carla running carla with RL method for now using ray to train a ppo agent he can learn turn left or right 20190226 using 8 channel rbg + depth + rgb + measurement_encode