And one proof is that the most capable autonomy solution available to the consumer today is implemented via a driver assist UI and is thus only "level 2", whereas you can put a radar and lane sensor into your junky 1990's sedan and be "level THREE (on certain blessed highways, unless the user needs to change lanes or there's any construction". Clearly one system is more valuable than the other, but not in a way captured by SAE.
It is not the job of the SAE to assign value. Their job is to define function and role. The consumer can look at the level and the ODD and figure out the value.
Tesla FSD is wildly more capable than any autonomy solution available to the consumer. Yet it remains "level 2" while silly traffic jam toys claim "level 3" based on minimized behavior and heavy geofencing. And that's dumb.
It's not dumb if you understand how the levels work. The SAE is very clear that L3 is not automatically better than L2. L3 is merely different than L2. They simply describe function and role and they define what is autonomous or not. L2 means the driver always has to pay attention and therefore it is not autonomous. L3 means that the driver sometimes does not have to pay attention and therefore it is limited autonomous. Tesla's "FSD" being L2 does not imply that it is less capable than L3, it simply means that it requires constant driver attention and therefore it is not autonomous. A system can be very capable and not be autonomous. There is nothing wrong with being L2. Again, L2 is not necessarily inferior to L3, it's just different than L3.
Your response, it seems, is again a purely semantic one. You take the SAE definition of "autonomy" as canon and then define SAE's categorization as correct since it meets its own defintion. And... that's dumb. Because it fails to capture value as perceived by actual consumers of actual products.
Their definition is canon. They are the society of automotive engineers. So they get to define autonomy from a technical, engineering point of view. You are confusing two different issues. One issue is the technical definition. The other issue is consumer value. Obviously, consumers get to assign value as they see fit but that is different from the technical definition. The SAE levels merely define the type of autonomous driving. It is up to consumers to look at the level + ODD and decide the value.
Problem is since ODD is not standardized - one OEM can claim L3 (but on tiny, tiny ODD) and other with very broad ODD could be L2. Consumers would think - looking just at the levels that the first OEM has a more useful system than the second OEM. That may not be the case.
We had the same kind of argument a year or two back.
As they say, it feels like déjà vu all over again.
Yes, it feels like deja vu. But this is a red herring. Consumers are not going to look at just the level, they are also going to look at the ODD.
Some people seem to want the SAE levels to be something they are not intended to be. They want the levels to be some type of ranking where consumers can just look at the level and know what system is better. They don't seem to understand that you are supposed to give the level and the ODD as a combo to describe a system. The level is only one piece. It is not meant to describe the entire autonomous driving system.
ps : I can understand why they didn't specify ODD. They probably didn't get consensus. May be its a good time for them to revisit this and define standard ODDs to help consumers easily compare among various vendors.
Defining standard ODDs is a bad idea because ODDs involve too many different factors, from geofencing, to road types, to weather conditions, to traffic speeds etc... So it is not possible to cover every possible type of ODD. And if you try to cover the main ones, you could miss one. What if you create a few standard ODDs and a manufacturer decides to deploy an autonomous driving system with an ODD you forgot? It is better to leave ODDs out of the levels so that the levels can cover different autonomous driving, regardless of what ODD a manufacturer chooses.
SAE J3016, page 32:
Conceptually, the role of a driving automation system vis-à-vis a user in performance of part or all of the DDT is orthogonal to the specific conditions under which it performs that role. For example, a specific implementation of adaptive cruise control may be intended to operate only at high speeds, only at low speeds, or at all speeds. For simplicity, however, this taxonomy collapses these two axes into a single set of levels of driving automation. Levels 1 through 4 expressly contemplate ODD limitations. In contrast, Level 5 (like Level 0) does not have ODD limitations (subject to the discussion in 8.8).Accordingly, accurately describing a feature (other than at Levels 0 and 5) requires identifying both its level of driving automation and its operational design domain (ODD). As provided in the definitions above, this combination of level of driving automation and ODD is called a usage specification, and a given feature satisfies a given usage specification.
Basically, Tesla fans are confusing levels with usage specification. They want the levels to be the usage specification when they are not. Level + ODD = usage specification.
