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

Tesla.com - "Transitioning to Tesla Vision"

This site may earn commission on affiliate links.
Also, can you link to any discussion where others have experienced this? This is the first I’ve heard of the headlights for AP issue. I sure hope you have a defective car and it can be fixed.
If I’m understanding your question correctly the two things I was mainly referring to with others having problems with AP was this post.
This was actually kind of the catalyst of my post as that was a reaction from one new to Tesla and why I worded the post as one from someone not new to Tesla.

The second and more directly related to the high beams were comments on my thread such as this.

 
Sorry no, I'm referring to this thread where OP who has both says that vision is objectively worse:

Pretty sure crossing into oncoming traffic due to not seeing lane lines has nothing to do with radar. That is objective proof of baloney. Rain washing out the vision system is no different to the with radar cars. Auto high beams as a requirement is interesting and hopefully they can make that smoother but it doesn't reduce functions. No way on earth tesla pulls this move if they are not 100% confident that they'll figure this out very quickly. Let's see how the aeb and other safety systems do in independent tests (iihs et al) THEN you can start up the sirens...for 2 weeks...until they improve it again...

Note, I'm not saying radar is bad. It just really doesnt change anything once you've got vision far enough along. The absolute worst plausible consequence would be permanently enforced longer following distances. Everything else vision can do what radar does (i.e., other than see through stuff), when sufficiently advanced. In absence of seeing through stuff, you can meet the same safety objectives by driving slower. What is clear is that the radar was on its way out regardless, the only actual issue is the timing which tesla is notoriously good at poorly transitioning between tech in the short term. To believe that Tesla would knowingly 'nerf' every model 3 sold...well, elon has his issues but he's not a complete idiot. This is kinda like the 'no one in the drivers seat' saga, took about a month but all of the loud mouths yelling about how that proved autopilot was doomed had to eat crow pretty hard. Tesla is not god, ap is not magic, but ap does have a large group of really smart, dedicated software engineers working on it with a pretty clear idea of what needs to be done. You have to bring that to table anytime you make an assessment.
 
Yes, that was my post on Reddit. Yeah, I know others don’t like the auto high beams, and I disabled it on my old one near immediately. The problem is you can’t even use cruise control now without them on for the new non-radar cars. So you have to decide between flashing people from behind with your brights which usually is a kind of road rage thing and driving manually.

It’s a little better if no one is around, but it still feels awkward even when alone them constantly going on and off. It may be an irrational fear, but I’m afraid there is going to be a cop on the side who sees it and pulls me over since it’s kind of suspicious. I don’t think there’s anything explicitly illegal about it, but the fact was if it makes me less comfortable than just driving for whatever reason than there’s no reason not to just drive. The whole point is making it less stressful, not more.
This is exactly why the not-yet-legal matrix headlights could become very important. The ability to run with high beams but selectively dim the beam for oncoming cars, highly reflective signs etc. could be a key benefit, assuming:
A. the matrix projector module is fitted,
B. it becomes legally approved in USA,
C. it actually works smoothly & seamlessly.

This last point is not a given; delay or uncertainty of the response, or inability of the matrix control to keep up with the vehicle suspension bouncing around, could easily result in strange super-bright flickering effects that would be very disconcerting to others. This would have the effect that Demonofelru is talking about: a great idea poorly executed that confuses or enrages people, instead of making everyone's life safer and better.

There's also the hardware-platform problem here:
All M3s and MYP for many months now have had some sort of projector lights, though AFAIK it's still not clear whether these have any matrix capability. The latest info points to a modest-resolution Samsung PixCell matrix headlight module built into a Hella light assembly, but whether these are already the 2nd-gen M3/MYP lights, or a 3rd-gen version yet to be fitted in production, is not yet clear. In any case, older M3/MYP and present MYLR don't have such matrix hardware.

So even if there could be a wonderful solution to maximize nighttime illumination for the Pure Vision AP, the approval, performance level and backwards-compatibility picture is kind of a mess.

As many users already know, there's been a somewhat ridiculed desire to wait for MYLR to be fitted with new(er) or new(est) headlights. I'm sensitive to the night-driving issue and I thought it'd be nice to have these, even on the uncertain possibility of future matrix-feature enablement. But now I see a perverse disincentive looming: even if AP really would benefit from fancier headlight technology, Tesla won't want this point to become too obvious - because it could generate a clamor for expensive retrofits to bring older cars up to snuff. Maybe they'll conclude it's better just to keep dumber headlight functionality for everyone, even a problematic Auto-high-beam implementation as described above.

I hope things play out better than this pessimistic scenario, but it's frustrating to think it could be so. As much as I try to see things from Tesla's side as they navigate changes and shortages, it looks like a bumpy ride with this latest roll-out, and I hate to think that they're painting themselves into any corners that could stifle future excellence.
 
The flashing highbeams seems to make AP nearly useless at night so I would say this certainly reduces functionality. The problem I have with the optimism is that removing radar seems to have been a business decision to sell a backlog of vehicles waiting for radar parts. Clearly they were not ready to sell fully functional cars sans radar and I don't trust their predictions about FSD stuff since they have a terrible track record.

Something is not right here. Paying customers are beta testing safety features. Features they thought were working and paid for. AP reduces the accident rate by nearly a factor of ten. Being forced to turn it off is not good.
 
Calibration can be done in the field (aka at home).

Cite? From reliable source?
The actual Tesla calibration procedure requires a 4 wheel alignment to be done as the first step: Tesla MODEL 3 SERVICE MANUAL: SENSOR - RADAR - FRONT (CALIBRATION)

Also, removal of the bumper requires two people and lifting the car to remove both front tires at the same time, usually done on a post lift: Tesla MODEL 3 SERVICE MANUAL: FASCIA - FRONT (REMOVE AND INSTALL)

So, sure, they can ignore their own service procedures, the one they would charge you for if you went in and paid, and just disassemble the whole front end of your car in your driveway and not align it first.
 
The actual Tesla calibration procedure requires a 4 wheel alignment to be done as the first step: Tesla MODEL 3 SERVICE MANUAL: SENSOR - RADAR - FRONT (CALIBRATION)

Also, removal of the bumper requires two people and lifting the car to remove both front tires at the same time, usually done on a post lift: Tesla MODEL 3 SERVICE MANUAL: FASCIA - FRONT (REMOVE AND INSTALL)

So, sure, they can ignore their own service procedures, the one they would charge you for if you went in and paid, and just disassemble the whole front end of your car in your driveway and not align it first.
If you have a valid ODIN token via Tesla Toolbox you can just click Calibrate Radar.

Also mobile service does alignments.

I looked into putting radar back on myself, extensively. Requires root though as you need to update a config file from RADAR:NONE to RADAR:CONTI. Root is near impossible to obtain without taking it apart which is well outside my comfort zone.
 
To believe that Tesla would knowingly 'nerf' every model 3 sold...well, elon has his issues but he's not a complete idiot.
You are overlooking the widely held belief that all of this came about due to radar part shortages that resulted in a hold on thousands of newly manufactured Y and 3 cars. The part shortage forced Tesla's hand on this. While I can certainly believe that vision-only was part of their roadmap, it is very clear that it was never part of their roadmap at this stage in the game. The state of the software limitations at the moment is proof enough. If their hand had not been forced due to part shortages, then I just can't imagine them rocking the boat in the middle of a model year for no real reason while nerfing the capabilities of their cars, however temporary that may be.
 
If their hand had not been forced due to part shortages, then I just can't imagine them rocking the boat in the middle of a model year for no real reason while nerfing the capabilities of their cars, however temporary that may be.
This is an excellent point! In addition, if their hand was forced due to a parts shortage then it's hard for me to be sanguine about their claim they will get all the new problems ironed out in two weeks. If then do then I will be extremely impressed.
 
Some interesting (albeit somewhat contrived) tests were performed by Greentheonly between a Model X with radar and Model Y without radar. Discussion is here:


Summary of it is:

Expt 1 - Cardboard box lined with some metal foil... should be good for providing some radar response.

- Model X with radar on AP mows it down, but does provide FCW warnings prior to collision
- Model Y without radar on AP mows it down with no FCW warning prior to collision

Expt 2 - Cardboard box with a pedestrian cutout pasted to it. Tesla's vision NNs classify this fairly well as a pedestrian

- Both Model X and Model Y automatically stop for the pedestrian cutout on AP
- Both Model X and Model Y run over the pedestrian cutout + box with no FCW, AEB when driven manually

Hard to draw very concrete conclusions from this but it clearly highlights that as things are today, cars with radar sensors do have some benefit over radar-less cars as evidenced by the FCW warning in Expt 1 for the Model X vs. none for the Model Y.

In general though, this ultimately highlights the need for cheap automotive lidar sensors to more reliably help assist with detecting obstacles on the road. Whatever Tesla's current vision-only approach is, it isn't really cutting it for obstacle detection on the road at the moment in some situations.
 
Last edited:
That is a terrible test for the effects of no radar. It's daytime. The road is straight and almost empty. Radar wouldn't give you any useful information. At best it would tell you there is a road in front of you. It might tell you it's raining.
I think the point there is only to show a counter example to reddit posts claiming vision-only performs poorly in the rain (or even in wet roads). It's not to contrast with radar. I think it's just to show it's not a given vision-only performs poorly in the rain, so there are other factors at play (I noted both cases complaining were of people that took delivery straight out in rain, but that may not necessarily have to do with it).
 
I think the point there is only to show a counter example to reddit posts claiming vision-only performs poorly in the rain (or even in wet roads). It's not to contrast with radar. I think it's just to show it's not a given vision-only performs poorly in the rain, so there are other factors at play (I noted both cases complaining were of people that took delivery straight out in rain, but that may not necessarily have to do with it).

Yeah, you nailed it. There are other factors at play: cars on the road.
The anecdotes I saw were the mist/ cloud kicked up by vehicles ahead of the Tesla impeded the AP’s vision such that it disabled. Not that it can’t keep active on an empty straight highway in the rain.

Aka without radar it lost “lock” / location awareness of those vehicles due to road mist.
 
Green does some more testing with Tesla Vision. He placed a brown box with foil (for great radar detection) in the middle of the road.

Model X with radar detected the box with a 30% probability and 60% of collision, warned the driver to take over but collided with box.

Model Y with Tesla Vision, did not warn the driver, and just hit the box.

When he placed a cardboard cut out of a person on the box, Model Y with Tesla Vision detected it as a pedestrian from afar and stopped perfectly.



IMO, this test illustrates the limitations of vision-only. Vision-only only works for what it is trained. Since Tesla Vision is not trained to see a box, it did not see the box. But Tesla Vision is trained to see pedestrians so it recognized what looked like a person and responded correctly.
 
Green does some more testing with Tesla Vision. He placed a brown box with foil (for great radar detection) in the middle of the road.

Model X with radar detected the box with a 30% probability and 60% of collision, warned the driver to take over but collided with box.

Model Y with Tesla Vision, did not warn the driver, and just hit the box.

When he placed a cardboard cut out of a person on the box, Model Y with Tesla Vision detected it as a pedestrian from afar and stopped perfectly.



IMO, this test illustrates the limitations of vision-only. Vision-only only works for what it is trained. Since Tesla Vision is not trained to see a box, it did not see the box. But Tesla Vision is trained to see pedestrians so it recognized what looked like a person and responded correctly.
I mostly agree with this but to play devils advocate, tesla vision successfully ignored a box while the radar thought the box was a car. It demonstrates the weakness of radar, it has no idea what it is looking at! But also the strength of radar, it can cover for something the vision system, can't classify, just with poor probabilities and insufficient confidence to actually take action other than a warning. Tesla vision also successfully errors on the side of caution and didn't hit a fake pedestrian, so in an actionable instance no change. I wonder how much general object detection will be in V9? Or if the current vision engine is basically V9, just without the city streets driving policy.