Study On Evolutionary Algorithms In Noisy Environments

Evolutionary Algorithms (EAs) based on the principle of biology have been widely investigated and utilized in the fields of intelligent control in recent years. EAs in nowadays, however, are mostly developed for deterministic situation. So those are not always able to acquire expectant performances under the noisy environments.This paper is devoted to the effects of environmental noise on evolutionary algorithms (EAs). Through the study of the sphere test function, the performances of genetic algorithms (GAs) and evolution strategies (ESs) are evaluated. It is shown that neither GAs nor ESs can find their global optimal solution in noisy environments, and with the increase of noise strength, their convergent abilities decrease.To improve the robustness of EAs in complex environments, reasons on their working mechanisms are discussed firstly. And then two methods for restraining noises are presented utilizing re-sampling and Kalman filter, which can improve the performances by filtering the fitness functions. The simulation studies are carried out, those haveshown that both of the two methods benefit to improvement of EAs under noisy environments.