If a level 3 car crashes, is the car responsible, or is the driver? It's ambiguous, and that's a huge problem.
Is a driver responsible when a Level 3 vehicle crashes? According to Level 3 definitions, it’s hard to tell.
That there is still no definitive answer to the question is seriously troubling, and so the debate over SAE Level 3 is flaring back up. The controversy is not about the nuts and bolts in the SAE J3016 technical standard put together by SAE International. Instead, the spotlight shines on a broad spectrum of interpretations and loose claims that automakers and tech suppliers are conjuring as they call their new vehicles the “Level 3 of Driving Automation.”
The debate within the industry spilled into the public arena with Tesla’s recent admission that neither Autopilot nor Full Self-Driving (FSD) will be autonomous systems, an admission that set off an enormous tweet storm.
This acknowledgement emerged when a Freedom of Information Act petition led to the release of an email thread between Tesla’s legal team and the California Dept. of Motor Vehicles. Tesla’s legal team told the DMV late last year that its ‘Autosteer on City Streets’ feature “continues to firmly root the vehicle in SAE Level 2 capability and does not make it autonomous under the DMV’s definition.”
Forget all Elon Musk’s bravado on FSD; Tesla is now confessing that its vehicle is still looking up at Level 3.
This was especially jarring, given that Tesla offered its customers a Full Self-Driving option — a $10,000 option you can pre-pay for for your Tesla, leading consumers to believe that their brand new Tesla would eventually live up to its “self-driving” descriptive.
Meanwhile, earlier this month, Honda surprised the global auto market by launching Honda Sensing Elite, an enhanced version of its advanced driver-assistance system (ADAS). The Japanese company called it “the first commercially available SAE Level 3 system” in the world, although the car itself is an extremely limited edition available only in Japan.
Over the last several years, auto industry observers and human-machine interface (HMI) specialists have voiced concerns over Level 3. The growing consensus is that automakers might do well to skip Level 3 altogether, due to vehicle-to-human-driver handover issues that have defied solution.
Phil Koopman, co-founder of Edge Case Research and professor at Carnegie Mellon University, is skeptical about Level 3, but isn’t dismissing it yet, either.
In opening remarks at the Roadmap to Next-gen EV&AV conference last week, Koopman stressed, “I’m not saying Level 3 shouldn’t happen, but I’m saying you can’t communicate what Level 3 is to average drivers” in a way it’s helpful to them. And that’s a huge problem.
In his view, SAE J3016 is an “engineering standard by engineers and for engineers.” For “ordinary audiences, you need a different framework” for driving safety, Koopman insists. He recently proposed a “User’s Guide to Vehicle Automation Modes,” proposing four categories of vehicle operation: Assistive, Supervised, Automated, and Autonomous.
During his speech at the virtual event, Koopman explained, “In the Assistive mode, the human role is traditional driving. The human takes care of driving safety… The next mode is the Supervised mode in which the humans are responsible — not for driving since the car does the actual driving task. But the human driver is responsible for driving safety. And that means eyes on the road.”
He explained that these two modes – Assistive and Supervised – are driver assistance. An “important line is crossed” when entering automated mode in which human drivers “are allowed to take eyes off the road because the car is entirely responsible for both driving and driving safety.”
EE Times asked Koopman why his four modes exclude a Level 3 equivalent.
He explained that although the SAE levels are well-defined, it’s a challenge even for people knowledgeable about safety and driving to digest what exactly the standard says — and that’s before getting to the footnotes.
“I struggled with it,” Koopman acknowledged. “I don’t see how someone without a lot of expertise who hasn’t even read the standard has any chance.”
Second, he noted, Level 3 poses a distinction based on “which functions belong to the car and which functions to the driver.” This is a proven recipe for confusion. “From a driver’s point of view, you still have to pay attention in Level 3. So, a driver would be hard pressed to tell the difference between Level 2 and Level 3, because some failures count, and some don’t. It’s a technical distinction that I’m not even sure I could tell which one [that belongs to the driver].”
Level 3 doesn’t go away, stressed Koopman. “But what you have to do is to ask if the driver is responsible for safety. It’s a yes/no question.”
In his opinion, there are assumptions made in Level 3 that in practice are so nebulous as to be useless, such as “you’re responsible for safety, except when you’re not,” or “when you’re not responsible depends on your understanding of the engineering details of your particular car.” Such distinctions are guaranteed to be unhelpful to human drivers.
Regulations on Level 3?
How regulators eventually deal with Level 3 remains to be seen. While J3016 defines matters clearly for engineers, Koopman said that if regulators don’t know what the standard exactly says, they might formulate regulations that contradict standards, or allow standards that materially override the regulations.
He reminded that he is not talking about any particular regulation. “But if you want to say Level 3, you have to buy into the standard.” After all, “The standard includes that the driver might need to take over control with zero warning time. And [with] no notice from the Automated Driving Systems, you’re supposed to just notice the cars are doing something weird, and you must intervene.”
Undoubtedly, Level 3 is subject to interpretation. But the standard itself is quite clear, noted Koopman. “The problem is that very few people talk about what’s actually in it.”
People know what they want in Level 3. Many think they know what’s there. But that’s different from what’s in it. “You actually have to read the standard.”
The point is that the driver needs to know, whether he or she is supposed to pay attention. “It’s a yes/ no question,” Koopman reiterated.
Regulations or vendors’ manuals might say Mr. Driver, you can take your eyes off the road. But when you do, look out, because “we’re going to blame you for taking your eyes off the road if there’s a crash.” Koopman said, “I don’t know that’s what it says. But my understanding is it’s an issue that has not quite settled yet.”
Bottom line: You can’t have it both ways. If paying attention is optional, then blaming an inattentive driver for an accident is an untenable option. “I know that’s not a system I’d be happy to see,” said Koopman.
This article was originally published on EE Times.
Former beat reporter, bureau chief, and editor in chief of EE Times, Junko Yoshida now spends a lot of her time covering the global electronics industry with a particular focus on China. Her beat has always been emerging technologies and business models that enable a new generation of consumer electronics. She is now adding the coverage of Chinaâ€™s semiconductor manufacturers, writing about machinations of fabs and fabless manufacturers. In addition, she covers automotive, Internet of Things, and wireless/networking for EE Times’ Designlines. She has been writing for EE Times since 1990.