Anyone understand when/how the benefits of the FSD single stack transition will translate into improved performance for highway autopilot on AP/EAP (non-FSD) cars?
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Single stack for AP/EAP?
- Thread starter Stellavator
- Start date
Tam
Well-Known Member
Current Highway module is old and not based on Vision Only technology and Tesla asked us what to do if there’s an disagreement between Vision, radar, and uss.
The new module for city streets has the goal for Vision only so we don't need to give Tesla a fusion solution.
The theory is that Vision is so superior so if you can unify all modules city streets, Highway, parking into one single Vision stack then your Highway driving will just as sublime as what you see on all those FSD beta youtube.
But I would caution that the result in theory might not be the same in reality.
KArnold
Active Member
I am anticipating some significant regression on highways when single stack is released.
Technically there are HW2.5 owners with EAP (eg. OP I imagine, based on the signature). I wonder if they will maintain the old stack with NoA for those users or if they have been able to scale down the FSDb stack for HW2.5 use.
Tam
Well-Known Member
This is the wisdom from experience: A new update does not mean progress!
I am OP and I do have a 2018 but I upgraded to the FSD computer. I do not have FSD though, just EAP. Wondering when/if I will see the advances in highway autopilot that will come with the vision FSD single stack.
I agree, likely a step back at times, but highway AP is a relatively easy implementation, and the current one is old. I think with everything Tesla has learned from FSD work, if translated into highway driving, it will be way better.
I agree, likely a step back at times, but highway AP is a relatively easy implementation, and the current one is old. I think with everything Tesla has learned from FSD work, if translated into highway driving, it will be way better.
Of course no one has any idea. In all likelihood it will take 6 (super optimistic) to 24 months after V11 is released before the legacy AP stack is replaced by the FSD Stack.
Also your statement "..highway AP is a relatively easy implementation...." is incorrect. Beta is completely different and "identifies" and reacts to all kinds of objects and situations that are simply ignored in the legacy AP Stack. It will require MANY months of fine tuning, edge case experience and real world training to get FSD Beta to a competency level equal to AP Legacy when driving at high speeds on highways.
Example of what FSD Beta might do wrong: A barrier set up separating lanes from construction. People working and traffic flowing at highway speeds. A construction worker walks towards the barrier and FSD Beta may "freak out" and implement a full on panic stop.
Last edited:
Agree with JulienW, and this is also why we sometimes hear the refrain, "Why does Tesla's TACC phantom brake so much when [insert other mfr car here] has no issues? Can I just get dumb cruise control instead?"
That said, since the existing FSDb code knows when you're local vs hwy (for the most part), it might choose a different set of "rules" to play by when it's on the highway, which might eliminate a lot of false positives (phantom braking) on highways that currently happen on local streets.
It's all speculation, but I suspect it hasn't been an easy transition, because this time last year, many of us were predicting full stack at the beginning of 2022.
"Elon Musk has noted that Tesla is looking to widen the release of FSD Beta V11 before Thanksgiving. The V11 update is notable since it is the first iteration of the company’s advanced driver-assist system that operates using a single software stack".
Just be ready for surprises for months and releases after it is merged, a sudden jerk of the steering while traveling at 45mph in less traffic city road is quite different from traveling at 70mph with a concrete barrier and other vehicles around you on the highway. From past experience, beta simplify means the software is not ready but needs to learn from real-world driving data.
FSD Beta V11 partial release notes mention single stack for highways only, Smart Summon to come next month, says Musk - Tesla Oracle
Read the Tesla FSD Beta V11 (2022.40.5) release notes (partial). Tesla has deployed single stack for highway only. Smart Summon to come next month, says Elon Musk.
www.teslaoracle.com
Bump.
With the announcement that v11.3 will finally be the single stack FSD, will this bring the FSD improvements to AP/EAP owners on highway driving?
With the announcement that v11.3 will finally be the single stack FSD, will this bring the FSD improvements to AP/EAP owners on highway driving?
It is not a true single stack yet (integrated stack would be better) since Summon and Auto Park, and Parking Lots are still MIA but.......
Likely the legacy AP stack will eventually be retired and replaced by the new code but.......don't hold your breath. It will likely take a year or more of testing and tweaking before the the new code is deemed more reliable and better suited to replace the legacy code. But no one knows for sure.
beatle
Active Member
Seems that Tesla would do themselves good to bring in more testing data by extending the single stack release to AP/EAP owners for highway use. Navigation on city streets just wouldn't work on non-FSD cars, though who knows how they distinguish "city streets" in the software.
JetJock3315
Member
Anyone have updated info on this topic. Want to upgrade my new Plaid X to EAP now but only if it is based on the new single stack for highway. Wondering if it might make a difference if the car first downloaded FSD Beta for a 1 month subscription prior to purchasing EAP.
Single stack is only on FSD Beta for now. No one here has a timeline when it will come to AP/EAP.
JetJock3315
Member
Anyone know if 2023.20.7 affected EAP since it is now a single production release even for non-FSD users?
Similar threads
