Reinforcement Learning with Kohonen-based State Aggregation for Obstacle Avoidance of a Mobile Robot