Six Levels of Autonomy for Self-driving Cars
Illustrated guide to different levels of autonomy for a self-driving car.
Autonomous vehicles have seen a significant improvement over the last decade. Since Tesla’s AutoPilot in 2015, there has been impactful research in the area from big companies like Ford, General Motors, Audi, etc. to successful startups such as Aurora, Agro AI, Pony AI, etc. With these advancements, a systematic way of defining the autonomy of a vehicle needs to be defined. In 2016, the Society of Automotive Engineers (SAE) released documentation, SAEJ3016, defining six different levels of automation for self-driving cars. In the year 2021, a detailed revision of the SAEJ3016 document was carried out that adds a few new terms clarifying the concepts and defining responsibility in the event of a failure.
In this article, we will go through these six levels of automation for self-driving cars. A summarized figure can be seen at the bottom of this article. The article also contains animated videos to help understand different terms related to self-driving cars.
Level 0: No Automation:
This is the baseline for other automation levels. Level 0 means no automation of any kind. The driver is responsible for all the tasks from steering to cruising.
Level 0: No drivers assistance or automation
Level 1: Partial Driver Assistance:
A level 1 vehicle includes partial assistance of driving. The term partial assistance is very broad and can be vague. So SAE defined it in terms of two important features. A level 1 vehicle has either of the following (but not both)
- Adaptive Cruise Control (ACC): Adaptive cruising is updated cruise control with a stop-and-go function. Adaptive Cruise Control has a stop-and-go function. When a slow-moving car comes ahead, the vehicle maintains a preset distance from the vehicle, until the slow-moving car moves away or speeds up. After that, the vehicle accelerates to the pre-set speed. The ACC does not disengage until the vehicle speed is reduced below a threshold. Hence ACC maintains the speed of the vehicle prioritizing the distance from the vehicle ahead.
- Lane Centering Assist: Conventional lane departure system warns the driver when the vehicle drifts off to the boundaries of the lane. The feedback is provided by steering wheel/seat vibration, warning sound, or screen display. Lane-centering controls the steering of the vehicle keeping it centered to the lane. It identifies the lane the vehicle is in and tries to maintain it by following its curvatures.
Both of these features can be seen in the animation below. The driver is responsible for carrying out driving actions and is required to be attentive and hands on the steering wheels at all times.
Level 1: Either Auto steering OR accelerating in limited conditions and areas
Level 2: Advanced Driver Assistance:
A level 2 vehicle is equipped with both adaptive cruise control and lane-centering feature. Hence steering, acceleration, and braking can be controlled simultaneously. This means that on the highway, the vehicle can carry out the majority of driving. The driver is still required to be attentive and with hands on the steering wheel. Tesla AutoPilot is classified as a Level 2 vehicle (as of the date the article was written, Jun 2021).
Level 2: Auto steering and accelerating in limited conditions and areas
Level 3: Conditional Autonomation:
Level 3 autonomous vehicle has advanced environmental perception. It can detect various objects around it such as traffic signals, traffic signs, pedestrians, buildings, etc. This detailed perception of the environment makes the vehicle advance enough to carry out situation-based tasks, such as stopping at the traffic signal and yielding for oncoming traffic, lane changing in case of a slow-moving vehicle ahead, traffic jam assistance, etc. Levels 3 and above are considered at automation, while the three levels below are considered as driver assistance. The driver is not required to keep their hands on the steering wheel all the time but should be ready to take over as soon as the vehicle autonomous system malfunctions.
Level 3: Partial automation in limited conditions and areas
Level 4: High Automation:
This is what people usually think when they hear the words autonomous driving. A level 4 car is capable of moving from point A to point B without human interference by safely following all the traffic rules. There is no human driver involved. The vehicle does all the driving. The steering wheel, accelerating, and braking pads are optional. The only caveat is that the vehicle can only drive when certain safety conditions are met. For example, the vehicle might not be able to drive itself autonomously due to bad weather, unknowns territories, etc. The vehicle’s driving is geo-fenced meaning that there are only certain regions where it will be able to drive itself.
Level 4: Complete automation in limited conditions and areas
Level 5: Complete Automation:
This is the ultimate autonomy for a vehicle. A level 5 vehicle can carry out all the driving tasks in all the conditions and areas. The vehicle is solely responsible for all the tasks and carrying them out safely. The steering wheel and accelerating/braking pads are optional.
Summary:
In this article, we looked at different levels of autonomy for a self-driving car. These levels are defined to track the advancements and set an understood expectation in the area of autonomous vehicles. These six levels of automation from Level 0 to Level 5 define the features, assigns responsibility, and clarifies the conditions in which those features are applicable. A detailed summary of these autonomy levels can be seen below
If this article was helpful to you or you want to learn more about Machine Learning and Data Science, follow Aqeel Anwar, or connect with me on LinkedIn or Twitter.
