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.
The next big milestone for FSD is 11. It is a significant upgrade and fundamental changes to several parts of the FSD stack including totally new way to train the perception NN.

From AI day and Lex Fridman interview we have a good sense of what might be included.

- Object permanence both temporal and spatial
- Moving from “bag of points” to objects in NN
- Creating a 3D vector representation of the environment all in NN
- Planner optimization using NN / Monte Carlo Tree Search (MCTS)
- Change from processed images to “photon count” / raw image
- Change from single image perception to surround video
- Merging of city, highway and parking lot stacks a.k.a. Single Stack

Lex Fridman Interview of Elon. Starting with FSD related topics.


Here is a detailed explanation of Beta 11 in "layman's language" by James Douma, interview done after Lex Podcast.


Here is the AI Day explanation by in 4 parts.


screenshot-teslamotorsclub.com-2022.01.26-21_30_17.png


Here is a useful blog post asking a few questions to Tesla about AI day. The useful part comes in comparison of Tesla's methods with Waymo and others (detailed papers linked).

 
Last edited:
Did some more interstate driving today and same results I have been having (blowing by almost every exit). Every time I’m in the far left lane and it makes no attempt to start getting over until I’m a mile away from my right turn. On one attempt it got to within one lane of the Exit lane but it was backup up. So Beta thought it would be good idea to just turn on the signal and stop on the Interstate.

On another it was in the HOV lane and passing the last double stripped line exit and it was not going to get out of the HOV lane (as it does EVERY time) so I turned on the signal to forced it out. It still stays in the left lane next to the HOV lane until a mile away from an Interstate split (I85 left lanes and I75 right lanes) where it needs to go right on I75. Usually the traffic is backed up and I would be trapped going I85 but it managed to just barely make it over to the right split but ONLY because traffic was moving. Again most times traffic is backed up and this will never happen.

All other parts of Interstate driving seems so much better on Beta than legacy AP stack but…….why is the Interstate logic so bad at anticipating needed strategy to exit. Waiting on the last mile is fine if you are on 3 lanes of moving Interstate. However in urban settings and needing to get over 6 or 8 lanes with heavy, slow, inconsistent speed traffic you can’t wait to the last mile.
 
Did some more interstate driving today and same results I have been having (blowing by almost every exit). Every time I’m in the far left lane and it makes no attempt to start getting over until I’m a mile away from my right turn.

Just manually signal yourself. There's no point in testing this aspect, as we know Tesla is improving it with every new release.
 
Did some more interstate driving today and same results I have been having (blowing by almost every exit). Every time I’m in the far left lane and it makes no attempt to start getting over until I’m a mile away from my right turn. On one attempt it got to within one lane of the Exit lane but it was backup up. So Beta thought it would be good idea to just turn on the signal and stop on the Interstate.

On another it was in the HOV lane and passing the last double stripped line exit and it was not going to get out of the HOV lane (as it does EVERY time) so I turned on the signal to forced it out. It still stays in the left lane next to the HOV lane until a mile away from an Interstate split (I85 left lanes and I75 right lanes) where it needs to go right on I75. Usually the traffic is backed up and I would be trapped going I85 but it managed to just barely make it over to the right split but ONLY because traffic was moving. Again most times traffic is backed up and this will never happen.

All other parts of Interstate driving seems so much better on Beta than legacy AP stack but…….why is the Interstate logic so bad at anticipating needed strategy to exit. Waiting on the last mile is fine if you are on 3 lanes of moving Interstate. However in urban settings and needing to get over 6 or 8 lanes with heavy, slow, inconsistent speed traffic you can’t wait to the last mile.
Which setting were you using? I had been using "assertive" and had the same lane movements as you (one lane at 1 mile, next lane at .5 miles) though I'm told if you change to "chill" it doubles the distance (one lane at 2 miles, one lane at 1 mile). I haven't had a chance to confirm this yet but if accurate this would really help with the exits with heavy traffic.
 
Which setting were you using? I had been using "assertive" and had the same lane movements as you (one lane at 1 mile, next lane at .5 miles) though I'm told if you change to "chill" it doubles the distance (one lane at 2 miles, one lane at 1 mile). I haven't had a chance to confirm this yet but if accurate this would really help with the exits with heavy traffic.
I'm on Average now (was on Assertive and changed a week ago). May give Chill a try but since I live and mostly drive in the city feel that Assertive is more applicable to in city driving.
 
I'm on Average now (was on Assertive and changed a week ago). May give Chill a try but since I live and mostly drive in the city feel that Assertive is more applicable to in city driving.
I've tried Average, Chill and Aggressive and none matter. FSDb is pretty bad at getting over to take some exits here in Maryland, while for others it's even better than NoA ever was. So not always a regression, it's just different, and hopefully better in 11.4.x.
 
  • Like
Reactions: RabidYak
I've tried Average, Chill and Aggressive and none matter. FSDb is pretty bad at getting over to take some exits here in Maryland, while for others it's even better than NoA ever was. So not always a regression, it's just different, and hopefully better in 11.4.x.
In some cases it appears to be caused by bad map data. For example, for one interchange I take the nav system shows that the right two lanes are acceptable, which is sort of true. The problem is that it shows that about a mile before the right most exit lanes starts. So FSDb moves over to the second lane from the right, which is correct according to the nav system for where we want to go, but then in a mile when the exit lane starts, it is now in the third lane from the right which doesn't work, and it often doesn't have space/time to move right to make it on its own. (And this is still with the most current map data.)

I'm not sure where this map data comes from or how we might go about getting it fixed. But it is something that Tesla will need to deal with eventually.

edit: I looked in OpenStreetMap and can see what is likely the issue. OSM shows the extra lane as soon as the lane starts widening, which is way before there is a painted line marking splitting the lane. So FSDb is seeing a wide lane where OSM is reporting two lanes. (FSDb should probably handle it better when the map data shows more lanes than it is seeing, like maybe realize that the "wide lane" it is seeing is actually two.)
 
Last edited:
Which setting were you using? I had been using "assertive" and had the same lane movements as you (one lane at 1 mile, next lane at .5 miles) though I'm told if you change to "chill" it doubles the distance (one lane at 2 miles, one lane at 1 mile). I haven't had a chance to confirm this yet but if accurate this would really help with the exits with heavy traffic.
They should improve the traffic data by lane, and then the planner should use the traffic data to see there is traffic in those lanes and it should get over sooner then later when there is heavy traffic (red).
 
I wouldn't worry about Tesla's ability to fix these minor lane selection issues down the line. Based on the 11.4 notes, Tesla is on the right track to fix them.

I'd rather Tesla focus on eliminating all safety related issues: creeping too far, stopping in the middle of intersections, unnecessary slowdowns, merging in presence of large trucks, running reds, etc.

I don't need to tell them any of this. They already know.

I'm very excited about the direction and progress of V11 if you can't tell. I can see the light for HW3, still not sure about 5x human safety, but even if HW3 gets to 2x human safety, that'd be incredible and would provide amazing value as a driver-assist feature.
 
I wouldn't worry about Tesla's ability to fix these minor lane selection issues down the line. Based on the 11.4 notes, Tesla is on the right track to fix them.

I'd rather Tesla focus on eliminating all safety related issues: creeping too far, stopping in the middle of intersections, unnecessary slowdowns, merging in presence of large trucks, running reds, etc.

I don't need to tell them any of this. They already know.

I'm very excited about the direction and progress of V11 if you can't tell. I can see the light for HW3, still not sure about 5x human safety, but even if HW3 gets to 2x human safety, that'd be incredible and would provide amazing value as a driver-assist feature.
And howdy. 2x!? Where is all that hope coming from?

They've played the cards they had for HW3 and now there's a little bit of mopping/wrapping before they move onto bigger and better things (HW4).
 
And howdy. 2x!? Where is all that hope coming from?

They've played the cards they had for HW3 and now there's a little bit of mopping/wrapping before they move onto bigger and better things (HW4).
Still some cards to play for HW3 after all. This improvement I believe was unexpected by most.
"the ability for the vehicle to stop when backing up and by increasing the top speed it can operate in a new update."

Tesla makes Automatic Emergency Braking even safer in update
 
Does anybody know if the FSD beta enrollment has reopened? I wonder if I subscribe how soon I will get 2023.15. Currently on 2023.12.1.
2023.6.15 will be the next FSDb release. I’m at 2023.2.12 and have been waiting for it for over a month. Been subscribed since a week before the NHTSA recall. I have been holding off on updates so that I don’t get beyond the current FSD branch. If you are currently on 2023.12.1 then you won’t get it for a while. 2023.6.15 needs to be released before they make a version beyond 2023.12.X. Tesla won’t downgrade your software for FSD. I would advise you not to update with “non-FSD” updates if you plan to subscribe. Regular updates are much farther along than the FSD versions.