Mobile Robots Navigation 448 on sensor values. The search space of behaviors is huge and designing suitable behaviors by hand is very difficult (11) therefore we use a genetic algorithm (GA) within the simulator in order to find appropriate behaviors. The GA selects from a population of robots (neural networks) using a fuzzy fitness function that considers various robotic motivations such as: the need for exploration (curiosity), the need to conserve its battery (energy), the desire to determine its location (orientation), and the capacity to return to its initial position (homing). This paper is organized as follows. Section 2 gives a description of the robotic system. Section 3 describes the entropy measures used for diversity evaluation. Section 4 introduces the experiments performed. In section 5 we describe and summarize our results. Finally, in section 6 some conclusions are drawn.