For almost a century, inventors have tried to build self-driving vehicles that don’t need human drivers to get around. The first such cars appeared in the 70s and 80s as prototypes. Now, it seems in the next few years we may finally be in a situation where autonomous vehicles are ready to become a reality.
In 2016, SAE International (formerly the Society of Automotive Engineers) broke down the path to driverless cars to six different levels of autonomy. This classification system replaced a similar one the U.S. Department of Transportation National Highway Traffic Safety Administration (NHTSA) published in 2013, and now both agencies use the SAE version exclusively.
It’s a bit controversial and confusing to both manufacturers and consumers, so we’ll explain them below.
What Are Self-Driving Vehicles, and What Do the Levels of Autonomous Driving Mean?
In the narrow sense, self-driving vehicles are just ones that can navigate their driving environments without human input. They may use cameras, sonar, radar and lidar to detect obstacles on the road, and they may use GPS and other coordinating systems to reach their destination. As of 2019, most cars still lack automated driving systems, though many have advanced driver assistance systems, such as automatic emergency braking and cruise control.
Full autonomy doesn’t look too far away, however, and some companies, such as Tesla and Waymo, already sell cars with significant hands-off features. Technological, cultural and legal barriers have made the transition to driverless vehicles hard to predict, so projections of automation aren’t set in stone. Still, the SAE’s six driving levels serve as a useful schematic for thinking about where this technology is headed, and automakers and policy experts can use it to make classifications when needed. The SAE treats each level as a benchmark for the range of driving tasks an autonomous driving system can achieve, from merging lanes to high-speed driving.
Each level has a number 0 through 5, with higher numbers indicating an increased level of autonomy in a vehicle. These numbers may appear on a company’s marketing materials, in a popular news article or in a press release about a new offering from a car company.
Too Many Levels?
Many believe levels of self-driving cars should be simplified into two levels to avoid confusion:
- Driver-in-Control – the vehicle has sophisticated semi-automated functions for driver assistance, but the driver is still ultimately responsible and in control (Level 2).
- Vehicle-in-Control – the vehicle is fully autonomous and does not require a driver. The vehicle will likely be restricted to certain locations and conditions (Level 4).
Here’s a chart with an overview of the current SAE Self-Driving Automation levels and with the two most important ones (Level 2 and Level 4) called out:
Check out what each self-driving car level means, and discover what to expect from a rapidly approaching future where autonomous cars are the norm.
Self-Driving Car Level 0: No Automation
Driver: Licensed driver with full attention and total control
Vehicle: No automation
Almost all cars ever made, from the Benz Patent-Motorwagen built in 1885 to the 2010 Toyota Prius, have lacked self-driving capacities of any kind. Such “Level 0” cars, as the SAE classifies them, may have simple warning or intervention systems, but they all require the driver to brake, steer, park, maneuver and accelerate manually.
At level 0, you have complete control over your car or truck. Your car may offer warnings, such as an audio alarm when you back up and get to close to another car, but you must apply the brake to prevent hitting the other vehicle.
Self-Driving Car Level 1: Driver Assistance (“hands-on”)
Driver: Licensed driver with full attention required
Vehicle: Semi-automated systems, like cruise control.
Level 1 cars have mechanisms that help with steering or speed, though not both at the same time. The driver must switch from one function to the other. Sometimes called the “hands-on” level, because drivers are expected to keep their hands on the steering wheel, this level of automation splits control of the vehicle between the driver and software. Adaptive cruise control is a common feature of Level 1 self-driving cars. When adaptive cruise control is activated, the cars speeds up and slows down on its own while the driver steers the car. Driving technology features for Level 1 automated vehicles can also include lane keeping assistance, which alerts the vehicle of drivers moving in and out of lanes; automatic emergency braking, which alerts the driver to a crash and permits full braking capacity; and parking assistance, which steers the car while the driver accelerates or decelerates.
Cars with adaptive cruise control include the 2014 Chevy Impala, Ford F150 and BMW 5 Series Sedan, which includes a Stop&Go function to automatically adjust the car’s distance to vehicles behind and in front of it. BMW 5 Series Sedans also include the BMW Personal CoPilot system and Collision and Pedestrian Warning with City Brake Activation system for extra convenience. Both of these features lower the risk of collisions dramatically. Many Level 1 cars have road edge detection, which tells the driver if they’re too close to the side of the road. In Europe, all newly launched vehicles must have lane-keeping assistance, automatic emergency braking and road edge detection to qualify for a four-star safety rating or higher.
Self-Driving Car Level 2: Partial Automation (“pay attention”)
Driver: Licensed driver with full attention required
Vehicle: Semi-automated systems, like steering, speed and braking.
This is the most common level in consumer-level cars with autopilot today and describes vehicles that can control steering and speed features. However, the driver is still responsible for the vehicle and must pay attention at all times, even if the Driver Monitoring System (DMS) allows the driver to be “hands-off” the steering wheel.
Tesla’s autopilot for the Model S, for example, can maneuver through traffic and accelerate or decelerate on its own. Human intervention still plays a big role in Level 2 cars, however, as the driver must keep their hands on the steering wheel in case the system fails to respond to a problem.
The General Motors Super Cruise feature, on the other hand, lets you can take your hands off the wheel, though cameras in the vehicle monitor your eyes and disable cruise control if you aren’t paying attention.
Other examples of Level 2 features include Audi Traffic Jam Assist, which guides the vehicle through slow-moving traffic, and Mercedes-Benz Driver Assistance, which can maintain the vehicle’s distance between cars at speeds of up to 210 km/h. Volvo Pilot Assist is another such feature, providing simultaneous steering assistance and distance-keeping from other cars.
For more current Level 2 vehicles, see Cars with Autopilot.
Self-Driving Car Level 3: Conditional Automation (“eyes off”)
Driver: Licensed driver with ability to take control, if required
Vehicle: Primary driving functions are automated under some conditions.
Conditional automation, Level 3 of the SAE classification system is the most controversial and debated level since lets the driver focus on things other than driving tasks while in the car, while still being ultimately responsible.
With Level 3 technology, the car doesn’t need your hands to stay on the steering wheel or your eyes to stay fixed on the road. You can read a book, watch a movie or play around on your phone. The only requirement is that you stay in the car, as Level 3 autonomous vehicles can still prompt the driver to react to problems they can’t deal with on their own, such as navigating road construction sites.
This level is so controversial since the driver is required to be present and responsible for the vehicle even though the vehicle is primarily in control and the driver may not be aware of the situation at all times. If there is a problem, this would make it very difficult for the driver to suddenly regain situational awareness and context of the situation when often split-second decisions are required.
Therefore most manufacturers will likely skip this Level 3 and make self-driving more black-and-white, with either the human driver is responsible and in-control (Level 0,1,2), or not (Level 4 and 5). Others, like Mobileye, are trying to improve on Level 2, calling their technology “Level 2+”, potentially adding more confusion.
That said, some auto manufacturers are pushing forward with marketing this level. For example, the Audi A8 has “Level 3 features”, though its Traffic Jam Pilot system, which lets the car navigate traffic autonomously, was canceled due to legal issues in the U.S. Level 3 cars can apply emergency brakes on their own, and they’re especially competent on highways, where pedestrians and intersections are less of an issue. Level 3 vehicles usually have audible warnings to alert drivers to take control of the car. If the driver fails to respond, the vehicle may pull to the side of the road until the driver re-engages.
Cars with this level of automation are still rare, as of 2020, because they require much more extensive safety features than Level 2 cars to meet regulatory standards. Should it become commonplace, a Level 3 system would mark a crucial milestone in the development of self-driving cars, as it would permit people to do work during their commutes, potentially shortening their workdays. Mobileye, a subsidiary of Intel, is working on a redundancy system combining lidar, radar and path-finding software to make Level 3 more feasible. BMW, meanwhile, hopes to have Level 3 cars on the road by 2021. BMW’s Personal CoPilot system is projected to let drivers cede steering, acceleration and braking to the car without requiring their full attention. Ford, by contrast, hopes to jump from Level 2 to Level 4 to avoid designing a system that requires any driver intervention at all.
Self-Driving Car Level 4: High Automation (“mind off”)
Driver: No Licensed driver required (passengers OK)
Vehicle: Primary driving functions are automated under most conditions.
Level 4 of the SAE classification system covers vehicles that don’t need any human interaction, even when it comes to safety. They can perform all driving functions on their own, allowing the driver to relax or sleep. No Level 4 car is commercially available in the U.S., but many companies are in the process of developing one, such as Google, Uber, Lyft and Waymo. Waymo’s Early Access program allows people to ride Level 4 cars in special locations in the country, such as Arizona, or in low-intensity situations, such as traffic jams.
Because Level 4 technology often still has trouble navigating through rain and snow today, the automation in current Level 4 cars may deactivate if the weather turns poor, forcing the driver to re-engage the driver controls or automatically parking somewhere if they refuse. In Level 4 vehicles, the driver can retake control of the car whenever they want. Also, Level 4 cars can handle road construction sites and other tricky circumstances as long as they fit the operating parameters of the car’s software.
Honda plans to have a Level 4 car on the road by 2026, and Google has built Level 4 prototypes, such as its Firefly pod-car, even though its vehicles still rank between Levels 2 and 3. Level 4 automation may radically expand popular support for self-driving cars in cities, since it means ride-sharing fleets with autonomous cars can advertise the technology to millions of people.
Self-Driving Car Level 5: Full Automation (“chauffeured”)
Driver: No Licensed driver required (passengers OK)
Vehicle: Primary driving functions are automated under all conditions.
Full automation is the ultimate goal of most self-driving car companies. At this level, human intervention is never necessary, even in the most challenging environments, such as dirt roads. In short, there are no drivers in Level 5 cars, only passengers. However, drivers can still take over whenever they want. In more advanced forms, such cars may resemble mobile living rooms, with comfortable seating that makes full use of the interior, and they may include amenities such as personal computers and refrigerators. Eventually, such cars may even have fold-out beds that allow passengers to sleep during road trips or take naps on the way home from work. Since licenses aren’t necessary for Level 5 cars, anyone could take them for a spin.
People with disabilities that make driving difficult or impossible, such as blindness or paralysis, might benefit the most from full automation. Also, ride-sharing fleets could transport passengers much more cheaply, as they wouldn’t need to pay drivers. Although commercial passenger cars at this level aren’t on the road yet, Nuro’s Level 5 grocery delivery truck is. This truck delivers food to your house and drives back to its parking area once it’s done. In addition, Waymo is currently developing a fleet of 600 Level 5 Chrysler Pacific hybrids, and BMW is working with Mobileye to get the first fully self-driving car into serial production.
Early Level 5 cars may stay confined to slow-paced urban areas while companies work out the final kinks of the technology and build up their library of real-world performance data.
Autonomous driving technology promises to be disruptive. As a result, it poses a lot of legal challenges. Who is responsible if a self-driving car hits someone: the driver, or the programmer who wrote the algorithm? Should a self-driving car get to prioritize which people to avoid hitting if it loses control? Lawyers, judges and policymakers need to hash out these messy details, and the companies designing these new cars need to tweak their algorithms to accord with new laws.
In addition, self-driving car companies need to build more public trust in their products. That takes time, and people tend to drag their feet when it comes to accepting new technology. Companies can be overly optimistic as well. Tesla promised that its cars would have full autonomy by 2018, and they still haven’t reached Level 4, as of 2019.
The overall forecast is clear, though. Driverless cars are expected to have far-reaching impacts on the U.S. economy. For example, the technology may allow for an auto-sharing economy in which people and organizations can share vehicles with ease. Mobileye even wants to develop multipurpose self-driving software for automakers to adopt without having to develop the technology in-house.
Once fully self-driving cars are widespread, businesses can transport goods much more cheaply, since they don’t have to hire professional drivers, and taxi services are apt to get cheaper for the same reason. The rate of accidents is bound to fall, too, since autonomous vehicles are much less error-prone than human beings. As a result, modern liability law and insurance plans may change significantly.
A major roll-out of self-driving cars is still far off, though. Most cars on the market with autonomous driving software are much too expensive for the average car buyer, and decades could pass before higher-level features reach the non-luxury parts of the automotive industry. Plus, public fears that self-driving cars may put millions of drivers out of work has led to push-back against the technology. Long-distance truckers, in particular, may have good reason to be worried, as their jobs are likely to be automated early. Still, the economic benefits are going to have widespread benefits overall. If self-driving taxis get cheap enough, for instance, many people can commute to work from far away without relying on a personal car or public transport.
Self-driving cars are poised to disrupt every aspect of society, from the cost of travel to employee productivity. SAE’s six-level classification system provides solid benchmarks for tracking how far the technology has come and where it’s headed. In a few decades, cars, as most people think of them, may become something else entirely.