Interesting new way of classifying autonomous driving

[diplomat33]

I found this blog from Dr. Philip Koopman, Edge Case Research & Carnegie Mellon University, where he proposes a new classification system for autonomous driving:

A Driver-Centric User’s Guide to Vehicle Automation Modes

Vehicle Automation Modes emphasize the responsibilities of a self-driving vehicle user By: Dr. Philip Koopman , Carnegie Mellon University V...

safeautonomy.blogspot.com

It does not replace the SAE levels but he feels it might be more helpful and less open to abuse or misunderstanding.

He proposes 4 "operational modes":

Assistive: A licensed human driver drives, and the vehicle assists.
This might include anti-lock brakes, stability control, cruise control, and automatic emergency braking. The driver always remains in the loop, exerting at least some form of continuous control over speed, lane keeping, or both. Most passenger vehicles on the road today are Assistive.

Driver liability: As with conventional human driving.

Supervised: The vehicle drives, but a human driver is responsible for ensuring safety.
Technology normally handles all aspects of the driving task. However, a licensed human driver is responsible for continuous monitoring of driving safety and taking over control instantly if something goes wrong. The driver is not expected to perform a continuous control function such as steering or speed control while in this operating mode. An effective driver monitoring system is required to ensure driver ability to take over when required.

Tesla's AP and GM's SuperCruise would fall under this category.

Driver liability: The human driver is responsible for safety unless the vehicle does something dangerous that is beyond a reasonable human driver capacity to intervene.

Automated: The vehicle performs the complete driving task.
A human driver is not required to operate the vehicle in this mode. However, a responsible person is required to ensure other aspects of vehicle safety such as buckling up the kids, proper securing of any cargo, and post-crash response. Simply put, in this operation mode the vehicle does the driving, but a responsible human is still the “captain of the ship” for handling everything except the driving. In some cases, there might be an expectation that a human driver moves the vehicle under manual control during portions of a trip that are not suitable for Automated operation.

Driver liability: The human driver is not responsible for driving errors, but is responsible for non-driving aspects of safety such as passenger safety, proper cargo loading, and post-crash situation management.

Autonomous: The whole vehicle is completely capable of operation with no human monitoring.
The vehicle can complete an entire driving mission under normal circumstances without human supervision. If something goes wrong, the vehicle is entirely responsible for alerting humans that it needs assistance, and for operating safely until that assistance is available. Things that might go wrong include not only encountering unforeseen situations and technology failures, but also flat tires, a battery fire, being hit by another vehicle, or all of these things at once. People in the vehicle, if there are any, might not be licensed drivers, and might not be capable of assuming the role of “captain of the ship.”

Driver liability: There is no human driver to blame for mistakes.

Here is a handy summary chart:

I kinda like this classification system. Thoughts?

 

[Dan D.]

I like it. I can imagine a car having several "modes" and you set one based on your mood, subject to an electronic acceptance of your driver's license.

Something is needed to break the disconnect between current/future ADAS/AP systems and the drivers who think their cars allow them to stop paying attention. The current SAE & industry practice is not giving me much reassurance that it is being taken very seriously, except the recent NHTSA reporting rules.

 

[diplomat33]

I can imagine a car having several "modes" and you set one based on your mood, subject to an electronic acceptance of your driver's license.

Yes. The blog mentions that cars can have several modes:

A single vehicle can employ various operational modes across its Operational Design Domain (ODD). What is most important is that at any particular time the vehicle and the driver both understand that the vehicle is in exactly one of the four operational modes so that the driver’s responsibilities remain clear. As a simplified example, the same car might operate as Autonomous in a specially equipped parking garage, Automated on limited access highways, Supervised on designated main roads, and Assistive at other times. In such a car it would be important to ensure that the human driver is aware of and capable of performing accompanying driver responsibilities when modes change.

 

[ATPMSD]

Not a fan of the term ”Other Safety.” I would prefer “Final Safety” or “Backup Safety”

 

[diplomat33]

Not a fan of the term ”Other Safety.” I would prefer “Final Safety” or “Backup Safety”

Yeah. "Other Safety" is the one aspect of the classification system that I am not sure I completely agree with.

Specifically, the blog says this:

Even more importantly, the SAE Levels say nothing about all the safety relevant tasks that a human driver does beyond actual driving. For example, someone has to make sure that the kids are buckled into their car seats. To actually deploy such vehicles, we need to cover the whole picture, in which driving is critical but only a piece of the safety puzzle.

The SAE levels only cover "actual driving" because that is all that an autonomous vehicle should really be responsible for. Buckling your kids in their cars seats is critical but it has nothing to do with autonomous driving. That is why the SAE does not cover "other safety" that is not related to actual driving.

 

[voyager]

Hm, isn't any classification a bit of a semantics discussion?
I still don't dislike SAE's Level 1-5 classification for ADS.

What I would like, is the acceptance of driving circumstances. I am not talking
weather conditions, as this is incorporated in the SAE classification,
but I am talking traffic complexity. Driving downtown SF is much more complicated
than cruising over B roads outside the city.

 

[diplomat33]

Hm, isn't any classification a bit of a semantics discussion?
I still don't dislike SAE's Level 1-5 classification for ADS.

Probably the simplest definition in my opinion of autonomous driving is "a computer driving a car instead of a human". But you still have questions like "what is driving?" and how do define all the different ways that a computer can drive a car. You could have a system where the computer drives the car but with a human as a fallback if something goes wrong, or a computer could drive a car but only on certain roads or in certain conditions, or you could have a computer that can drive a car in all conditions and never asks the human for help. How do you classify all those different variations of "a computer is driving a car". And if the computer is steering, braking and following a route, does that count as "driving" if it can't react to certain situations? Also, how good or how safe does the computer driving have to be accepted as the "computer driving a car"? So there are a lot of questions that have to be answered.

Having studied J3016 in depth, I think the SAE levels make a lot of sense. J3016 certainly does a good job of defining the key terms related to autonomous driving and offers different classes of autonomous driving based on the role of the driver.

I would add that the SAE levels are not really levels. They are more like classes of autonomous driving.

What I would like, is the acceptance of driving circumstances. I am not talking
weather conditions, as this is incorporated in the SAE classification,
but I am talking traffic complexity. Driving downtown SF is much more complicated
than cruising over B roads outside the city.

Well, I think that is part of the challenge of developing autonomous driving. Depending on where and when you want to use your autonomous driving, driving might be easier or more difficult, and present different driving situations to solve. That is where the ODD comes in. Certainly, if you deploy say driverless L4 in downtown NYC that is more impressive than deploying driverless L4 trucks that only do a single route at night on empty highways.

 

[voyager]

IMO we are sooooooo far away from the only phase that's truly interesting, and that is SAE Level 5 autonomy,
that we might conclude that self-driving should not be used within the built environment and only outside the city.

We may then be able to skip a lot of frustrating experimental phases in which many EV developers are trying to
find the holy grail, which may take years, if ever possible.

What strikes me about ALL threads regarding vehicle autonomy, is the focus on technology, optics and AI.
What sets a human driver (not under the influence of anything) apart from any AV system or a chimpansee is:
situational awareness and the split-sec ability to make conscious decisions.

 

[Daniel in SD]

IMO we are sooooooo far away from the only phase that's truly interesting, and that is SAE Level 5 autonomy,
that we might conclude that self-driving should not be used within the built environment and only outside the city.

We may then be able to skip a lot of frustrating experimental phases in which many EV developers are trying to
find the holy grail, which may take years, if ever possible.

What strikes me about ALL threads regarding vehicle autonomy, is the focus on technology, optics and AI.
What sets a human driver (not under the influence of anything) apart from any AV system or a chimpansee is:
situational awareness and the split-sec ability to make conscious decisions.

A robotaxi that work on 99% of paved roads within a city isn't interesting?
"eyes off" automated driving that works in good weather on all controlled access highways in the US wouldn't be interesting?
There are a lot of interesting applications that will be possible before AVs can drive everywhere and under every condition that a human can.

This classification system seems almost the same as the SAE categories.

 

[cucubits]

I don't think Tesla will like this chart too much. If we go by the classification Tesla should be somewhere between 1 and 2. The car does indeed do the driving part but the driver still needs to have the hands on the wheel, not just eyes on the road.

 

[diplomat33]

IMO we are sooooooo far away from the only phase that's truly interesting, and that is SAE Level 5 autonomy,

I think it would be more precise to say that we are far away from L5 that is safe to deploy. Remember the SAE does not define safety, it just defines capability. We certainly could deploy L5 now, it would just be uneven, handling some driving great but struggle, behave badly, or get into accidents in other situations. Basically, the safety would be ok in some areas but not ok in other areas. Thus, for now, companies are geofencing their autonomous driving to where they are confident it is safe enough.

 

[EVNow]

Just more of the same. Theoretical and useless at present.

ODD - unless that is defined, it’s simply not useful.

 

[Daniel in SD]

Just more of the same. Theoretical and useless at present.

ODD - unless that is defined, it’s simply not useful.

I think it's safe to assume that any "automated" or "autonomous" vehicle under this classification will have a defined ODD.

 

[StealthLSU]

I don't think any classification will work the way it is currently being developed.

Cars like whichever one in Germany can do such a limited situation on the autobahn but is level 3 is weird to call it level 3. Even with this new system, you can't call it one of these classifications because it just depends on the situation.

And even the original is flawed since you could theoretically geofence a one block area and make it level 4 autonomous. But it would not really help anyone and doesn't fit the spirit of what a level 4 car is.

Best system is probably just giving the full description of what it does and not worry about what level it falls into.

 

[Daniel in SD]

I don't think any classification will work the way it is currently being developed.

Cars like whichever one in Germany can do such a limited situation on the autobahn but is level 3 is weird to call it level 3. Even with this new system, you can't call it one of these classifications because it just depends on the situation.

And even the original is flawed since you could theoretically geofence a one block area and make it level 4 autonomous. But it would not really help anyone and doesn't fit the spirit of what a level 4 car is.

Best system is probably just giving the full description of what it does and not worry about what level it falls into.

That's the Operational Design Domain of the system. Neither this nor the SAE taxonomy are meant to classify vehicles, they're meant to classify systems. The Mercedes you're talking about is also Level 2 because it has a "supervised" automation mode as well.
Tesla is planning to release FSD later this year as an L4 system but it won't work everywhere so you'll still have an L2 system that you can use outside the L4 system's ODD.

 

[cucubits]

Tesla is planning to release FSD later this year

 

[diplomat33]

Here is an article on a new classification for autonomous driving:

What Are the Levels of Automated Driving? Think of Your Hands, Feet, Eyes and Brain

I’m going to dive a little deeper however and explore the relationship between our own senses and vehicle automation.

groundtruthautonomy.com

Basically the system would be based on whether the driver needs to do braking and/or steering, keep their eyes on the road or keep their brain on the road:

• Feet-on/hands-on/eyes-on/brain-on
• Feet-OFF/hands-on/eyes-on/brain-on
• Feet-OFF/hands-OFF/eyes-on/brain-on
• Feet-OFF/hands-OFF/eyes-OFF/brain-on
• Feet-OFF/hands-OFF/eyes-OFF/brain-OFF

What is interesting is that this classification system mirrors what some people think the SAE Levels are. We see diagrams that interpret L2 as "hands-on" and L3 as "eyes off" or L4 as "mind off". The problem is that there is some nuance there because some companies are promoting "hands-off L2". So we cannot say that all L2 is "hands-on". This system would be more deliberate and clearer, I think.

This system would probably be clearer for the public since, I think, everyone understands if a system requires your hands on the wheel or not or requires your eyes on the road or not.

 

[2101Guy]

which one of those has the CEO of Tesla promised with the current hardware in place, and by when?

 

[Daniel in SD]

Seems fine but like you said these are basically the SAE levels. I would classify Autopilot as "hands-off" too since it does not allow any steering input while it is engaged.