Teach a simulated quadcopter to fly with reinforcement learning and deep deterministic policy gradients