The Five Levels of Autonomous Driving

Autonomous driving requires a variety of complex systems that work together. These include collision avoidance algorithms, lane tracking systems and motion planning.


Software constructs an internal map of the road environment using input from sensors like radar, lasers and high-powered cameras. This information is passed to localization, trajectory planning and control blocks.

Level 1

Autonomous driving is commonly split into five levels, thanks to a taxonomy developed by SAE International (previously the Society of Automotive Engineers). Here’s how it works:

At this level of autonomy, a driver must be fully engaged. This includes interacting with the steering wheel and other controls, as well as monitoring road conditions at all times. Features that fall under this level include ADAS features such as blind spot monitoring and lane departure warning.

Level 2 takes things up a notch by allowing the driver to keep their eyes off the road for brief periods of time. That said, the car will still need to be on-hand to intervene when needed and will alert the driver when it is no longer able to perform its tasks in certain conditions. Examples of level 2 automation are Tesla’s Autopilot or Nissan’s ProPilot.

Level 4 is where things start to get really interesting. At this level, the driver no longer needs to interact with the vehicle and can focus on other activities such as watching a movie or working on a laptop, although they will need to remain ready to take control in case of an emergency.

Level 2

Level 2 takes the driver assistance systems found in many modern cars up a notch. The human driver still needs to keep an eye on the road (no rush hour naps allowed), but the system can handle some of the dynamic driving tasks such as cruise control, lane keeping, and even automatic parking.

This system can also take over combined longitudinal and lateral driving functions, but only in certain conditions or scenarios. It is similar to what TuSimple’s automated trucking system offers, and Waymo’s relatively small self-driving taxi runs are currently offering. Breakthroughs in computing such as the NVIDIA Xavier system-on-chip make this level of autonomy possible for automakers to offer to consumers.

This level of autonomy isn’t available in any new vehicles to the public yet, though. It falls into a legal gray area, and there are worries that drivers may become complacent behind the wheel of a car capable of taking over at any time. This is what prompted Audi to kill off its A8’s Traffic Jam Pilot feature. However, this could change in the future as more autonomous features are added to existing cars.

Level 3

At level 3, the driver can turn their hands off, relax in the seat and even take a nap (though we still recommend staying awake!). The car will still monitor the surroundings and can even make autonomous decisions for itself, like accelerating past a slow-moving vehicle. However, the driver must be ready to take control should the system fail and remain alert to the situation at all times.

This level is a bit more advanced than the current highway cruise controls you’ll find in vehicles from Tesla, Ford, GM and Subaru, which offer hand-off driving functions for highway speeds only. The 2019 Audi A8L will be the first production Level 3 vehicle on the market and features a hands-free system called Traffic Jam Pilot that combines a lidar scanner, sensor fusion, powerful processing power and built-in redundancies.

This level is known as conditional automation because it can only be used under certain conditions, such as highways at specific speeds. For example, Mercedes’ DRIVE PILOT is currently limited to highways and can only operate at speeds up to 60 km/h.

Level 4

Level 4 autonomy is the sci-fi vision of cars that take on everything and drive you around. It’s also the goal of Google’s Waymo self-driving car, and it promises to let you relax on your daily commute or long road trip by watching Netflix or even falling asleep in the driver’s seat. The key difference here is that the vehicle will still ask you to take over and will be able to tell whether or not it’s safe to do so with less than a two-second comms lag.

This is the most advanced autonomous driving mode yet, but you can’t really call it self-driving. Using a variety of sensors and cameras, Level 3 autonomy can monitor the environment, control the throttle and brakes, enter and exit freeways, navigate traffic lights and signage, maintain speed and distance from other vehicles, and even park itself. That’s a lot of what a human driver does to get from A to B, and you’ll have the power to take back the wheel at any time. It’s a little like putting your car in cruise control, but more sophisticated.

Level 5

SAE Level 5 autonomous vehicles are fully self-driving and able to operate on any road or geographic area. There are currently no cars that have reached this stage but automakers and technology companies are racing to be first to market with a commercially viable model.

At this point, you can sit back and relax. The car handles the driving, steering and braking on highways and other open roads. But a driver must remain in the vehicle to take control if the system fails or if driving conditions change suddenly, such as heavy snow.

Achieving Level 5 will be a massive undertaking and requires a large amount of data to be processed in real time. That’s why engineers need powerful solvers like Ansys to understand how weather affects sensor performance and how to optimize sensors for different conditions.