An angular distance dependent alternation model for real-coded genetic algorithms

When we use genetic algorithms to solve any type of problems, it is important to maintain the diversity of populations for avoiding early stage stagnation or falling into local minima. We propose an angular distance dependent alternation (ADDA) model as a generation alternation model on real-coded genetic algorithms (GA) to improve its performance by maintaining adequate diversity of populations. The basic concept of the ADDA is that all of offspring generated by crossover operations will be clustered by a corresponding parent based on the angular distance metric and will be transposed from the parent. We compare performance of the proposed alternation model with previous family based Minimal Generation Gap (MGG) model and Distance Dependent Alternation (DDA) model. Using with the multi-parental Unimodal Normal Distribution Crossover (UNDX-m), the ADDA model shows good performance on three typical benchmark problems.