-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Tesla.com - "Transitioning to Tesla Vision"
- Thread starter run-the-joules
- Start date
Cheburashka
Active Member
If radar is so bad and they are only weeks away from a vision only system, what's the point of keeping radar on the MS and MX? Sounds like they'll have everything figured out by the time those land in customers hands.
Good on NHTSA to respond so quickly to these developments. From reading an article about it, it sounds like Tesla briefed them on their change to camera-only configurations and they've decided to revoke the check marks for those safety features for the new models for now.
Not sure about other customers, but I am not going to be taking delivery of a Model Y unless these boxes end up getting checked again to the satisfaction of the NHTSA. Literally, the primary feature about the Tesla that attracted me to it (beyond obviously the no-ICE aspect of it) was it's safety rating in terms of crash tests (which remains exceptional) and all the additional safety features that came along with it (and a subset of these are standard across all modern cars these days). I can wait till they prove out to the NHTSA that they can earn those check marks back. Maybe it will be trivial, maybe not.
And if they can't, then that will just prove this to be a crappy move for their MY and M3 customers. Makes me wonder if they kept the sensors for their European sales because they probably wouldn't even be able to get away with selling in Europe without the existing system due to regulations.
I was worried yesterday when I first heard this no-radar news, but have since done some reading about Elon Musk's "pure vision" philosophy and feel a little better. First off, while it stands to reason that a system with vision only (cameras) can never be as capable as that same system augmented with radar/lidar as well, this doesn't mean a vision-only system cannot be good enough. I think the real issue here is probably cost. Radar and Lidar are more expensive, and continuing to work on a sensor fusion approach divides efforts along two paths.
Personally I would prefer the radar stay because it's well understood and I know it can do some things very well that are currently difficult with vision, but I am starting to believe that there may be merit to the idea of abandoning radar in order to focus 100% on a single modality approach for the future. This may be especially true if the new design is scalable so that any number of inexpensive cameras can be added at other positions on (or off) the car and incorporated into the network and optimized using computer vision and machine learning algorithms. It's a bold move and definitely involves some risk, but it seems like the sort of think Elon Musk does pretty well.
So basically I think performance is going to suffer for the short term as Tesla's systems adjust to working without the radar input, but in the long term it could be a big win.
Other food for thought. What unexpected benefits might there be for us to have a car that now navigates using only passive sensor technology?
Personally I would prefer the radar stay because it's well understood and I know it can do some things very well that are currently difficult with vision, but I am starting to believe that there may be merit to the idea of abandoning radar in order to focus 100% on a single modality approach for the future. This may be especially true if the new design is scalable so that any number of inexpensive cameras can be added at other positions on (or off) the car and incorporated into the network and optimized using computer vision and machine learning algorithms. It's a bold move and definitely involves some risk, but it seems like the sort of think Elon Musk does pretty well.
So basically I think performance is going to suffer for the short term as Tesla's systems adjust to working without the radar input, but in the long term it could be a big win.
Other food for thought. What unexpected benefits might there be for us to have a car that now navigates using only passive sensor technology?
I'm the one that posted Tesla had removed AEB from their OWN WEBSITE on the day they also removed radar. A surprising bit of honesty from Tesla.
Then it suddenly got a lot of discussion on Reddit, and volia, Tesla added it back.
But then, when NHTSA notices this irregularity, they ask Tesla if the cars have AEB, and Tesla says NO.... And tesla is keeping radar on cars in countries that require AEB....
This is not "regulatory," Tesla themselves indicated the new cars don't have AEB on their own website. NHTSA picked up on this, asked Tesla if it has AEB, and updated their ratings. Then Tesla changed back again. So yeah, now Tesla has to re-prove it, because they brought this on themselves. And of course, it appears Tesla actually DOESN'T have AEB on these cars anymore, so this is exactly what should be happening. The only thing the regulator is doing is preventing Tesla from lying to customers about the features the car has.
The real unethical thing here now is Tesla listing AEB and other features on their site as standard features that they are telling regulators are not there.
Last edited:
There is literally no argument except cost that can ever be had for "why is it a good thing that they removed a sensor from my vehicle?" More data is always good if used correctly. If they really prove vision only is as good as vision+radar, they can always just stop processing radar. They cannot go the other way in a car with no radar.
Read the vision only stuff carefully- they never say vision only is better than vision+radar or vision+lidar. They just say it can be good enough.
That is what I am saying as well. It can never be better to have less information. But having more information (radar + cameras) is only useful if you know how to use this information effectively. Musk seems to be betting that in the future, a bunch of simple cameras and some really smart AI processing will be the most efficient way to get the job done.
Having owned multiple Subarus with Eyesight over the years I can confirm that the vision-only system works well, including lane-centering, ACC and AEB. Furthermore, all features work quite well whether it's city driving or highway driving which in the case of ACC in city situations at the minimum distance indicates that the system has a resolution far greater than the "few feet" suggested here.
To answer momo3605's earlier question, I have only very rarely had Eyesight disabled by adverse weather, be it direct sunlight, snow, or rain, but it does happen. Greenmore's point about Eyesight being a finely tuned stereo camera system is on-point. While possible, it will be difficult for Tesla to achieve this same level of performance from a vision-only system not designed for this purpose.
While the (allegedly temporary) removal of some FSD features is quite unfortunate and poor treatment by Tesla of their best (or let's be honest: most gullible) FSD-paying customers, the alleged (based on the NHSTA status) issues with core safety features such as AEB and the like is absolutely unacceptable especially as this is not spelled out in the disclaimer customers agree to when taking delivery. We'll have to see if this was rectified with the firmware delivered this weekend. If these key safety features are missing it could also affect insurance premiums. On the flip side, it doesn't make sense that something like lane departure warning which is very obviously a vision-based system would be affected by the removal of radar. I am going to take a stab in the dark and suggest that the NHTSA's change was related to new software that had not yet been fully tested and proven out. This would explain why the non-radar dependent systems were also marked as not standard.
As someone with a pending M3 order that just got delayed 2 more weeks, I'm a little glad it did as I'll hopefully have more time before having to make a decision on whether or not to accept the vehicle. Personally I'm fine without summon and other FSD toys (just my opinion, flame someone else or chock it up to I can't afford $10k for it) but not okay without the core safety features such as AEB, lane departure, etc..
To answer momo3605's earlier question, I have only very rarely had Eyesight disabled by adverse weather, be it direct sunlight, snow, or rain, but it does happen. Greenmore's point about Eyesight being a finely tuned stereo camera system is on-point. While possible, it will be difficult for Tesla to achieve this same level of performance from a vision-only system not designed for this purpose.
While the (allegedly temporary) removal of some FSD features is quite unfortunate and poor treatment by Tesla of their best (or let's be honest: most gullible) FSD-paying customers, the alleged (based on the NHSTA status) issues with core safety features such as AEB and the like is absolutely unacceptable especially as this is not spelled out in the disclaimer customers agree to when taking delivery. We'll have to see if this was rectified with the firmware delivered this weekend. If these key safety features are missing it could also affect insurance premiums. On the flip side, it doesn't make sense that something like lane departure warning which is very obviously a vision-based system would be affected by the removal of radar. I am going to take a stab in the dark and suggest that the NHTSA's change was related to new software that had not yet been fully tested and proven out. This would explain why the non-radar dependent systems were also marked as not standard.
As someone with a pending M3 order that just got delayed 2 more weeks, I'm a little glad it did as I'll hopefully have more time before having to make a decision on whether or not to accept the vehicle. Personally I'm fine without summon and other FSD toys (just my opinion, flame someone else or chock it up to I can't afford $10k for it) but not okay without the core safety features such as AEB, lane departure, etc..
Dogs_In_A_Pile
Member
With my insurance company it definitely will increase premiums which are already high. The loss of radar and move to cameras doesn't really bother me much. Increased insurance premiums bothers me a lot.
Like I've been saying here along with a few others, if Musk was really betting on this and believed that they are ready for this change, then the redesigned S and X would not include a radar. This isn't so much a proof of Tesla's conviction about their vision-only philosophy, but rather an experimentation on a large segment of their consumer base, born mostly out of supply chain issues and a desire to continue shipping vehicles to customers.
It's pretty clear that all the recent noise from Musk about the vision-only approach has been a planned PR move to mask the fact that they were going to have to gimp the M3 and MY, at least for the short-term, and potentially indefinitely just so they could continue selling their cars due to the parts shortage.
I have no issues with Tesla going down the vision-only path. But to me, this is transparently not a move made by Tesla in good-faith. If they were truly ready for this change they would have already proven to themselves that vision-only could handle everything just as well, if not better than the previous hybrid system. They would also then be prepared to seamlessly switch their software over to provide the exact same current feature-set rather than gimping it as they have at the moment.
Daniel in SD
(supervised)
Maybe lane departure warning and assistance are disabled when forward collision is possible. You might not want the car tugging at the wheel or beeping when you’re trying to avoid a collision.
Yep, "betting" is the right word.
Same guy that said 5 years ago "self driving was a solved problem". And when he was saying that, he put radar on the cars. And FSD was a year away.
And then 2 years ago said Radar was a fundamental importance. And FSD was a year away.
And now says it isn't, even though when he removed radar, he had to remove emergency braking. And FSD is a year away....
Maybe he's right and totally ignoring radar will get us to L3+ autonomy faster. But maybe it won't. The only thing you know is that if he's wrong, your car won't have radar to be enabled with a software update.
The question is, do you want to buy a car based on his bet? Lots of us have in the past (the bet on FSD existing at all in any suggested timeframe), and many of us feel like we lost that bet.
You're playing against the house here. If they don't need radar, they save money leaving it off. If they do need radar, well, can I interest you in a paid for upgrade, or a 2024 model year car?
Last edited:
That's an interesting and valid point. Though don't all those features turn off once you provide a manual input such as steering or braking? On a true but funny point: based on what I've seen in other cars the damn things will ABSOLUTELY beep at you for crossing a lane or similar, regardless of what peril you're trying to avoid. It's like the engineers were tasked with maximizing beep count rather than safety haha!
stopcrazypp
Well-Known Member
The only problem with this theory is that they did the same thing with AP2 and AP2.5 when there was no supply shortage. This is just how Tesla operates.
Agreed. I never meant to suggest this is proof of Tesla's conviction about Vision-Only. It is more likely proof of a supply chain issue, as you say, and they are making sure that their more expensive models (S and X) are not effected by the shortage. I am talking more about his vision for the future. I think that eventually Musk expects vision-only plus AI to be the best approach and I think the idea has merit, but it is not quite ready yet and the shortage of parts is forcing his hand early. For now he feels confident enough that the system won't be degraded so much that he has to stop selling cars, but definitely not confident enough to claim it's on par with Radar.
If you are a member of consumer reports, go to their website and write them to get some visibility on this issue. When CR brought up braking issues before, and gave the M3 a bad rating, Tesla was forced into doing something about it.
If you write CR, quote the NHTSA info. At a minimum, it will put pressure on them and bring the attention this needs to the public. Why would I want to buy a less safe car, even short term. Get it right, and then bring it back.
If you write CR, quote the NHTSA info. At a minimum, it will put pressure on them and bring the attention this needs to the public. Why would I want to buy a less safe car, even short term. Get it right, and then bring it back.
I won't deny that Tesla can make sudden decisions. But in this case, given some of the corroborating stories about BMW struggling with sourcing radar for their cars and even leaving it entirely out of some of their current cars, the part shortage theory does seem quite plausible as what precipitated this sudden change. I don't doubt that Tesla was planning to try and transition to vision-only. But I don't think they really wanted to be forced into the current timeframe with a severely restricted set of features for the new cars without radar.
Tesla changed radar suppliers between HW2 and 2.5. Are you sure that they hadn't run out of HW2 radars before the software was ready and had to drop HW2.5 radars in?
Like now, the made that change before they had the SW ready, and they didn't tell prospective buyers. That became a huge deal when Consumer Reports dropped their ratings and exposed it.
Similar threads
- Article
