Learning the local search range for genetic optimisation in nonstationary environments

We examine a modification to the genetic algorithm. The variable local search (VLS) operator was designed to enable the genetic algorithm based online optimisers to track optima of time-varying dynamic systems. This feature is not to the detriment of its ability to provide sound results for the stationary environments. The operator matches the level of diversity introduced into the population with the "degree" of the environmental change by increasing population diversity only gradually. The paper also shows that the performance of the designed tracking method can be further enhanced by integrating it with a simple exemplar-based incremental learning technique. It is believed that the designed technique will prove beneficial in the application of the genetic algorithm based approaches to industrial control problems.