The difference between autonomous and driverless cars

Ralf Herrtwich, Head of Automotive at HERE
Berlin 52° 30' 57.852" N, 13° 22' 37.128" E

Automated, self-driving, autonomous and driverless – there are many different terms for vehicles that can drive themselves and, perhaps surprisingly, they don’t all mean the same thing.

The clearest of the terms is probably ‘driverless’. A vehicle that has no driver needs to be able to do everything by itself in the area of its designated use, so it has to be fully automated. And, as ‘fully’ hints when it comes to automation, there are actually different levels.

Understanding the different levels

SAE, the Society of Automotive Engineers, defines six levels of automation, starting at Level 0 for no automation, and leading up to Level 5 for full automation. The simplest automated tasks can be found in Level 1, with features including adaptive cruise control and parking assistance, where vehicles are capable of performing specific tasks but a driver is still required to take control at any time.

Level 2 is called partial automation, and this is the automation level that can be seen in the market today with the Mercedes-Benz E-Class or Tesla S. It’s a system capable of driving the vehicle by itself under certain conditions and on certain roads, but the driver still needs to supervise what the car does. The driver, and not the OEM, is accountable for road safety, which makes it much easier for an OEM to deploy such a system, as less validation is needed. However, it’s important that the driver understands that he or she needs to supervise the car even if it feels as though the vehicle can handle everything.

Things become even trickier at Level 3, which is called conditional automation. Here the car should be able to monitor its environment completely, but can call upon the driver as a fallback. The driver does not need to pay attention, but needs to be ready to take control at any time. It’s a difficult task because the driver may have looked away from the road and will first need to assess the situation when called upon.

Level 3 is also tricky for the system developer, as SAE doesn’t define how much time is required for the driver to get back into the loop. "A few seconds" is a statement that you often hear. Well, every second at 120km/h is 33 meters travelled – so any emergency situation is coming towards the driver very quickly (or vice versa). Different OEMs will make different design choices here; we’ll see some systems being called "Level 3" in the next years and it will be interesting to watch what their operating constraints are and in which cases they need the driver to help.

Things get clearer, albeit not easier to implement or validate, at the remaining levels of automation. Level 4 is called high automation; such a car should be able to do everything by itself without driver involvement in some so-called "driving modes". A driving mode is a traffic scenario such as highway driving or maneuvering in a parking garage. Level 4 is what Google (with their self-driving car project called Waymo) is striving for: vehicles that can drive on a certain mapped road network in a specific area.

Level 4 cars need to be able to cope with all situations, but they may, for example, never enter a highway because their sensors are not designed for the range needed to travel at high speeds: the faster the vehicle goes, the further its sensors have to see to bring the vehicle to a safe stop in case the driving path is blocked. "Driving mode" is the key to the introduction of highly-automated cars. It's comparatively easy to develop a Level 4 self-parking car for a garage, especially if humans are not allowed to walk in this car park, but it's much more difficult to make this work across a small city. Google has been trying for eight years now in Mountain View, and has not yet released its cars for any regular mobility service

Finally, there’s Level 5, which is full automation. SAE says such a system should be able to handle all driving tasks in all driving modes and under all environmental conditions, just like a human driver. At Level 4 one could still say that in fog or snow the system would not work, but a car at Level 5 will need to master even such adverse weather conditions. Bad weather is a key enemy to vehicle sensors (and sometimes to actuators), so it’s no coincidence that so many vehicle automation experiments take place in sunny California, Nevada, or Arizona.

Level 5 is naturally the most difficult to achieve, but unfortunately many business cases for new forms of mobility only make sense if cars are fully automated. It’s hard to imagine an automated Uber fleet being grounded just because of snow in New York City.

Which terms are applicable to which automation level?

Starting at Level 2, you can call a car autonomous because it makes its own driving decisions. Following this argument, you can also call it self-driving, although the term seems more adequate for cars at Level 4 and 5. These are also the only levels at which a car can become driverless, which conversely does not mean it must never have a driver: just like the autopilot in airplanes, the vehicle autopilot may have an on/off switch.

And just because a vehicle can perform the driving task itself does not mean that the driver wants it to do so. Even with ever-increasing traffic density, there will still be situations where it is sheer fun to drive and drivers want to be in charge. Whether they are then up to the task with less and less “self” driving experience is another matter...

What does this all mean for HERE?

This means we will be busy for quite some time accompanying and supporting the development of vehicles at different levels of automation. Developers on the OEM side will still need some time to figure out the increasingly difficult problems as they climb up the automation ladder, and more and more fresh and precise data can and will be collected in our location databases to make their tasks easier.

Topics: Automotive, Autonomous cars, Features, Editor's picks

Comments