Publicado el 06/02/2013
Complex trajectories for robot navigation can be decomposed in elementary movement primitives. Reinforcement Learning is employed to learn the movement primitives, as well as the combination of primitives, to reach a desired goal position while avoiding obstacles.