An evolutionary algorithm to track changes of optimum value locations in dynamic environments

Abstract

Non-stationary, or dynamic, problems change over time. There exist a variety of forms of dynamism. The concept of dynamic environments in the context of this paper means that the fitness landscape changes during the run of an evolutionary algorithm. Genetic diversity is crucial to provide the necessary adaptability of the algorithm to changes. Two mechanism of macromutation are incorporated to the algorithm to maintain genetic diversity in the population. The algorithm was tested on a set of dynamic testing functions provided by a dynamic fitness problem generator. The main goal was to determinate the algorithm´s ability to reacting to changes of optimum values that alter their locations, so that the optimum value can still be tracked when dimensional and multimodal scalability in the functions is adjusted. The effectiveness and limitations of the proposed algorithm is discussed from results empirically obtained.