Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Tesla Vision NOA aces ADAS test vs Mobileye Supervision (Zeekr 001) and others

This site may earn commission on affiliate links.
I did a search and didn't see this posted elsewhere, but I stumbled on a fairly interesting set of tests done on the ADAS systems in the leading EVs in China in November 2022. They did 10 different scenarios (7 daylight, 3 at night) including a bunch of tests that typical standardized testing don't get to test (only made possible because they rented out a national test center that has a full scale model of public roads). Video has English subtitles so is fairly easy to follow.

Some background: Tesla is a 2022 Model 3 equipped with the FSD package, but video mentions in China it does not respond to traffic lights even though it recognizes them (so effectively it is just the same as EAP/NOA). It has no radar (AKA it is running Tesla Vision), given that was removed in around November 2021 (video mentions this too versus other cars that have radar):
Giga Shanghai follows Fremont's lead and ditches radar

For all the flack Tesla gets for phantom braking in the forums, removing radar in Tesla Vision, and all the FUD about FSD Beta running over child dummies, you would think they have the least capable, most unsafe, ADAS system. These tests show how the Tesla system is tuned to be safer in more scenarios (at the cost of false positives).

Test results (out of 10 tests):
Tesla Model 3: 10 good
Zeekr 001: 4 good, 2 average
Changan SL03: 2 good, 1 average
BYD Han EV: 1 good
BYD Seal (Atto 4): 2 average

LightingScenarioBYD Han EVBYD Seal (Atto 4)Changan SL03Tesla Model 3Zeekr 001
DayFollow car at red lightPoorPoorAverageGoodAverage
DayCross intersection with no carPoorPoorPoorGoodPoor
DayCross intersection with carPoorAveragePoorGoodPoor
DayMerging Car in motionGoodAverageGoodGoodGood
DayPedestrians running red lightPoorPoorPoorGoodAverage
DayMerging car stopped on sidePoorPoorGoodGoodGood
DayTraffic ConesPoorPoorPoorGoodPoor
NightLeft turn into crossing pedestrianPoorPoorPoorGoodPoor
NightCrossing BicyclePoorPoorPoorGoodGood
NightDisabled vehicle in fog, opposing headlightsPoorPoorPoorGoodGood

Unfortunately they didn't give details on the sensor mix of each one, but they did mention none of them are equipped with Lidar, but rather a mix of cameras, radar, and ultrasonic sensors. From my search the SL03 was promoted to have Lidar, but it might not be equipped in this particular car (or didn't end up in production).

SL03 uses a Qualcomm platform for its ADAS, similar to what GM is looking to move to for Supercruise.

They didn't test the Nio ET5 (which does have Lidar) because it was a preproduction car and didn't have the software ready yet.

The Tesla passed with flying colors getting a good rating on 10/10 tests, with others not even coming close. Zeeker 001 with the Mobileye SuperVision came a distant second, although in 2 tests reviewers note it did better than the Tesla: in the night bike crossing test it was able to perform an "Antelope Avoidance" maneuver instead of just braking, and in the disabled smoking car test it was able to turn on the fog lights automatically as it approached (an improvement after the latest software update, where in a previous test it did poorly responding to a vehicle in fog).

 
Last edited:
it has been posted. This is just one test by a tester out of a bunch of other tests by other testers.

What you link is a completely different test, looks like the standard Euro AEB test done with AP/NOA disabled.
AEB Pedestrian | Euro NCAP

The above tests instead are testing with NOA enabled vs other ADAS systems, not testing the AEB (which in Teslas err against false positives, so would rather not brake/slow down, the exact opposite of how things operate in AP).
 
  • Like
Reactions: GSP
What you link is a completely different test, looks like the standard Euro AEB test done with AP/NOA disabled.
AEB Pedestrian | Euro NCAP

The above tests instead are testing with NOA enabled vs other ADAS systems, not testing the AEB (which in Teslas err against false positives, so would rather not brake/slow down, the exact opposite of how things operate in AP).
If that is your criteria. There are abunch of other tests where Tesla doesn't do aswell as others in. I have watched a bunch of china tests. I just have to remember the video and post it later.

Basically This test is only popular because Tesla came up on top. And this test is missing other cars that WOULD beat Tesla. For example Huawei was doing a test where there's a stopped car in tunnel entrance and Tesla on AP/NOA tried to ram the car every time.

TLDR: There's a bunch of ADAS tests in China, car reviews is a popular thing. Tesla rarely comes out on top.

EDIT: Some of the test in the video you posted are AEB tests.
 
The above tests instead are testing with NOA enabled vs other ADAS systems, not testing the AEB (which in Teslas err against false positives, so would rather not brake/slow down, the exact opposite of how things operate in AP).
Test (Stalled Car in Tunnel Entrance and Cone in the road)
Model Y Tested failed multiple tests.
3:58-5:50

There are a bunch of tests with different conditions and scenarios.
I can find them if you are really interested.
 
Last edited:
If that is your criteria. There are abunch of other tests where Tesla doesn't do aswell as others in. I have watched a bunch of china tests. I just have to remember the video and post it later.

Basically This test is only popular because Tesla came up on top. And this test is missing other cars that WOULD beat Tesla. For example Huawei was doing a test where there's a stopped car in tunnel entrance and Tesla on AP/NOA tried to ram the car every time.

TLDR: There's a bunch of ADAS tests in China, car reviews is a popular thing. Tesla rarely comes out on top.
Seems like you are just trying to nitpick individual tests that show Tesla doing worse. All I'm doing is posting a set of comprehensive tests done with NOA/Autosteer enabled (which is not that common). You haven't linked that this set was ever posted in this forum.
EDIT: Some of the test in the video you posted are AEB tests.
No it's not, all of them were done with Autosteer enabled (you can see in the pedestrian braking tests the blue lane lines from Autosteer). In some of the other cars, the AEB had to kick in to stop the car, but the cars were all in its respective ACC/lane keeping modes for the tests as they approached the targets (although they did not show the camera views in all of them leading up to it).
pedestrian_test.jpg

bike_test.jpg


Left turn test was only exception I can see that didn't have Autosteer on for obvious reasons (Autosteer can't make left turns and if you take over steering it falls back to TACC), but TACC was on (see the 30 max symbol next to speed limit). The words on the right side "高阶驾驶辅助开启中" appear to translate to "Advanced driving assistance is on" and they show it in every test, even if they don't show the in car camera view every time.
left_turn_test.jpg
 
Last edited:
NOA would have a teal path predictor from the front hood of the Tesla car icon.

Any mode missing that teal path predictor means it's not NOA.
Thanks for the correction. I should have said Autosteer instead of NOA. I forgot the type of road they were testing would have made it so NOA would not be operating. Basically my main point was that some form of AP was on versus other tests where it's purely AEB and no ADAS was on (sometimes the tester is even applying accelerator).
 
  • Like
Reactions: Tam
As evidence that Tesla AEB errs against false positive (especially when driver is applying accelerator), I found this set of tests on the latest Model Y (without ultrasonics) just done 3 weeks ago in Taiwan. Luckily the video has subtitles and autotranslate should make it possible to follow along:

Basically it tests AEB vs TACC against a 130cm (4'8") tall dummy between two orange cones in rain. Here are test results:
1) 30 km/h accelerator applied (FCW Medium): fail to detect and no action at all
2) 30 km/h accelerator applied (FCW Early): detected last minute and AEB braked but still hit
3) 40 km/h accelerator applied: detected last minute and AEB braked but still hit
4) 30 km/h accelerator released to coast: detected last minute and AEB braked but still hit
5) 30 km/h accelerator released to coast: detected last minute and AEB braked in time
6) 30 km/h TACC: detected and braked gradually stopping far from dummy
7) 40 km/h TACC: detected and braked gradually stopping far from dummy
8) 50 km/h TACC: detected and braked (AEB activated) but still hit
9) 50 km/h TACC: detected and braked gradually stopping far from dummy
10) 60 km/h TACC: detected and braked gradually stopping far from dummy

They noted it was very hard to coast the vehicle due to the regen braking on accelerator pedal, so that part might not be that consistent versus other cars they tested in the past.
 
Last edited:
  • Informative
Reactions: emmz0r
Zeeker 001 with the Mobileye SuperVision came a distant second

I could be wrong but I don't think that was SuperVision in those tests. AFAIK, SuperVision is only activated on highways right now and only in "lead customers". SuperVision has not been released for city driving yet. So I doubt they were using SuperVision in those tests. It is more likely using Zeekr's base ADAS or AEB systems.
 
Seems like you are just trying to nitpick individual tests that show Tesla doing worse. All I'm doing is posting a set of comprehensive tests done with NOA/Autosteer enabled (which is not that common). You haven't linked that this set was ever posted in this forum.
No you are the one nitpicking. There are dozens of these tests and they are done all the time.
you call this one "comprehensive" because it fits your bias.

This was already posted last week by @EVNow

Lastly I posted another "comprehensive" test which Tesla didn't come out on top which you COMPLETELY IGNORED. It was predictable. I can post 10 more similar tests and you will completely ignore them also. Because you already have your conclusions. This is why the "test" went viral. A China Tesla fan finally found a test out of dozens (probably up to 100 at this point) where Tesla performed best at and wrote a internet article about it.
No it's not, all of them were done with Autosteer enabled (you can see in the pedestrian braking tests the blue lane lines from Autosteer). In some of the other cars, the AEB had to kick in to stop the car, but the cars were all in its respective ACC/lane keeping modes for the tests as they approached the targets (although they did not show the camera views in all of them leading up to it).
The event they were testing weren't all autosteer related, which is why i said some were AEB related. Nevertheless it doesn't matter, there are dozens of these tests and you like other Tesla fan ignore the rest and found one where Tesla came up on top and are running with it.
Left turn test was only exception I can see that didn't have Autosteer on for obvious reasons (Autosteer can't make left turns and if you take over steering it falls back to TACC), but TACC was on (see the 30 max symbol next to speed limit). The words on the right side "高阶驾驶辅助开启中" appear to translate to "Advanced driving assistance is on" and they show it in every test, even if they don't show the in car camera view every time.
But It wasn't autosteer related. It was AEB related and was AEB that kicked in. Anyway it doesn't matter. None of this matters which is the point. You are putting this test on a pedestal when there are dozens of these tests which you ignore.
As evidence that Tesla AEB errs against false positive (especially when driver is applying accelerator), I found this set of tests on the latest Model Y (without ultrasonics) just done 3 weeks ago in Taiwan. Luckily the video has subtitles and autotranslate should make it possible to follow along:

Basically it tests AEB vs TACC against a 130cm (4'8") tall dummy between two orange cones in rain. Here are test results:
1) 30 km/h accelerator applied (FCW Medium): fail to detect and no action at all
2) 30 km/h accelerator applied (FCW Early): detected last minute and AEB braked but still hit
3) 40 km/h accelerator applied: detected last minute and AEB braked but still hit
4) 30 km/h accelerator released to coast: detected last minute and AEB braked but still hit
5) 30 km/h accelerator released to coast: detected last minute and AEB braked in time
6) 30 km/h TACC: detected and braked gradually stopping far from dummy
7) 40 km/h TACC: detected and braked gradually stopping far from dummy
8) 50 km/h TACC: detected and braked (AEB activated) but still hit
9) 50 km/h TACC: detected and braked gradually stopping far from dummy
10) 60 km/h TACC: detected and braked gradually stopping far from dummy

They noted it was very hard to coast the vehicle due to the regen braking on accelerator pedal, so that part might not be that consistent versus other cars they tested in the past.
And Mobileye's SuperVision AEB and Huawei ADS AEB would ace all of these tests just like in the video i showed you which you conveniently ignored.

You actually think this is some flex when its a HUGE flaw of Tesla's active safety system. You are literally grasping at straws at this point. Maybe I should gather all the dozens of ADAS tests just to see the excuses you come up with for each one.

I already posted two which you ignored.
 
Last edited:
  • Like
Reactions: ilikewhey
I could be wrong but I don't think that was SuperVision in those tests. AFAIK, SuperVision is only activated on highways right now and only in "lead customers". SuperVision has not been released for city driving yet. So I doubt they were using SuperVision in those tests. It is more likely using Zeekr's base ADAS or AEB systems.
Yeah its not the SuperVsion derived from Mobileye's L4 development, but its still an offering under the SuperVision full package. Its the base ADAS/AEB system that Mobileye created in 2021/2022 using EyeQ4-L/M equivalent perception output from their EyeQ5. Its a separate system so they can sell highway assist and base active safety system by itself (using an EyeQ4) without the more expensive full supervision end to end suite.
 
I could be wrong but I don't think that was SuperVision in those tests. AFAIK, SuperVision is only activated on highways right now and only in "lead customers". SuperVision has not been released for city driving yet. So I doubt they were using SuperVision in those tests. It is more likely using Zeekr's base ADAS or AEB systems.
As mentioned in the other thread, SuperVision is an ADAS package that performs all the functions (even the parking assist) it does not refer solely to highway or city activation. There is no separate base ADAS. You can see it in their page (which even has the Zeekr 001 as the demo vehicle):

The upcoming update is the equivalent to FSD Beta, which as above does not exist in China (they don't even have traffic light control yet). It would be interesting to see a retest perhaps a year later when the software has been updated (and other cars with lidar can join in).

At any rate, as I noted, all the cars tested were specifically noted that they had their ADAS enabled (including lane keeping where applicable).
 
No you are the one nitpicking. There are dozens of these tests and they are done all the time.
you call this one "comprehensive" because it fits your bias.
I said it's comprehensive because it compares bunch of different cars in different scenarios and conditions.
This was already posted last week by @EVNow
I didn't find it in my search, did you have a link?
Lastly I posted another "comprehensive" test which Tesla didn't come out on top which you COMPLETELY IGNORED.
No you didn't, that was just a standard AEB test, where Tesla successfully stopped at some speeds and didn't in others.

It was predictable. I can post 10 more similar tests and you will completely ignore them also. Because you already have your conclusions. This is why the "test" went viral. A China Tesla fan finally found a test out of dozens (probably up to 100 at this point) where Tesla performed best at and wrote a internet article about it.
No it's not, it was from a series of tests of 6 EVs that included range tests and the Tesla only came at a tied second overall. They were not cherry picking a set of tests that were designed to favor the Tesla.
The event they were testing weren't all autosteer related, which is why i said some were AEB related. Nevertheless it doesn't matter, there are dozens of these tests and you like other Tesla fan ignore the rest and found one where Tesla came up on top and are running with it.

But It wasn't autosteer related. It was AEB related and was AEB that kicked in. Anyway it doesn't matter. None of this matters which is the point. You are putting this test on a pedestal when there are dozens of these tests which you ignore.
It does matter because how Teslas respond in AEB is completely different from how they respond in TACC or Autosteer (as noted in the Taiwan test posted). This explains all the phantom braking in AP, while it practically never happens while car is driven with AP disabled.
And Mobileye's SuperVision AEB and Huawei ADS AEB would ace all of these tests just like in the video i showed you which you conveniently ignored.
I watched the first video, it was just a standard AEB test with AP disabled. The second link took forever to load on mobile, so I didn't bother watching it previously, but I just did now. The Tesla definitely failed on the given tests, but from my understanding, it was going significantly faster than the other vehicles due to the Tesla not slowing down for the curves (something I notice too from daily use, in a lot of curves it goes in way too fast if I don't lower my set speed ahead of time).
You actually think this is some flex when its a HUGE flaw of Tesla's active safety system. You are literally grasping at straws at this point. Maybe I should gather all the dozens of ADAS tests just to see the excuses you come up with for each one.

I already posted two which you ignored.
 
As mentioned in the other thread, SuperVision is an ADAS package that performs all the functions (even the parking assist) it does not refer solely to highway or city activation. There is no separate base ADAS. You can see it in their page (which even has the Zeekr 001 as the demo vehicle):

The upcoming update is the equivalent to FSD Beta, which as above does not exist in China (they don't even have traffic light control yet). It would be interesting to see a retest perhaps a year later when the software has been updated (and other cars with lidar can join in).

At any rate, as I noted, all the cars tested were specifically noted that they had their ADAS enabled (including lane keeping where applicable).

Yes it is part of the SuperVision package but it does not include REM or RSS. The tests were only with the base ADAS, not the advanced ADAS with REM and RSS. That is why I say it is misleading. You say it is a test of SuperVision but it was a only a test of the basic ADAS part of SuperVision, not the full SuperVision suite.
 
Yes it is part of the SuperVision package but it does not include REM or RSS. The tests were only with the base ADAS, not the advanced ADAS with REM and RSS. That is why I say it is misleading. You say it is a test of SuperVision but it was a only a test of the basic ADAS part of SuperVision, not the full SuperVision suite.
Under that criteria, any test that is done without FSD Beta would be considered misleading as it wouldn't be the "full FSD suite". I don't think many would buy that argument.

What is being tested is equivalent to Tesla's EAP/NOA package, which is the max of what is enabled in China (as others point out, actually for this particular road it was just on normal Autosteer, not even NOA).
 
Under that criteria, any test that is done without FSD Beta would be considered misleading as it wouldn't be the "full FSD suite". I don't think many would buy that argument.

It is not misleading if you clearly state what is being tested. If I say I am testing Tesla's autosteer and I say that, that's not misleading. But if I say I am testing Tesla FSD when I am only testing Tesla autosteer, that is misleading. If you had stated they were testing normal autosteer and AEB, I would have had no issue with your statement. But you stated it was testing SuperVision when it was testing a small subset of SuperVision. That's misleading.

What is being tested is equivalent to Tesla's NOA package, which is the max of what is enabled in China (as others point out, actually for this particular road it was just on normal Autosteer, not even NOA).

No it is not equivalent to Tesla's NOA. Yes, it was basic autosteer that was being tested.
 
Please clarify --- phantom braking using TACC only is similar to AP on or off ?
I count both under the same thing, although Autosteer was not specifically tested in that Taiwan test. Plenty of phantom braking happens under TACC also (that's why people keep asking for a dumb cruise control option). But you very rarely if ever hear of a cases just driving with both disabled, the AEB system almost never activates, and even if it does, it brakes much less than other systems.

I remember when it first started, AEB only slowed a max of 25mph, it would still crash into whatever obstacle ahead (although this had gradually changed throughout the years, perhaps because various AEB test standards expect the car to stop completely). I wouldn't be surprised if the AEB system is still tuned to just pass all the various standard AEB tests and not much beyond that. Tesla probably doesn't want to be subject to a bunch of those AEB recalls other manufacturers have been (although they had done that in that one FSD update back in 2021).
 
  • Helpful
Reactions: SageBrush
Anyone existing on either extreme end of the spectrum is of course doing themselves a disservice. The real world is full of nuance, and lots of ADAS nuance is evidenced in this thread.

My own opinion is that phantom braking is a result of heightened sensitivity in the OEDR and so you’d expect the system to perform really well in tests like these. Tesla could ramp
down the sensitivity and reduce phantom braking but it would likely impact the results in safety tests, heck there’s probably a simple slider for this in the dev tools.

The next question would be where the value line is drawn insofar that phantom braking can discourage people from using these systems in the first place and thus negate any benefits. I think it’s a balancing act between heightened sensitivity improving safety but also generating more false positives.

That’s how I imagine it works anyways