On emergence of group behavior in a genetically controlled autonomous agent system

This paper presents a genetically controlled means for a group of autonomous robots to change their sensory, behavioral, and learning capabilities in an attempt to collectively achieve such tasks as surrounding a set of target objects. We first describe the essential building blocks of the autonomous mobile robots, such as their sensing capability, primitive behavior, and behavioral learning and selection mechanism. We then demonstrate how a genetic algorithm can be applied to evolve some globally optimal group behavior among the robots, by way of selecting the appropriate sensory and behavioral parameters as well as specific task configurations.