Autonomous cars have already logged millions of miles, yet they still feel like something out of The Jetsons (even George had to manage a stick shift). Part of the future shock comes from the fact we still don’t know the exact meaning of “autonomous” or synonymous phrases like “self-driving” and “driverless.” If a car has a human sitting in the front seat waiting to take over at a given moment, is it truly “driverless”? How “autonomous” are those slick Lexuses that automatically park themselves in commercials (and, I assume, wealthy neighborhoods) versus Tesla’s new autopilot-mode fleet?
It’s a discussion the auto and tech industries are currently having. In 2014, the Society of Automotive Engineers (SAE) drafted a taxonomy to evaluate cars based on the sophistication of their autonomous features. The result is a system to rank cars’ intelligence from “high school hooptie” to “Lindsay Lohan star vehicle.” “It is a reference point in many ways for the industry,” says Jeremy Carlson, a senior autonomous driving analyst at IHS. (The National Highway Traffic Safety Administration has released a similar set of guidelines.)
Having these standards is more important than ever. Investment in driverless cars has exploded in an effort spearheaded by tech giants. Traditional automakers, sensing a sea change in the ways people use vehicles, are also getting in on the action as they prepare to compete (or collaborate) with Silicon Valley. This week Volkswagen poured $300 million into the European ride-hailing service Gett, while Toyota invested an undisclosed sum in Uber. These deals follow Apple’s $1 billion investment in the Chinese taxi app Didi Chuxing, General Motors’ $500 million bet on Lyft, and BMW’s tie-up with Chinese internet giant Baidu.
The market for autonomous cars is suddenly very crowded, yet showrooms are completely empty. There’s still no firm date for when you’ll be able to buy a self-driving car at a dealership or hail one to take you home from the bar. But understanding the automation levels helps provide a clearer picture of the progress carmakers have made so far, and the many miles they still have to travel. Here’s the breakdown:
Level 0: No Automation
This includes not only the the Model T (duh) but also fairly modern vehicles with features like cruise control or collision warning systems. There’s nothing in these vehicles that’s not directly controlled by the driver, Carlson explains. In other words, cars have been dumb for a very, very long time.
Level 1: Driver Assistance
Driver assistance actually has been around since the ’90s in the form of adaptive cruise control, which lets cars adjust their speed on the highway amid moving traffic. More recently, automakers have introduced lane-assist technology that will keep a car between the dotted lines if it starts to drift, as well as vehicles that can park themselves.
Level 2: Partial Automation
If MTV comes through with a Pimp My Ride reboot, an L2 is the best car Xzibit could offer a flabbergasted teen. L2s can control both acceleration-deceleration and steering, which means they can basically drive themselves while on a highway. The autopilot mode on Tesla’s Model S is a great example of this technology — people have even napped in their Teslas while on autopilot, though the company does not suggest this. A human is supposed to supervise all of a vehicle’s movements at Level 2.
Level 3: Conditional Automation
This is the automobile’s bone-toss moment. The functionality of a Level 3 vehicle is largely similar to Level 2, according to Carlson. The primary difference is the car itself is responsible for driving and must request that the driver intervene if something goes wrong. “It’s about the driver stepping back and relinquishing primary control of the vehicle, but in some cases, it’ll be necessary for that vehicle to call the driver back in, to get them up to speed and to have them take control again,” Carlson explains. “In many cases that’s going to be a safety-critical situation.”
Carlson expects many traditional automakers to roll out Level 3 vehicles in the next two years because they offer more convenience than less advanced cars without being as complex to build as higher-level ones. However, some companies, such as Google and Ford, believe that requiring a human driver to take control at a moment’s notice is actually more dangerous than ceding total control to a computer. That’s why they’re skipping Level 3 and going straight to …
Level 4: High Automation
Those pod-like Google cars that have been zipping around Mountain View? Those are L4s, driverless cars that don’t require humans at all. They can take passengers to a requested location whether it requires speeding down a freeway or making a left turn in congested city traffic.
L4s are the cars that have the potential to truly change how we travel. But they’ll have restrictions on where they can go, and when they can go there. They’ll likely be limited to specific, well-mapped areas via geo-fences, and they may not be able to handle inclement weather, which throws off their sensors. That makes them less than ideal for personal ownership, but perfect for a taxi service like Uber or Lyft (Uber reportedly wants to buy up a ton of driverless cars ASAP). IHS expects a limited number of L4 cars to be available for purchase by 2020, mostly to be used by ride-sharing companies. Ford and Google have also pegged the year as a viable launch window.
Level 5: Full Automation
A car that, in the words of SAE, can drive itself “under all roadway and environmental conditions.” It’s basically the Batmobile or, if imbued with a strong sense of moral righteousness, Optimus Prime. Point is, we’re very far off from a vehicle that can truly navigate any patch of land it comes across. “Level 5 is like an endpoint,” Carlson says. “We’re talking about decades upon decades to cover every single situation that might possibly present itself.”