Learning to Play the BRIO Labyrinth Game

Jan Hendrik Metzen; Elsa Andrea Kirchner; Larbi Abdenebaoui; Frank Kirchner

In: KI - Künstliche Intelligenz, German Journal on Artificial Intelligence - Organ des Fachbereiches "Künstliche Intelligenz" der Gesellschaft für Informatik e.V. (KI), Vol. 3/09 - Themenheft ''Reinforcement Learning'', Pages 34-37, Böttcher IT Verlag, 2009.


In this article, we present the BRIO labyrinth system, a challenging real-world, robotic system suited as testbed for an empirical analysis and comparison of Reinforcement Learning (RL) algorithms. We discuss the development of both the testbed itself as well as a realistic simulation of the system and first experiments with RL algorithms on the simulated BRIO system. Furthermore, we briefly sketch some experiments that will be conducted with human probands on the rio system that will yield some insights into sensorimotor learning of humans.

German Research Center for Artificial Intelligence
Deutsches Forschungszentrum für Künstliche Intelligenz