What are Autonomous Driving Levels?

18-12-2020 | By Sam Brown

There is no doubt that self-driving vehicles will one day become widespread and available to all. But when talking about the stage of development, the term levels is often used. What are self-driving levels, why are they important, and what level is humanity currently at?

What is a self-driving car?

The term “self-driving” has been around for a long time, but when the term self-driving or autonomous is applied to products, its meaning tends to change. In the truest sense, an autonomous vehicle would drive itself entirely in all scenarios with no user input, ever. However, such vehicles do not exist, but some are still sold as having autonomous capabilities and features.

Even vehicles that are being designed to run themselves still require the user to be aware of the road and potential dangers so that in the event of an emergency the driver can take back control (similar to a co-pilot). This confusion of the term “autonomous” and “self-driving” has even resulted in road deaths with users operating “auto-pilot modes” as if the car can drive entirely by itself. 

Some autonomous features that modern vehicles can integrate include lane-steering, adaptive cruise, emergency braking, and pedestrian detection. However, these features only enhance a drivers capability, but the driver is still required to operate the vehicle. Since there are no vehicles in existence that can entirely drive itself, self-driving vehicles are still a work of fiction.

What are the “levels” in autonomous vehicles?

A large amount of confusion that surrounds the definition of autonomous vehicles led the automotive standard body SAE International to create a level system that clearly defines each level of autonomous capabilities. Therefore, a vehicle that can do limited autonomous activities may have a level 1 rating while a more advanced vehicle could have level 2.

Level 0 – No autonomous driving

At this level, the driver is required to operate the vehicles at all times. However, this level can still include warning features such as parking sensors, ice warning, and other systems, but at the end of the day, all the decisions literally rest in the hands of the driver. 

Level 1 – Driver assistance

At this level of automation, a vehicle provides a singular activity that it can automate such as cruise control. However, even though the vehicle can control acceleration, deceleration, and steering, the human driver is still solely responsible for the vehicle's safety. For example, when cruise control is enabled, the driver must still have hold of the wheel, feet at the pedals, and check the surroundings. This level is called “driver assistance”, and as the name suggests, assists the driver. 

Level 2 – Partial automation

At this level of automation, the vehicle can combine different assistant systems to be able to control the vehicle with limited capabilities. For example, a level 2 autonomous vehicle could provide lane-changing capabilities which control both the steering and the acceleration/deceleration while ensuring that it is safe to do so using sensors. However, just like level 1, the driver is still responsible for the system overall and must have hands on the wheel and feet at the pedal while monitoring the environment. But, multiple controls are being executed in co-ordination by the system.

Level 3 – Conditional Automation

Until level 3, the human driver has always been responsible for monitoring the driving environment. While the car can make basic decisions on driving, the driver is responsible for ensuring safety.

Level 3, however, is where the role in monitoring shifts to the system instead of the driver. This means that the vehicle can perform most driving tasks, including general driving, parking, and route navigation while also avoiding obstacles and following road signs. However, despite the systems ability to drive the vehicle, human intervention is still required should something go wrong. Another way to better understand what level 3 automation is capable of is that the driver should be able to watch a movie while driving; the driver is alerted to potential problems and expected to react in a reasonable amount of time. 

Level 4 – High Automation

Level 4 is where a vehicle can drive itself from start to finish without any human intervention in controlled environments. For example, such a vehicle may be “geo-fenced” to main roads meaning that if the vehicle enters the wilderness (i.e. no roads), then it is required to be driven by a human. Such a system would include self-driving taxis with no driver, and while a human driver can intervene, the vehicle should be able to pull over safely during any event. 

Level 5 – Full automation

Level 5 is the sci-fi autonomous driving level whereby a vehicle has complete control in all environments. A driver can get out of their seat, go the back, make dinner, and host a tea party while the vehicle climbs up Mt. Everest at 45° to the plane. The vehicle makes all decisions; the human occupant provides no input and is never required to do so. Such a vehicle could be fitted to allow human driving, but this would more likely be an optional feature.

Where is humanity currently at?

There is no doubt that current technology allows for level 5 automation in limited cases (such as a race track). However, cars with “autonomous features” are rarely beyond level 2 (such as Tesla’s autopilot system), and all require constant human attention. While such levels may be able to drive a car autonomously, they cannot do so reliably and do not have the needed miles of experience (typically in the millions of miles). 

One vehicle on the market, the Audi A8, almost had level 3 capabilities, but the difficulty in implementation has seen Audi move away from level 3 automation for the time being. However, it is expected that their system will be introduced into the next generation of A8’s.  Tesla is also looking towards the development of level 3 systems as are other auto-makers. 

Fully autonomous vehicles will become dependent on a wide range of different technologies, including LiDAR, RADAR, SONAR, GPS, 5G, and Wi-Fi. With software-defined cars on the horizon, it won't be long before autonomous systems become commonplace, and the use of software-defined systems will enable older cars to take on newer technologies.

Read More

By Sam Brown