Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register
This site may earn commission on affiliate links.
And at least during the HW2.x-3 transition it was a bit of a controversy that all this personal info (including things like login info for streaming services) was stored unencrypted for anyone who dumpster (or junkyard) dived the old MCUs to pull off at will.
Definitely there are setting items for FSD itself saved like following distance and mode (aggressive etc).

Question is - are there items that Tesla is storing that help with FSD. Like some help in selecting the correct lane or help with a stop sign. Tesla could definitely customize the drives that way ... Don't know whether they do.

There have been questions and speculations around this for years. We have all experienced the drives seem to get a little better as time goes by ...
 
  • Like
Reactions: sleepydoc
I typically drive in a way to maximize comfort and not being annoyed by FSD. So this lack of control is a bit disconcerting in highway driving. I think I just need to play a more active role if I think it will miss an exit.
I already do this by changing lanes as I think fit esp. near exits. But I also ignore a lot of lane change confirmation requests. So, will have to figure out how to make V11 work ... Hopefully the minimize lane changes setting will work to actually do what it says.
 
  • Like
Reactions: JHCCAZ
@verygreen can chime in if he'd like-- I THINK those are stored on the media computer-- not the driving computer.

And at least during the HW2.x-3 transition it was a bit of a controversy that all this personal info (including things like login info for streaming services) was stored unencrypted for anyone who dumpster (or junkyard) dived the old MCUs to pull off at will.

But then nobody can make your car behave badly in actual driving if they change your saved seat position like they could if they could hack neural network weights so it makes at least SOME sense that stuff would be less secure.
there are some limited settings that could be persisted. There are actually two (three, but third is only for dev cars) sets, some ap-related stuff in ap firmwares, some ap related stuff on infotainment most of which (but not all) you can control in the settings UI.

And then there are maps. I was exploring mapping recently and discovered that not only Tesla has those pretty frequent incremental updates, but when you have a route set - it asks the mothership and receives a route outline with very detailed info including where various intersections, crosswalks, stop signs, traffic control devices, speed limits, lanes, ... are. Every time you use "navigate to" function (and a bunch of that data is fleet-collected). As you can imagine updates in that sort of thing can have big impact on performance of AP without any code changes.
 
Again to be clear I was talking about on a specific release. For different software releases certain behaviors may be permanently (statistically) altered.
Yeah, that's what I meant. Every release will have its own set of average behaviors, and they will persist until the next one. I don't think they're tweaking parameters on the fly, at least on the scale of wide beta releases.
 
  • Like
Reactions: AlanSubie4Life
there are some limited settings that could be persisted. There are actually two (three, but third is only for dev cars) sets, some ap-related stuff in ap firmwares, some ap related stuff on infotainment most of which (but not all) you can control in the settings UI.

And then there are maps. I was exploring mapping recently and discovered that not only Tesla has those pretty frequent incremental updates, but when you have a route set - it asks the mothership and receives a route outline with very detailed info including where various intersections, crosswalks, stop signs, traffic control devices, speed limits, lanes, ... are. Every time you use "navigate to" function (and a bunch of that data is fleet-collected). As you can imagine updates in that sort of thing can have big impact on performance of AP without any code changes.
Thanks for that great information. What you just described about maps - route-specific data that is downloaded in near real time - is potentially very significant, and fairly surprising (to me) in a number of ways:

First, because most users agree that a great number of today's FSD foibles seem to stem from mapping errors or inadequacies.

Second, because the map database and updates have been kind of mysterious and often are assumed to be chronically out of date.

Third, because there's always a background debate smoldering here, about Tesla's supposed lack of interest or dependence on detailed maps for its real time driving performance.

Fourth, because I for one had assumed that drive-time data connectivity was basically irrelevant to Tesla driving operations, things like media streaming voice command processing, data collection uploads and so on are all things that don't really affect actual driving, so lack of connectivity is a temporary inconvenience.

But what you just revealed is something that could make a real difference in FSD capability, depending on whether the data connection is available. It begs a few further questions (answers for which may be unknown now, but if you have any insight please comment):
  1. Is this drive time server download a new thing, or something that has been added recently? Is it applicable only to FSD beta, or has it been around for legacy NoA as well?
  2. What happens if the call to the Mothership server fails during the route planning or route execution? Is there a timeout, or does it keep trying as the drive proceeds along the route? Upon failure, does the information devolve to the in-car stored map database (with presumably coarser and/or very stale information), or does it remember data from the most recent drive on or near that route?
  3. Do you think this map details feature will be an integral enabler of things like parking lot mapping for Actual Smart Summon and presumably required robotaxi pickup/dropoff operation?
Finally, is it your sense that this discovered capability is indeed highly significant for the upcoming development of FSD, or am I extrapolating too far from what you explained? Thanks so much for whatever additional commentary you can provide.
 
Last edited:
I haven’t. Ever.
May be we are going a little too OT .... but there are two specific things that come to my mind.

1. On a straight road on a frequent route, with one of the releases, the car would suddenly swerve left at a particular place. It happened every time - and I even posted a snapshot of it here. One fine day, it just stopped happening. And it has not happened again .... I've done frame by frame analysis and I don't know why.

2. At a newish traffic light, the car would erroneously take the right turn lane instead of straight. Used to happen almost always - but again one fine day stopped happening.
 
No - that was more about 11.4 onwards. Not a bug fix release .... since he is talking about making big changes in some NNs.


This is my interpretation as well. My guess is the subtext is "One more round of refinement needed [in order to achieve a truly unified single-stack on all of our neural networks]."

Not necessarily "...needed [before we continue rolling out V11]."
 
Does it say somewhere about a third of testers?
I did the math in another thread here:

 
  • Like
Reactions: DanCar