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

Update enables cabin camera for driver monitoring....I guess excluding models S & X?

This site may earn commission on affiliate links.
Lawsuits and threats of regulatory action are eventually going to catch up with Tesla.
No model S/X made until mid 2021 has a camera, and the 3/Y doesn't have IR illumination for night (the new S/X does).

They're screwed if they need this to protect themselves, because the HW is either missing or insufficent in most cars.

How will Tesla argue that the DMC is needed while letting a bunch of cars drive around without it, or it being non-functional 50% of the day?
 
No model S/X made until mid 2021 has a camera, and the 3/Y doesn't have IR illumination for night (the new S/X does).

They're screwed if they need this to protect themselves, because the HW is either missing or insufficent in most cars.

How will Tesla argue that the DMC is needed while letting a bunch of cars drive around without it, or it being non-functional 50% of the day?
I don't think it'll be necessarily Tesla making that argument, but regulators or advocacy groups. If more camera based restrictions are only a software update away, they may make the argument Tesla should be forced to add it (after all, Tesla did do the three strikes restriction and increased nag frequency based on NHTSA investigation). The recent investigation can lead to this (we know NTSB has been pushing NHTSA on this).
 
Optical Cameras need light. We normally get that from the sun.
The sun goes away at night. Thus we need artificial illumination.
If we want light humans can see, we use a normal light. But if we want light that doesn't distract humans, we use infrared, which humans cannot see but cameras can.

The IR we use is right near the visible spectrum. The IR you mean by heat is far infrared, which requires completely different cameras due to drastically different wavelengths. It also doesn't let you do things like see if you're looking at a phone.

Functionally, the driver camera in older cars cannot see you at night, as it has no reliable light source to illuminate your face.
 
  • Informative
Reactions: jsmay311
after all, Tesla did do the three strikes restriction and increased nag frequency based on NHTSA investigation
Source? I've always found it funny that Tesla disables AP if you ignore it, while also advertising that it's safer than a human. I never heard it was forced by the NHTSA.

But we're in agreement- it will be an external force that makes them use the camera based DMS. However, that doesn't change the fact they will be screwed if they need it.

I'm kind of amazed they are dipping their toes in it voluntarily as it is, given how not all AP2 cars have it, and how fixated they appear to be on the torque sensor. As you say that seems to open the door to it only being a SW update away.
 
Source? I've always found it funny that Tesla disables AP if you ignore it, while also advertising that it's safer than a human. I never heard it was forced by the NHTSA.

But we're in agreement- it will be an external force that makes them use the camera based DMS. However, that doesn't change the fact they will be screwed if they need it.

I'm kind of amazed they are dipping their toes in it voluntarily as it is, given how not all AP2 cars have it, and how fixated they appear to be on the torque sensor. As you say that seems to open the door to it only being a SW update away.
The ODI report on page 7 specifically mentions the V8.0 update with the three strikes and the nag intervals mentioned seem to be based on the newer updates (the first release of AP allowed much longer times before nag). While the report doesn't explicitly mention NHTSA saying it was required, it's definitely a significant factor that allowed Tesla off the hook without doing anything further (I highly doubt if Tesla didn't do the change, NHTSA would let them off with no changes to the system).

NTSB is not shy they want Tesla to do even more, and NHTSA previously ignored them, but the new investigations that have started may change things.

Tesla could be doing the current updates to attempt the same thing (saying they are improving the monitoring system already), but that does open up the possibility of regulators asking for more (especially if they examine the details and find that the camera system works more like an OR function).
 
While the report doesn't explicitly mention NHTSA saying it was required, it's definitely a significant factor that allowed Tesla off the hook without doing anything further (I highly doubt if Tesla didn't do the change, NHTSA would let them off with no changes to the system).

I appreciate your data here, but as you say, there is no solid evidence that NHTSA required the 3 strikes. In fact, the document says in July of 2016, while 8.0 came out in September:

A safety-related defect trend has not been identified at this time and further examination of this issue does not appear to be warranted.

I fully agree that the NHTSA is paying more and more attention to Tesla. Not convinced that was true in 2016, and that they were the source of the three strikes rule.

As I say, I still find it amazing that Tesla simultaneously argues that AP is safer than a human, while also disabling it if they believe the driver is non-attentive. It's completely illogical (and to be clear, it's obvious that AP on it's own is not safer, which is why they disable it)
 
I appreciate your data here, but as you say, there is no solid evidence that NHTSA required the 3 strikes. In fact, the document says in July of 2016, while 8.0 came out in September:



I fully agree that the NHTSA is paying more and more attention to Tesla. Not convinced that was true in 2016, and that they were the source of the three strikes rule.
Well the update was significant enough to warrant mention even in the conclusion and it was a critical sign that Tesla was looking at HMI factors. We won't really know unless we wind back the clock where Tesla didn't do the update and see.
Tesla appears to have recognized HMI factors, such as the potential for driver distraction, in its design process for the Autopilot system. Tesla's design included a hands-on the steering wheel system for monitoring driver engagement. That system has been updated to further reinforce the need for driver engagement through a "strike out" strategy. Drivers that do not respond to visual cues in the driver monitoring system alerts may "strike out" and lose Autopilot function for the remainder of the drive cycle.

As I say, I still find it amazing that Tesla simultaneously argues that AP is safer than a human, while also disabling it if they believe the driver is non-attentive. It's completely illogical (and to be clear, it's obvious that AP on it's own is not safer, which is why they disable it)
Tesla is arguing when used properly (meaning driver paying attention) AP is safer than just the driver driving alone without AP. The disabling it as punishment is consistent with this (necessary as a disincentive against using it improperly) and doesn't contradict that, so not seeing a logic failure there.
 
Tesla is arguing when used properly (meaning driver paying attention) AP is safer than just the driver driving alone without AP. The disabling it as punishment is consistent with this (necessary as a disincentive against using it improperly) and doesn't contradict that, so not seeing a logic failure there.
The logic failure is to say that AP's functions outside of driver monitoring have any involvement in the reduction of crashes. We know that AP cannot be used when you are not attentive, and that lack of attention is a obvious reason for a crash. In fact, I'd say that when the car is used properly, the driver is paying attention, regardless of AP being on or off.

I'd like to see the statistics of crashes when the vehicle is being used properly vs not. I bet those statistics are just as good if not better than when it's on AP, if we're allowed to remove speeding, inattentiveness, substance use, etc. AP gets to remove them and even actively disable itself when the driver is not attentive, why don't we do that for the whole car crash?

I thought the point of driver assistance features were to assist real humans and their failures, not to only work when the human is at their best.

Now I want to go make an eye tracking camera with a light on it. I can call it the super safety device. The light is on only when you're looking right ahead. I can sell it as reducing accidents by 90% whenever it's "Active." ;)
 
Just got the latest 2021.24.4 update.

Seems like I have all the latest updates, except no update to the Cabin Camera on my radar & FSD equipped model.

Going back and looking at release notes posted elsewhere it seems Cabin Camera and Cabin Camera Updates are only select regions. I thought it was worldwide before. Either way wouldn't US be included?
 
Last edited:
Optical Cameras need light. We normally get that from the sun.
The sun goes away at night. Thus we need artificial illumination.
If we want light humans can see, we use a normal light. But if we want light that doesn't distract humans, we use infrared, which humans cannot see but cameras can.

The IR we use is right near the visible spectrum. The IR you mean by heat is far infrared, which requires completely different cameras due to drastically different wavelengths. It also doesn't let you do things like see if you're looking at a phone.

Functionally, the driver camera in older cars cannot see you at night, as it has no reliable light source to illuminate your face.
Yes you are correct about all that. I thought I had read months ago that the driver facing camera was specifically a thermal imaging device, which I remember thinking was kind of unexpected and advanced. Upon reading further today I understand that it was not even designed for this purpose, but more for future robotaxi passenger monitoring (which BTW would have similar challenges at night).

So in order to make it do secondary duty, or re-tasked primary duty as an IR driver monitoring camera, it would need to be sensitive in the near infrared and then require illumination diode(s) for nighttime. Do we even know if it has sensitivity in that range? Alternatively, is the illumination from the screen, even in nighttime mode, enough for driver monitoring in the visible spectrum? It doesn't take much illumination to get an adequate image for this task; somewhat noisy (i.e. gaihed-up) is not a real problem.

I do believe that the comma.ai device has an infrared camera/illuminator on the back for this purpose. That is meant to be mounted in a variety of different cars and so in that case there's no guarantee of any particular illumination source from the dashboard.
 
Do we even know if it has sensitivity in that range?
The new S/X cars have IR illumination. If they need to retrofit IR to cars, they'll probably swap out the whole housing and do the camera + lighting at the same time, so it doesn't really matter if the current camera has IR sensitivity, and Tesla has already determined they need this in the newest cars (so much for "all cars have all HW needed...")

The screen at night puts out very little light on purpose. It switches to black and goes very dim. I'd be surprised if the monitoring camera, which has to deal with direct sunlight can also deal with minimal light at night and still be able to see things like your pupils for gaze direction monitoring.
 
  • Like
Reactions: jsmay311
The cars have all the hardware needed for FSD, they were not saying they have all the hardware they need for the level 3 where the driver is still involved, just for level 5 where the driver is stretched out across the back seat...

I think they were just going to go with the steering wheel to check involvement but then other people started saying you need cameras monitoring the driver, and the 3/Y had that RoboTaxi camera in it but it's not originally for driver monitoring
 
Should be easy enough to swap out the mirror or the mirror and forward camera assemblies.

Data for camera already exists if they need to swap out. IR could be added to new assemblies.

IR could also be added to swapped cabin lights assembly as another option as power exists there two.

At a minimum add IR somewhere in cabin and keep existing cabin. Otherwise add cabin IR and swap cabin camera, which would require some / more updated coding.

Any of above is minor cost in HW / coding to implement in existing 3/Y.
 
  • Like
Reactions: JHCCAZ
But what about the S/X from before this recent refresh?

That's more challenging. Power is obviously there. Software wouldn't be to hard. But it's going to be a lot more involved to get a data connection up there "if" it needs a separate cable.

Good for Tesla anyway that most of what they need to do volume wise is S/Y. The S/X is low volume and they can limit to only cars with FSD on an add requested basis.

Still doable and without too much cost, but more complexity / time.