W-Learning for Behavior Coordination on a Real Khepera
Dimitrij Surmeli, Horst–Michael Groß · 2001
We utilize HUMPHRYS' W-Learning on a real robot Khepera to coordinate two behaviors in a maze: first, to avoid obstacles and second, to find a location marked by a certain color (e.g., where food can be found). We describe the experimental setup and compare results of the individual agents with one monolithic agent solving both tasks and the agents coordinated by W-Learning. We demonstrate the feasibility of W-Learning on a real robot.