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

Speculation about EAP and FSD codelines

This site may earn commission on affiliate links.
I presume the AP1 and AP2 are also separate though related firmware versions?
AP1 is part of regular firmware.
AP2 is not but has a pointer to it.
AP1 and AP2 codebases are totally separate for multiple reasons mostly due to totally different architectures and hardware.

MCU firmwares for AP1, AP2 and no-AP are 100% identical byte for byte (same image is downloaded by every car no matter when it was made).
 
Essentially, What I infer from earliers posts, There would be multiple firmwares , Instrument Cluster, MCU, AP1, AP2 , anything else to be listed ?

But how this "Versioning is determined" ? Earlier it used to be "XX.XX.XXX" At any given Update there could be 1 or more firmwares could be updated. It should follow a sequence. Its bit complicated to imagine from what I can imagine .
 
Essentially, What I infer from earliers posts, There would be multiple firmwares , Instrument Cluster, MCU, AP1, AP2 , anything else to be listed ?

But how this "Versioning is determined" ? Earlier it used to be "XX.XX.XXX" At any given Update there could be 1 or more firmwares could be updated. It should follow a sequence. Its bit complicated to imagine from what I can imagine .
IC+MCU+AP1+GTW+things like motors, MBS, HPWC are the same firmware (single image that contains all the parts for everything). IC and MCU even run like 90% of code in a shared manner and as such have 100% copy of the firmware image of each other.
The firmware update on MCU looks at car setup and then updates bits that make sense for the given car config.
AP2 firmware is separately packaged. AP2.5 right now is different packaging of the same codebase as AP2.0, but we'll see how that changes once HW 2.5 ships in cars accessible to regular people, possibly similar to MCU would be single image that works on AP2 and AP2.5 and just uses what's there.
It's yet unclear how the firmware for the other node in hw2.5 would be packaged (but I did not look too much into it), possibly would be more or less identical to first node?

ICE (the big screen) of model 3 is yet another firmware bundle that seems to be different, but has shared parts with S/X. There are (yet unconfirmed) hints that ICE is x86 based.
 
Hi,

I want to ad a data point to the FSD effort by Tesla.

Some time ago there where reports of Nico Rosberg visiting Tesla.

Today an article was published in a German newspaper about this visit.


Nico Rosberg im Interview: „Beim autonomen Fahren vermisse ich nichts“

I am currently on the go, so will summarize this in the evening in English.

Quote:
Das haben Sie gerade erst erlebt in Kalifornien. Wie fühlt es sich an, im Auto zu sitzen und Passagier ohne Fahrer zu sein?


Ich fand es überhaupt nicht merkwürdig, weil ich volles Vertrauen in die Kompetenz der Ingenieure habe. Als wir mit einem Tesla unterwegs waren im Silicon Valley, hatten wir auf dem Beifahrersitz einen Laptop, darauf wurde das Kamerabild gespielt, wurde also gezeigt, was das Auto in dem Moment wahrnimmt. Menschen, andere Autos, Ampeln, all das ist darauf digital markiert. Der Computer scannt permanent die Umgebung ab und kann darauf reagieren. Für den Notfall musste allerdings noch ein Ingenieur am Steuer sitzen. In Kalifornien ist das autonome Fahren nicht erlaubt, anders als in Texas oder Arizona. Dort darf man ohne eine Person am Steuer unterwegs sein. Es gibt Familien, die das schon genau so machen.

End-Quote

Thank you!
 
I did a google translate. Mind you I know about 4 words of German, but it looks like it did a passable job, at least the gist of it... That said, I don’t know any families in Texas that drive around without anyone in the driver seat...lol

Any idea how long ago he visited Tesla?

“You've just experienced this in California. How does it feel to sit in the car and to be a passenger without a driver?


I did not find it strange at all because I had full confidence in the engineers' abilities. When we were on the road with a Tesla in Silicon Valley, we had a laptop on the passenger seat, the camera was played, so it was shown what the car perceived at the moment. People, other cars, traffic lights, all this is digitally marked. The computer continuously scans the environment and can react to it. In case of emergency, however, an engineer had to be at the wheel. In California, autonomous driving is not allowed, unlike in Texas or Arizona. There you can drive without a person at the wheel. There are families who do it exactly the same way.”
 
  • Like
Reactions: sreehyd and zmarty
My reasoning for EAP and FSD being separate codes is based on this;

#1 Elon is sane
#2 EAP doesn't show any promise of coast to coast autonomous capability
#3 Elon believes, that coast to coast autonomous drive is possible within 5 months

Conclusion; there must be separate FSD code, which is far more advanced than EAP.

If anyone of those 3 premises is not true, then conclusion is not necessarily true.
 
My reasoning for EAP and FSD being separate codes is based on this;

#1 Elon is sane
#2 EAP doesn't show any promise of coast to coast autonomous capability
#3 Elon believes, that coast to coast autonomous drive is possible within 5 months

Conclusion; there must be separate FSD code, which is far more advanced than EAP.

If anyone of those 3 premises is not true, then conclusion is not necessarily true.

Of course there is a separate codebase for FSD, I think we know enough to conclude that. I believe it is probably based on Nvidia's efforts and collaboration on some level (at least the video was, I believe).

EAP is probably just a crude emulation of AP1 pasted onto a codebase imported from AP1, that they could get quickly going. One bit of thinking that supports this is the MobilEye on AP2 board angle (there was even room for it, an empty area on the board). AP2 was originally going to run like AP1 at first, until the FSD codebase is completed, and a MobilEye chip was going to be there for this purpose...

When MobilEye ditched them (and I do believe MobilEye did ditch them, possibly after the Brown crash, possibly due to Tesla's reluctance to share data etc.), they found themselves in having to develop something for the meanwhile. So EAP as we now know it, was born, somewhere between spring/summer and October of 2016, probably.

Remember, we saw autonomous-looking hardware on a Model S already back in 2015. I think Tesla has been working on the Drive PX / Nvidia angle for some time now, while also entertaining MobilEye based alternatives (perhaps the AP 1.5 angle with two forward cameras and rear radars) on the side.