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

FSD Beta 10.69

This site may earn commission on affiliate links.
Pleasantly surprised that 10.69.2.2 visualized these traffic calming chicanes with the occupancy network gray blobs and slowed down to pass through without issue:
chicane front.jpg


As you can see, plenty of other vehicles (including past versions of FSD Beta) have run over the right side:
chicane right.jpg


I tested several around here querying OpenStreetMap data for traffic_calming=chicane, and the latest FSD Beta handled them all in both directions!
 
Lame Tesla apologist response. Typical. Let’s deprioritize school bus and children safety, and spend more time on Chuck Cook’s turn, fart sounds and games. Got it.
So many things to work on, so sorry they didn't work on your most important issue. Still no reason to shout at us. BTW, I've noted the school bus and school zone issue here for many months, so let's not be so quick with the name calling.

BTW, the people who do the fart sounds and games are not the people who program FSD. They are the MCU programmers. Whole different crowd. Same goes for the holiday light show developed by the controls group. They don't develop FSD either. There's lots of different software sets on a Tesla, developed by groups that don't work on self driving software. I suspect you already know that, but just don't care.
 
But why is this not a priority for Tesla? That’s the operative question - after a year of this.
Short answer - because they’ve made other things a priority.

As someone who’s been on FSDb for almost a year I can appreciate the improvements that have been made (along with some regressions.) Tesla has finite resources and they clearly can’t fix everything at once. If they had fixed the school bus issue then people would be complaining how it can’t handle a left turn.
 
Personal preference, guys. No way to prevent this. Not even a human can stop smoothly. We just need to get used to how FSD drives, we can't expect everyone to be happy with how it drives; these expectations are unreasonable.
Wait…haven’t you spent the last 20 pages complaining and nitpicking how FSD accelerates too slow, brakes too late, and doesn’t get unprotected left turns absolutely perfect. (and vociferously disagreeing when someone had the audacity to say that it was a preference of driving style?)
Disagree. My preference is that FSDb should slow down for red lights way earlier than what it does. It usually sees car/light in the distance before slowing down, so it should be doable. It’s inefficient and jarring coming up to a red light at 50, and then it slamming on the brakes to slow down at the last minute
Fixed it for you.
 
Exactly. But why? With all of those lame YT videos of people showing FSDb running over children dummies, you would think they would make it a priority. Not a good look, IMHO.

As someone who’s been on FSDb for almost a year I can appreciate the improvements that have been made (along with some regressions.) Tesla has finite resources and they clearly can’t fix everything at once. If they had fixed the school bus issue then people would be complaining how it can’t handle a left turn.
 
  • Like
Reactions: Supcom
because they’ve made other things a priority
Indeed, especially because FSD Beta is explicitly human supervised and currently designed for that, Tesla can prioritize things differently than if deploying a robotaxi. Before 10.69's occupancy network, FSD Beta would be much more likely to run into stuff that it doesn't explicitly know about such as green bollards, so Tesla prioritized building a general solution (as opposed to specifically teaching the neural networks about an endless long tail of objects that could be in the road).
 
C’mon @momo3605 …sigh. I insist emojis and /s are not needed. I would have thought you would have seen some of my posts in the last 100 pages. I expected responses from others…but…

To be clear: Humans are way better at anticipating and stopping than FSD.
Lol didn’t realize it was sarcasm. In hindsight I should’ve known. My bad 😁
 
  • Funny
Reactions: FSDtester#1
A simple concrete slab and rough framing so does not qualify as being half done a house lol, that’s maybe 30% at best!

Similarly I think people are so critical of FSD purely because the bar was set so high, it could never measure up against the promise of generalized robotaxis right around the corner despite how remarkable it is to have this technology out on public roads.
Wrong again.
You're forgetting that by this/that stage the design process has been done, documentation, approvals, etc. By the time the slab is down and the frame is up the process is approaching done. And pending disasters there is nothing likely to stop you from ending up with a complete house.
 
Important lesson I learned today that might come in handy for others: Pay attention to your navigation directions prior to activating FSDb.

My route this morning had two possible right turns. The first is "No right on red" so I drove past it to the second right turn which allows turns on red. I turned on FSDb just as I cleared the intersection for the first right turn, assuming the navigation would re-route and FSDb would continue along to the second turn without an issue. But instead, the navigation delayed rerouting, and FSDb tried what felt like an emergency maneuver to turn into a parking garage that happened to be between the two turns. If I hadn't been gripping the wheel, I don't know if FSDb would have successfully avoided the curbs in its attempt to make an emergency turn into the parking garage.
 
  • Informative
Reactions: FSDtester#1
For me the most noticeable latency is when the lead vehicle slows. You can watch FSD quickly fall behind as the lead vehicle gets closer and FSD needs to apply more and more excessive braking action. In those cases FSD feels like it's processing data over 1 sec old - a time when the lead vehicle is more distant and traveling faster. Of course the lead vehicle's distance should get closer as both vehicles slow but the velocity differential drives FSD's urgency to brake harder and harder. There's a few scenarios like this that suggesting latency is an issue

As I think about the abrupt steering could, in part, be intentional to reduce as much system delay as possible.

If they don't have a buffer or delay, people will complain it's too jerky and robotic.
 
  • Funny
Reactions: AlanSubie4Life
Wrong again.
You're forgetting that by this/that stage the design process has been done, documentation, approvals, etc. By the time the slab is down and the frame is up the process is approaching done. And pending disasters there is nothing likely to stop you from ending up with a complete house.
Do you work in construction? I'd be interested in seeing a milestone payment schedule or SOV that has 50% completion for a slab on grade and rough framing. You don't claim significant progress for submitting permit applications and then waiting for approval, and design costs scale with the complexity of construction that follows. Physical work in the field is what consumes the biggest portion of $$$, and detailed finishing is what takes the most time.

There are plethora parallels to draw between FSD's progress and this comic
 
Never got a response from the fellow from NYC about the specifics of his drive, but I found this video of a 10.69.2.2 drive in Manhattan, which does show the difficulty in dense-urban environments:


It seems like FSDb is way too timid for Manhattan, but it's interesting he's able to avoid full disengagements many times just by pressing the accelerator. Starting at 1:06 you can see FSDb stops for dense foot-traffic, but it's able to safely steer through two crosswalks with pedestrians when he presses the accelerator. Shows me it's probably capable of the kind of weaving required for driving in Manhattan, but it's just tuned to be too cautious at the moment.