Keywords: learning logic rules; predicate grounding; reinforcement learning; robotics; two-level planning.