Safe state abstraction and discounting in hierarchical reinforcement learning [electronic resource] / Bernhard Hengst