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

2023.12...

This site may earn commission on affiliate links.
I've been playing around with TACC only a bit to see what the car would do near junctions.

I'm finding it quite reliably reduces to dead slow at junctions and roundabouts. Anyone else seen the same?

Sort of related, I noticed something very surprising yesterday. I use basic AP a lot on all the windy single lane rural roads round here (don't even have EAP/FSD). Surprised how many roads it actually works on to be honest - plenty that don't have painted central, or any lines. Even one recently where roadworks have just torn the surface off to redo it and it still stayed on the right side. But what it never seems to do is make any real attempt to slow down for curves (unless simply following another car, of course). Sometimes a little bit, but usually if there's a tighter turn that can't be taken at the set cruise limit I will have to intervene - either dabbing brakes to disengage AP or, if I was prepared enough, rolling down the speed limit a little with the wheel in time for it to reduce itself. If I do nothing it seems to reach its lateral g limit and creep to the side and ultimately have a stroke unless the turn eases and it can recover.

But today I took my normal 'highway' exit which is a sweeping fairly tight turn so you have to slow down almost immediately after leaving the highway lane...
turn.png


I always have to disengage here as generally can't roll down the speed quick enough. So it's a dab of brake then regen usually does enough (depending on how fast I was going on the highway!) But today, before I even moved for the brake it started to slow down by itself and took the entire turn absolutely perfectly staying in AP (well technically TACC only as my manual lane departure deactivates the Autosteer part). No car ahead to follow. Almost took my breath away, lol! This made me dare to wait a little longer than usual to see if it would even stop for the T-junction, but of course it was going to barrel straight out without a care so I stamped on the brakes!

No idea if this is some new behavior, or some local fleet learning thing (does it even do that?) or just something always inconsistent that might or might not have worked before. Interesting though. Basic AP on rural roads would be much more useable if it behaved like this all the time round bends.
 
No idea if this is some new behavior, or some local fleet learning thing (does it even do that?)

Yes, it learns typical speeds that the rest of the fleet takes a junction at - it treats it kind of like a temporary speed limit change. So if it was "fleet speed" that you experienced, you would have seen the AP set speed automatically reduced. If the set speed remained at whatever you had it set to, and it still slowed down for the junction, then this would suggest something rather more interesting is going on.
 
  • Informative
Reactions: LiamPope
Yes, it learns typical speeds that the rest of the fleet takes a junction at - it treats it kind of like a temporary speed limit change. So if it was "fleet speed" that you experienced, you would have seen the AP set speed automatically reduced. If the set speed remained at whatever you had it set to, and it still slowed down for the junction, then this would suggest something rather more interesting is going on.

Ah I see. I didn't pay attention to what the max speed was saying. It felt like an appropriate response to the turn, but could just as easily have been a step change in the set max which it gradually reduced to, giving that impression. I'll check next time.
 
Sort of related, I noticed something very surprising yesterday. I use basic AP a lot on all the windy single lane rural roads round here (don't even have EAP/FSD). Surprised how many roads it actually works on to be honest - plenty that don't have painted central, or any lines. Even one recently where roadworks have just torn the surface off to redo it and it still stayed on the right side. But what it never seems to do is make any real attempt to slow down for curves (unless simply following another car, of course). Sometimes a little bit, but usually if there's a tighter turn that can't be taken at the set cruise limit I will have to intervene - either dabbing brakes to disengage AP or, if I was prepared enough, rolling down the speed limit a little with the wheel in time for it to reduce itself. If I do nothing it seems to reach its lateral g limit and creep to the side and ultimately have a stroke unless the turn eases and it can recover.

But today I took my normal 'highway' exit which is a sweeping fairly tight turn so you have to slow down almost immediately after leaving the highway lane...
View attachment 946496

I always have to disengage here as generally can't roll down the speed quick enough. So it's a dab of brake then regen usually does enough (depending on how fast I was going on the highway!) But today, before I even moved for the brake it started to slow down by itself and took the entire turn absolutely perfectly staying in AP (well technically TACC only as my manual lane departure deactivates the Autosteer part). No car ahead to follow. Almost took my breath away, lol! This made me dare to wait a little longer than usual to see if it would even stop for the T-junction, but of course it was going to barrel straight out without a care so I stamped on the brakes!

No idea if this is some new behavior, or some local fleet learning thing (does it even do that?) or just something always inconsistent that might or might not have worked before. Interesting though. Basic AP on rural roads would be much more useable if it behaved like this all the time round bends.
Which software version?
 
Anyone noticed you can now view ALL the cameras from the service menu? Kinda nice to make sure they are all working and actually see what those pillar cams can see for once.

I'm not totally sure when this appeared but I do believe it was recent?
 
Spent a bit of time as a passenger on 2023.12.9.1 in urban streets with lots of 20mph around, especially in side roads.

Previously the car would set the sped limit the moment it picked up a speed sign, even if clearly on an unrelated road. Certainly with non urban streets.

But yesterday, even though speed signs on adjacent roads were clearly picked up and visualised, I did not notice any occasion where it incorrectly set the speed limit to match.

Wondering if the car is now more spatially aware of where speed signs are, and the relevance that limit has on the cars current path.
 
I am struggling to imagine how vision can get road signs right 100% of the time ... some yobs have turned the one in my village round to face the other way ...

... and also (my day job is Database) struggling to understand why there isn't a public GPS database which is 100% accurate (except for mobile works) for everything in the street that matters. Even that, given the need to get permission for road works, well in advance, should be a no brainer. Even emergency roadworks - burst water main / whatever - has a procedure.

Am I missing something obvious that makes an accurate GPS database impossible?
 
There's actually a whole UNECE project for building such a database, that's supposed to encompass things like temporary restrictions. Who knows how long that'll be if it ever happens though.. trying to get the authorities of multiple countries to record their roadworks in the same way..

 
Last edited:
  • Like
Reactions: WannabeOwner
I am struggling to imagine how vision can get road signs right 100% of the time ... some yobs have turned the one in my village round to face the other way ...

... and also (my day job is Database) struggling to understand why there isn't a public GPS database which is 100% accurate (except for mobile works) for everything in the street that matters. Even that, given the need to get permission for road works, well in advance, should be a no brainer. Even emergency roadworks - burst water main / whatever - has a procedure.

Am I missing something obvious that makes an accurate GPS database impossible?
I remember the WDA (Water Data Archive) in the 90s, still seeing new initiatives relating to the same
 
Spent a bit of time as a passenger on 2023.12.9.1 in urban streets with lots of 20mph around, especially in side roads.

Previously the car would set the sped limit the moment it picked up a speed sign, even if clearly on an unrelated road. Certainly with non urban streets.

But yesterday, even though speed signs on adjacent roads were clearly picked up and visualised, I did not notice any occasion where it incorrectly set the speed limit to match.

Wondering if the car is now more spatially aware of where speed signs are, and the relevance that limit has on the cars current path.
When I was 12.9.1 it was still picking up a rural side road 30 sign on a 50 limit road. Worse still, the sign was correctly positioned. Still doing it on 20.4.1 too!
It also randomly picks up a 20 school limit which is time of day based anyway. It’s a regular short trip I make 4 or 5 times a week at odd times of day. Picks the sign up less than once a week!
 
I am struggling to imagine how vision can get road signs right 100% of the time ... some yobs have turned the one in my village round to face the other way ...

... and also (my day job is Database) struggling to understand why there isn't a public GPS database which is 100% accurate (except for mobile works) for everything in the street that matters. Even that, given the need to get permission for road works, well in advance, should be a no brainer. Even emergency roadworks - burst water main / whatever - has a procedure.

Am I missing something obvious that makes an accurate GPS database impossible?
I think Tesla use more than one database. Back in early 2020 it did not respond to vision based speed limits and my car seemed to change either before or after the limit. I thought it might use Open Street Map (OSM) Database. OpenStreetMap
Even now it sometimes drops to a 30 limit for no reason in a 50 or 60 limit road. I wonder if locals are messing with the database for their own nefarious intents. I also have a slip road near me which goes from 30 to 40 to 60? to 70 all in about 100m. The car increases speed at each limit even if cruse was turned on in the 30 limit. Something it does not do anywhere else - dangerous? All makes me wonder if some speed limit "databases" override others? OSM had an app called OsmAnd its a bit memory hungry but gives you an idea what detail is on this database.
 
Last edited:
... and also (my day job is Database) struggling to understand why there isn't a public GPS database which is 100% accurate (except for mobile works) for everything in the street that matters.

There are at least 3, if not more, agencies involved. Within same nation (forgetting any overlap between England, Wales and Scotland), you are likely to have national, county and councils all involved, often with the latter not knowing their (responsibility) boundaries. Throw in the likes of TfL etc…

Does that answer your question 🤔
 
  • Like
Reactions: WannabeOwner
Does that answer your question

Nope! We have all sorts of similar "competing data" to aggregate in our organisation. We manage just fine and have a very high degree of accuracy (even though we have zero control over the source data). Organisations which are set up for each of them to have a standard, even if they have no design-intent to be compatible (which in itself would be daft), should be within the ken of competent people to aggregate.