In this paper, Genetic Algorithms (GAs) with changing fitness function applied to mobile robots are investigated. The use of GAs with Random Immigrants where the individual with the lowest fitness and its next neighbors are replaced in every generation is proposed to deal with such problems. In order to avoid that the individuals with the best fitness take the new individuals to extinction, these ones are preserved in a subpopulation. This simple approach can take the system to a self-organization behavior, which can be useful to maintain the diversity of the solutions and, then, to allow the GA to escape from local optima induced by the changes in the problem. Simulations of a mobile robot controlled by an Artificial Neural Network configured by the proposed GA are presented.
Genetic Algorithms; Mobile Robots; Self-Organization; Optimization
