Obstacle Avoidance through Reinforcement Learning