Invited Speaker

Martin Riedmiller

Martin Riedmiller is a Professor (C4) for Neuroinformatics at the University of Osnabrueck, Germany. At this University he he leads the Neuroinformatics Group that deals with systems that can learn or adapt to complex dynamic environments. His main research areas are machine learning, neuroinformatics and robotics. He is co-author of more than forty papers in international journals and conferences and he is team leader of one of the world most successful robotic soccer teams - Brainstormers. Brainstormers Simulation league team was two times 2nd place and two times 3rd place in RoboCup world championship (plus several European awards) and Brainstormers Tribots middle-size league team was German Open champion in 2004.

Making Robots Learn the Right Things

Reinforcement learning is a powerful paradigm to enable systems to learn from scratch -by only providing the training information of success or failure. The research project 'Brainstormers' - a robotic soccer club - explores the principle capabilities of such learning algorithms, their behaviour in practical, complex and competitive environments and their integration in larger software systems. Both benefits and open problems of reinforcement learning methods are discussed on several example applications.