The common misconception when it comes to automation is that it just happens. Like the Terminator stepping from a crackling energy bubble transported from the future, it just arrives fully intact, ready to go.
Roboticists are quick to point out that reality occurs much more slowly. Automation, instead, happens one small step at a time over the course of many years.
Such is the case with cars, which are on their way to becoming self-driving. But, as per the truism of robotic reality, it’s happening more gradually than some proponents may suggest. Also, the process has been under way for decades.
When it comes to vehicle autonomy, or a car’s ability to drive itself, the steps are generally represented as a scale comprised of six levels. The scale was developed by the U.S.-based Society of Automotive Engineers, an organization founded in 1905 that sets professional standards for the auto industry.
Here’s a guide to the SAE’s vehicle autonomy scale, along with the latest expert thinking on when the six levels either were achieved or might be.
Level zero
Cars with some form of alert systems and limited automatic controls, such as lane departure warning and blind spot intervention, classify as level zero.
Such features are quite old. Lane departure warning, for example, has been around since the early 2000s. Blind spot intervention, where the car momentarily steers itself away from an oncoming vehicle, was first introduced nearly a decade ago.
Many started as options on luxury cars, but are gradually becoming standard throughout manufacturers’ lineups. Still, the majority of vehicles currently on the road are considered level zero and effectively have no ability to drive themselves.
Level one
If the driver is controlling either the steering or the speed of the vehicle, but not necessarily the other, it’s level one.
Adaptive cruise control, where the car uses radar and other sensors to automatically maintain a constant speed in relation to the vehicle in front of it, is a good example. Parking assistance, where the car steers itself but the driver controls the speed, is another good example. Both were first introduced in the 1990s and are now becoming commonplace.
Regardless of feature, drivers of level one cars must have their hands on the steering wheel and be ready to take control at a moment’s notice.
Level two
Level two cars are able to take control of both speed and steering, allowing drivers to take their hands off the wheel and feet off the pedals for short periods of time. The driver still needs to pay attention and be ready to take over quickly, but it’s otherwise a limited form of self-driving.
Many new cars on the market have such capabilities, with manufacturers offering them as options. Nissan’s ProPilot Assist or Tesla’s AutoPilot are both examples. Competition between automakers is fierce and the functions are spreading quickly as a result, meaning that level two could quickly become available in most new vehicles over the next few years.
Level three
Here’s where things start to get interesting, since it’s where the future comes into play and where the vehicle becomes responsible for monitoring more of its environment than its driver.
The SAE refers to level three as the “eyes off” phase, meaning the driver can take his or her eyes off the road and focus on something else, such as their smartphone. The driver must still be able to take over for limited situations, such as driving around an unexpected construction zone or accident.
“It’s going to tell you, ‘I’m not able to handle this scenario for whatever reason and you have to take control,’” says Ross McKenzie, managing director of the Waterloo Centre for Automotive Research in Waterloo, Ont. “In the absence of that warning, I wouldn’t need to have my hands on the wheel or my eyes on the road.”
German manufacturer Audi touted its 2018 A8 luxury sedan as the first commercial car capable of level three self-driving, but what’s possible and what’s permissible are two different things. Traffic regulators in many countries, including Canada, aren’t allowing level three cars on the roads except for testing.
Mr. McKenzie expects level three cars will become commonplace on highways within the next 10 years and on city streets within 15.
Angela Schoellig, head of the Dynamic Systems Lab at the University of Toronto, says it could be sooner – within the next two or three years on highways. However, as with each of the SAE’s advanced autonomy levels, geography will be primary determinants. Canada is likely to see later deployment as a result.
“Is it winter on a highway in Toronto or is it a straightforward neighbourhood somewhere in Florida or California?” she says. “This isn’t necessarily reflected in those autonomy levels. With very few exceptions, everyone has tested in the best possible conditions.”
Level four
The thought of sleeping behind the wheel – safely – seems too good to be true, but it’s what’s being promised by what the SAE refers to as the “mind off” phase.
Level four cars will be able to fully drive themselves, including in uncertain situations, at which point they might safely park and request the human – who can’t really be called a driver anymore – to take over. The idea is that the human is responsible for only a small portion of the overall driving.
Though companies including Google’s Waymo unit and General Motors are testing level four cars on city streets, observers say the technology realistically isn’t there yet. It could be another 15 years before such vehicles are safely deployed as a result, or up to 25 years for more complex journeys that include highways, city streets and other road types.
“The capabilities are not entirely self-sufficient. There isn’t enough intelligence built into the software to handle every possible scenario,” Mr. McKenzie says.
“The real world is so complex and has so many facets,” Dr. Schoellig says. “Guaranteeing that everywhere in the world is extremely difficult.”
Level five
Cars without steering wheels or pedals seem like science-fiction and widespread deployment could indeed be far off, but they are likely to see more constrained use in the near term.
Fully robotic cars are already being deployed in geographically limited zones that don’t otherwise have vehicle traffic. The University of Michigan, for example, launched driverless shuttles on campus this year. The University of Waterloo is planning to do the same next year.
Colin Dhillon, chief technical officer for the Automotive Parts Manufacturers' Association, expects fully robotic level five cars will be introduced as early as 2020 and 2021, but expecting them on any sort of a wide scale isn’t realistic until 2025 to 2030.
The industry is hoping that regulators, who are keeping a close eye on developments, continue to allow the expansion of tests.
“The introduction is already happening,” he says. “The challenge is not to stifle the whole movement.”