Improving Efficiency of a Genetic Algorithm Applied to Multi-robot Tactic Operation

There are two important issues in the Genetic Algorithm searching and optimization process: population diversity and selective pressure. These two issues are strongly related and have direct impact on the search efficiency. Two factors that directly influence these issues are often ignored: overlapping populations and fitness scaling. In this paper we address the use of overlapping populations and fitness scaling in a Genetic Algorithm (GA) applied to multi-robot squad formation and coordination. The robotic task is performed over a natural disaster scenario (a forest fire simulation). The robot squad mission is surrounding the fire and avoiding fire’s propagation based on the strategy proposed by the GA. Simulations have been carried out with several GA parameters (several types of scaling and different degrees of overlapping) in order to obtain the most efficient optimization for group formation and task execution. Simulations results show that the use of overlapping population and fitness scaling present better results than non-overlapping population and unscaled fitness.