Learning Complex Robot Control Using Evolutionary Behavior Based Systems
Yohannes Kassahun, Jakob Schwendner, José de Gea Fernández, Mark Edgington, Frank Kirchner
In GECCO '09: Proceedings of the 11th Annual conference on Genetic and evolutionary computation, (GECCO-2009), 08.7.-12.7.2009, Montréal, o.A., pages 129-136, 2009. ISBN: 978-1-60558-325-9.
Evolving a monolithic solution for complex robotic problems is hard. One of the reasons for this is the difficulty of defining a global fitness function that leads to a solution with desired operating properties. The problem with a global fitness function is that it may not reward intermidiate solutions that would ultimately lead to the desired operating properties. A possible way to solve such a problem is to decompose the solution space into smaller subsolutions with lower number of intrinsic dimensions. In this paper, we apply the design principles of behavior based systems to decompose a complex robot control task into subsolutions and show how to incrementally modify the fitness function that results in desired operating properties as the subsolutions are learned, and avoids the need to learn the coordination of behaviors separately. We demonstrate our method using learning to control the Quadrocopter flying vehicle.
Genetics-Based Machine Learning, Learning Classifier Systems