genetic algorithms

A genetic algorithm is a population based evolutionary algorithm inspired from natural evolution. A major drawback of this algorithm is the decrease of population diversity throughout execution. Several methods are used to prevent this drawback, including island model approach and diffusion model approach. In this study, a new approach (Earth Model) is proposed in order to slow down this decrease of population diversity. In this model, population is distributed randomly over a unit sphere and mating is constrained with a critical spherical distance. The model is tested on Euclidian Travelling Salesman Problem. Initial tests show that the decrease in diversity is slower in Earth Model and total distance of the best solution of the test problem is approximately 4.8% lower than traditional genetic algorithm.