The five stages of autonomous driving

From hype to disappointment and back down to earth: Autonomous driving is not coming as quickly as expected, but remains a trend for the future. So far, the mobility industry has taken about half of the five stages. An overview of the five levels - and a classification of where we stand.

The stairway to the robot car should have five levels. At least that's the definition put forward by the international engineering organization SAE in 2014: From starting level 0 to a fully automated vehicle at level 5, the paper mapped out in detail the planned development of autonomous car capabilities. Almost a decade later, however, the classification has become outdated: some levels now appear fragile, while others have been newly introduced.

"One reason why automated driving is not yet so widespread is the many steps that need to be taken in terms of development and testing before such systems come onto the market. Only when we are sure that the system works reliably will it be put on the road," explains Sven Lanwer, Head of Driver Assistance and Automated Driving at Bosch.

Level 1

Assisted driving is already widespread in various forms right down to the lower vehicle segments. As the name suggests, assistants help with driving. For example, they warn of vehicles in the blind spot, when leaving the lane or maintain the set distance to the vehicle in front. The assistants only assist, the person behind the wheel still has to drive.

Level 2

In "semi-automated driving", the car can already take over individual driving tasks itself, but must always be monitored by the driver. Most car manufacturers have currently reached this stage. An example of this is a traffic jam assistant that steers, brakes and starts up again independently in stop-and-go traffic. Although the driver does not have to actively steer, they are not allowed to take their hands off the wheel. If they do, the car prompts them to take hold of the steering wheel after a few seconds. Most recently, an expansion stage of the technology has been established, known as "Level 2 Plus" or "Level 2 Hands Free", which explicitly allows the driver to let go of the steering wheel and also tolerates this for a longer period of time.

Level 3

In "highly automated driving", the car takes over various functions in certain traffic situations and can, for example, handle longer stretches of highway completely on its own. However, the driver must be able to take over the steering wheel again after a request from the car. In theory, several car models are already capable of this technology, but only a few are currently allowed to use it legally, including the Mercedes S-Class and the BMW 5 Series. There are restrictions on speed and road type, for example - level 3 driving in Germany is only permitted on suitable sections of highway and at a maximum speed of 130 km/h.

Level 4

If the vehicle is "fully automated", the car can master specific applications completely on its own - from highway driving to highly complex urban traffic situations. Meanwhile, the driver can sleep, for example, and is no longer liable in the event of damage or traffic violations. This is where we leave the area that private cars still master today. Fully automated vehicles currently include robotaxis or robo-shuttles from mobility service providers that operate in localized areas. Another example is the "Automated Valet Parking" service offered by Bosch and Mercedes in Germany, where cars find their own parking space in specially equipped parking garages.

Level 5

In "autonomous driving", the occupants become complete passengers, without even needing a steering wheel or pedals. The car can handle all driving tasks on its own. On any road, in any weather and in the most complex traffic situations. At the beginning of the decade, engineers were still hoping to reach the top level by the mid-2020s. This will not happen. Some industry representatives doubt whether it will ever get that far. Not least because the costs would probably be so high that it would hardly be worthwhile. Bosch expert Lanwer is aware of another problem: "For example, it's very difficult to get an autonomous vehicle onto a lifting platform if you can't control it." For the time being, Level 5 does not see itself on the market because it currently offers no advantage over Level 4.

Criticism of the five-level model

In the meantime, the five-level model is coming under increasing criticism. The Federal Highway Research Institute (BASt) wants to provide a better overview and is proposing a new model for the classification of robot cars with just three modes: assisted, automated and autonomous. With the simpler classification, the authority wants to simplify communication about driverless driving, especially for users, and prevent misunderstandings. The proposal is that the previous terminology system should only be used in specialist circles in future.

BASt uses the term "autonomous mode" when the computer is permanently acting as the vehicle driver, as is envisaged for robotaxis in the near future. This state corresponds to SAE levels 4 and 5. The numerical designations of the SAE system remain compatible with the BASt designations and, according to the authority, should continue to be used in communication between experts.

Read also:

The Federal Highway Research Institute (BASt) criticizes the existing five-level model for classifying autonomous vehicles and proposes a simplified three-mode model instead. In this new system, 'autonomous mode' refers to scenarios where the computer acts as the vehicle driver, such as in future robotaxi operations, which correlates to SAE levels 4 and 5.

Given the Federal Highway Research Institute's ongoing efforts to improve communication about driverless vehicles, there's a need for the specialized terminology of the previous SAE system to be used only among experts in the future. This shift aims to avoid potential misunderstandings between specialists and users of autonomous driving technologies.

Additionally, the robotics and automotive industries have had to grapple with challenges in recent years, such as the development and testing of reliable autonomous driving systems to satisfy regulatory requirements. For instance, as Sven Lanwer, Head of Driver Assistance and Automated Driving at Bosch, explained, it's crucial to ensure the system's reliability before introducing it onto the market. This process typically involves tackling numerous technical challenges, including developing safe communication protocols between different vehicles and robotics for complex traffic scenarios.

The Federal Highway Research Institute's proposed three-mode system may simplify communication about driverless technology, but the challenges of developing autonomous driving capabilities still require the active engagement of industry experts, researchers, and regulatory bodies.

Source: www.ntv.de