Speeding Up Relational Reinforcement Learning through the Use of an Incremental First Order Decision Tree Learner