Towards Hierarchical Self-Optimization in Autonomous Groups of Mobile Robots

We present a real-world scenario for investigating and demonstrating hierarchical self-optimization in autonomous groups of mobile robots. The scenario is highly dynamic and easily expandable. It offers adequate starting points for the integration of hierarchical self-optimization. Reinforcement learning, e. g., can be introduced in order to improve the individual behavior of a single robot. Also swarm intelligence algorithms can improve the overall team behavior with respect to common goals. A reference behavior system incorporating a dynamic role assignment and hierarchical state machines was implemented and has been applied to the miniature robot BeBot. The system was evaluated by conducting several tests.