By Lance Eliot, the AI Trends Insider
An exasperating situation exists about the nature and scope of self-driving driverless autonomous cars.
The existing published standard that depicts the automation levels has become a well-worn and oft-cited cornerstone to discussing and understanding the aspects of self-driving cars.
In a sense, that is quite handy, and we can be thankful that a standard exists, albeit the mutterings and outright complaints by some insiders regarding the weaknesses and warts of the standard (more on this in a moment).
Unfortunately, despite the good facets of having such a standard, the levels are often mistakenly portrayed in the news and other media outlets.
This creates confusion.
Lots of it.
In addition, this then confounds efforts to compare the various experimental tryouts underway on our public roadways, and pretty much messes with the minds of the regulators and the public-at-large about what a self-driving car does or ought to do.
I’d like to take a moment herein and provide some clarification about the standard and especially the role of Level 4 and Level 5.
The shorthand notation used to refer to the standard is to say that it is the SAE J3016.
Let’s unpack that notation.
This standards document is one of many promulgated by the SAE (Society of Automotive Engineers). To easily refer to the numerous standards associated with automobiles and other types of vehicles, they are each assigned a numbered reference. The one that focuses on the automation levels for self-driving cars has the designation of J3016.
Thus, the quickest way to refer to the standard is by simply referring to SAE J3016.
Overview Of The Levels
The SAE J3016 has a somewhat lofty and technical title of “Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles” and fits within the rubric of the Surface Vehicle Recommended Practice aspects.
I’m not going to cover the levels that require the presence of a human driver, which are Level 0, Level 1, Level 2, and Level 3.
By-and-large, those lower levels of automation are rather self-apparent and involve having automation that augments a human driver.
Cars that require a human driver will often co-share the driving task, meaning that human drivers and automation of the car are supposed to work hand-in-hand while driving the car. These types of cars are properly described as being semi-autonomous vehicles rather than autonomous vehicles (AVs), and typically contain a variety of automated add-ons that are known as ADAS (Advanced Driver-Assistance Systems).
Since semi-autonomous cars require a human driver, the adoption of those types of cars won’t be markedly different than driving conventional vehicles, so I’m not going to include Level 0 to Level 3 in this discussion.
For semi-autonomous cars, it is equally important that I mention a disturbing aspect that’s been arising, namely that in spite of those human drivers that keep posting videos of themselves falling asleep at the wheel of a Level 2 or Level 3 car, we all need to avoid being misled into believing that the driver can take away their attention from the driving task while driving a semi-autonomous car.
You are the responsible party for the driving actions of the vehicle, regardless of how much automation might be tossed into a Level 2 or Level 3.
You might be wondering which vehicles today are at which level of automation.
As an example, existing Tesla’s are considered at a Level 2 and are gradually via the expansion of AutoPilot becoming closer to Level 3.
Meanwhile, Waymo and many others are aiming at Level 4 and Level 5 (I mention just Waymo because they are generally accepted as the furthest along on such efforts and are also widely known among the public, but there are plenty of other Level 4 and Level 5 efforts underway).
When referring to Level 4 and Level 5, I’ve found it handy to refer to those topmost levels as being true self-driving cars.
True self-driving cars are ones that Artificial Intelligence (AI) drives the car entirely on its own and there isn’t any human assistance required during the driving task.
There is not yet a true self-driving car at Level 5, which we don’t yet even know if this will be possible to achieve, and nor how long it will take to get there.
Meanwhile, the Level 4 efforts are gradually trying to get some traction by undergoing very narrow and selective public roadway trials, though there is controversy over whether this testing should be allowed per se (we are all life-or-death guinea pigs in an experiment taking place on our highways and byways, some point out).
I’m guessing you are likely curious to know more about how Level 4 and Level 5 are the same and how they differ, so let’s jump further into the details.
For my framework about AI autonomous cars, see the link here: https://aitrends.com/ai-insider/framework-ai-self-driving-driverless-cars-big-picture/
Why this is a moonshot effort, see my explanation here: https://aitrends.com/ai-insider/self-driving-car-mother-ai-projects-moonshot/
For more about the levels as a type of Richter scale, see my discussion here: https://aitrends.com/ai-insider/richter-scale-levels-self-driving-cars/
For the argument about bifurcating the levels, see my explanation here: https://aitrends.com/ai-insider/reframing-ai-levels-for-self-driving-cars-bifurcation-of-autonomy/
Co-Sharing The Driving Task Versus Solo Automation
For Level 4 and Level 5 true self-driving vehicles, there won’t necessarily be a human driver involved in the driving task.
This is what makes them so special and catches our attention and interest.
We’ve all seen science fiction movies that have cars that drive via AI only (no human driver), and it’s quite a promising technical achievement, along with freeing humans from having to know how to drive or arrange for a human driver when wanting to use a car.
Many hope that human driverless cars, which I’m referring to as true self-driving cars, will bring forth mobility-for-all, enabling those that are mobility disadvantaged or mobility marginally to finally have ready access to riding in cars.
This discussion about human drivers brings up one source of confusion about Level 4 and Level 5.
The formal definition says that there is no requirement that a human driver must be available for a Level 4 and nor for a Level 5 self-driving car.
Take a moment to reflect upon that use of the word “must” and you’ll realize that just because you might say that something doesn’t have to happen, it does not preclude allowing it to happen.
In other words, automakers and self-driving tech firms are able to decide to allow for human drivers to function inside a Level 4 or Level 5 self-driving car, if wishing to do so.
Perhaps a particular self-driving car might have driving controls included for humans that might want to drive the self-driving car, choosing to do so whenever they might please. Likewise, if the self-driving car does have human accessible driving controls, the AI might decide to hand over the driving to the human driver, assuming that a human driver happens to be present in the car at the time of seeking to do the hand off (another provision would be to allow for remote drivers).
There is a rub to this.
If you believe that the AI will be a safer driver than human drivers, presumably due to the aspect that the AI won’t drive drunk and otherwise drive in human faltering ways, you would aim to intentionally prevent human drivers from being able to drive.
By opening the door to the notion of allowing humans to drive a Level 4 or Level 5, a true self-driving car is said to potentially open a Pandora’s box.
As such, many are intending to remove driver controls from within true self-driving cars.
This would not only inhibit a human driver from trying to take over the driving from the AI, it also frees up the interior of the car to be redesigned. For example, future designs of such cars showcase that without the need for a steering wheel and pedals, you can reconfigure the interior, perhaps having swivel seats or reclining seats to allow passengers to take a nap while riding in a self-driving car.
There are valid reasons to consider allowing for human driving at Level 4 and Level 5.
Suppose the on-board AI system becomes messed-up or frozen, perhaps due to a car accident that involves getting rear-ended and harms the computer processors, and there’s no means to now drive the car.
How will someone get the self-driving car off the roadway and out of traffic when it has become a multi-ton dead weight without any driving capacity?
The human driving controls aspect is a tradeoff that is still being debated.
In short, true self-driving cars at Level 4 or Level 5:
- Are not required to allow for human driving
- Are not required to disallow human driving
- May allow human driving to occur if the automaker wishes to include it
- Some worry that allowing for human driving is a bad idea
- Removal of driving controls also offers to free up space inside the car
- Automakers and self-driving tech makers are allowed flexibility in this regard
For why remote piloting or operating of self-driving cars is generally eschewed, see my explanation here: https://aitrends.com/ai-insider/remote-piloting-is-a-self-driving-car-crutch/
To be wary of fake news about self-driving cars, see my tips here: https://aitrends.com/ai-insider/ai-fake-news-about-self-driving-cars/
The ethical implications of AI driving systems are significant, see my indication here: https://aitrends.com/selfdrivingcars/ethically-ambiguous-self-driving-cars/
Be aware of the pitfalls of normalization of deviance when it comes to self-driving cars, here’s my call to arms: https://aitrends.com/ai-insider/normalization-of-deviance-endangers-ai-self-driving-cars/
Level 4 Versus Level 5: Key Differences
Okay, so both Level 4 and Level 5 are about being driverless, though this is not a hard-and-fast rule since the maker of such self-driving cars can opt to allow for human driving.
Next, let’s consider what is decidedly the biggest and most important difference between Level 4 and Level 5.
It has to do with the scope of where and when a true self-driving car is going to drive.
Let’s ponder scope aspects.
Suppose you are trying to invent a new kind of screwdriver.
You work on it and are able to make it work on screws that have a flat or slotted top. That’s great, but it won’t yet work on screws that have a crosshead, known as Phillips.
Is the screwdriver useful even though it can’t yet handle crosshead screws?
Sure, the screwdriver is still handy, though its scope is narrow, and you’d like to somehow have the screwdriver be functional for crosshead screws too.
Maybe devising a screwdriver that can handle both types of screws is hard to figure out.
Indeed, perhaps you come up with a second screwdriver that is devoted to crosshead screws, and thus now have two screwdrivers.
Notice that both types of screwdrivers have the same overarching purpose, yet for now, they are each of their own types.
Your hope is to someday produce one screwdriver that can handle all types of screws, including flat top, Phillips, and even others like hex key, etc.
What does this have to do with Level 4 and Level 5 self-driving cars?
Level 5 is easy to understand, it is essentially a true self-driving car that can go anywhere and at any time, assuming that a human driver could do the same.
I mention the caveat that the Level 5 goes wherever and whenever a human driver could do the same since we aren’t going to require the Level 5 to go beyond what humans could do in terms of reasonably driving a car.
For example, if a human driver can’t drive a car across an unnavigable rushing stream, it’s not fair to assume that the AI could do so since as a driving task the driving effort has to do with being able to drive the car in reasonable ways. Adding AI to a car doesn’t mean it can magically fly or leap across an otherwise unnavigable rushing stream.
Level 5 is what most people have in their minds when you tell them that true self-driving cars are going to one day be on our roadways.
How does Level 4 differ?
Level 4 is like the screwdrivers that I mentioned, namely that for Level 4 true self-driving cars there is a scope that an automaker or self-driving tech firm can impose upon the drivability of the AI.
One automaker might make a Level 4 self-driving car that works within downtown San Francisco and only functions in sunny weather. That’s the scope, similar to saying that you have a screwdriver that works on flat head screws, but not on other types of screws.
Thus, this particular Level 4 self-driving car won’t work when you decide to take it to say Los Angeles (which is outside the scope of San Francisco).
Or, it won’t work in San Francisco if it is a rainy day (since this is outside the scope of sunny weather).
It is “easier” to make a Level 4 than a Level 5, for the same reason that it is easier to make a series of differing screwdrivers for different situations than it is to have a universal one.
Nearly everyone in the Level 4 and Level 5 game is starting with developing Level 4, and then hoping to gradually expand and extend the Level 4 into becoming a Level 5.
Level 5 though is a tall order.
Humans are actually quite capable drivers overall, being able to drive a car in sunny weather, in rainy weather, in snow, and in cities, and in suburbs, and in many other circumstances.
The path to Level 5 seems to be a divide-and-conquer approach, let’s do narrowly scoped uses, and eventually turn this into the all-encompassing multi-use tool.
These scopes that I’m alluding to are formally called ODD (Operational Design Domain).
Here’s a quote from the standard that helps indicate the Level 4 versus Level 5 capabilities of the Automated Driving System (ADS):
- Level 4: “Permits engagement only within its ODD”
- Level 5: “Permits engagement of the ADS under all driver-manageable on-road conditions”
Here’s the formal definition of ODD:
- “Operating conditions under which a given driving automation system or feature thereof is specifically designed to function, including, but not limited to, environmental, geographical, and time-of-day restrictions, and/or the requisite presence or absence of certain traffic or roadway characteristics.”
These ODD’s can be whatever an automaker or self-driving tech firm decides it should be, such as saying that their self-driving car won’t work at nighttime and only during the day. Meanwhile, some other automaker offers a self-driving car that will work during nighttime and daytime, but maybe it won’t work in the rain.
It is confusing to the public and the media that these ODD’s aren’t defined in a standardized way.
In other words, an automaker or self-driving tech firm can decide to divine their own proprietary ODD’s.
This means that if you see a self-driving car driving past you, there’s no immediate way to know what its scope consists of.
Maybe it has been devised to go only in a 10-block radius of where you happen to see it, or maybe it can go across your state to other cities. Perhaps it can work in the rain, but not in snow. And so on.
For self-driving cars that are going to be used for ridesharing, you’ll have no direct means of knowing what the scope of that self-driving car might be.
Presumably, hopefully, at least whoever owns the self-driving car will make available the scope limitations, perhaps stating as such when you request a ridesharing pick-up involving a self-driving car.
- Level 4 is a self-driving car that has some kind of bounded scope of where and when it will drive
- The bounded scope is formally called its ODD (Operational Design Domain)
- There is no standardized set of ODD’s and thus they can vary by automaker or tech maker
- Level 5 has no bounded scope per se and thus essentially encompasses all feasible ODD’s
- Both Level 4 and Level 5 are allowed to be limited to what is humanly drivable
- All the levels of the standard, including Level 4 and Level 5, apply only to on-road driving and thus off-road driving is not a requirement by the standard
For more details about ODDs, see my indication at this link here: https://www.aitrends.com/ai-insider/amalgamating-of-operational-design-domains-odds-for-ai-self-driving-cars/
On the topic of off-road self-driving cars, here’s my details elicitation: https://www.aitrends.com/ai-insider/off-roading-as-a-challenging-use-case-for-ai-autonomous-cars/
I’ve urged that there must be a Chief Safety Officer at self-driving car makers, here’s the scoop: https://www.aitrends.com/ai-insider/chief-safety-officers-needed-in-ai-the-case-of-ai-self-driving-cars/
Expect that lawsuits are going to gradually become a significant part of the self-driving car industry, see my explanatory details here: https://aitrends.com/selfdrivingcars/self-driving-car-lawsuits-bonanza-ahead/
For those readers that know the SAE J3016 by heart, you’ll notice that I’ve tried to simplify the language used to describe the levels of automation, and as a result, the herein description is not quite as precise as the actual standard, though nonetheless provides a fair, readable, and accurate depiction.
Those of you that were paying close attention, you might have noticed that there is an important and somewhat hidden and unheralded phrase that came up in the levels of automation, namely the indication of off-road versus on-road.
Many don’t realize that the SAE J3016 considers the standard to apply to on-road driving and does not necessarily apply to off-road driving.
Here’s the definition of on-road:
- “On-road refers to publicly accessible roadways (including parking areas and private campuses that permit public access) that collectively serve users of vehicles of all classes and driving automation levels (including no driving automation), as well as motorcyclists, pedal cyclists, and pedestrians.”
Ponder this important nuance.
As mentioned, Level 4 and Level 5 involve driving acts that are human drivers feasible, but there is a caveat to that aspect.
The definition caveat is that it includes only on-road driving and excludes, therefore, off-road driving (though, presumably off-road driving is another one of those optional additions).
Human drivers can drive off-road, there’s no doubt about that.
The definition for Level 4 and Level 5 is only with respect to humans being able to drive on-road.
Few realize that this limitation or constraint exists in the standard.
If you have the time to do so, you ought to consider reading the entire SAE J3016 document, especially for those interested in self-driving cars. There are lots of twists and turns included, and I assure you it is as engaging as reading a mystery story or drama novella.
Self-driving cars are going to undeniably impact our society.
It is therefore crucial that we all be sufficiently informed, and make sure that when you refer to a rose, it is a rose, and not something else since our words do matter and what we mean by them is vital to discussion and understanding.
Copyright 2020 Dr. Lance Eliot
This content is originally posted on AI Trends.
[Ed. Note: For reader’s interested in Dr. Eliot’s ongoing business analyses about the advent of self-driving cars, see his online Forbes column: https://forbes.com/sites/lanceeliot/]