[UK] 2024.20 Software Update

[Tony Hoyle]

It's down to the mapping.. eg. they've never enabled auto lane change on the A555 despite it being a dual carriageway with a central barrier.

 

[MrBadger]

You will probably find that the mapping for those sections of road have potential uncontrolled access.

 

[JM_M3]

Auto lane change worked fine again today

 

[aq42]

It’s kind of user error, caused by Tesla poor implementation.

Once you start indicating, the clock starts, so if it can’t get one set of wheels over the white line within a set timeframe the car will abort, often violently.

There may be some real reasons for this, such as a car behind moving out behind you into the same lane as you were about to move into - it may come from several cars back.

But the common one/user error/poor implementation is not being in a position that the car detects you holding the wheel early enough into sequence that the car has time to get a set of wheels over the line. Sometimes it’s obvious that car hasn’t detected you so you have time to ‘let the car know you have applied pressure’ (often described as wiggling the wheel) but there is a window where you do apply turning force, car starts manoeuvring (in other words it hasn’t timed out before starting the manoeuvre) but does not get a set of wheels over the line in time so aborts mid manoeuvre. The car really should be in a position to know whether it can or cannot get a set of wheels over the line but for some reason thinks it can, start’s manoeuvre, then doesn’t make it in time so aborts. It’s easily reproducible once you realise the scenario and hence easy to avoid.

Aborting mid manoeuvre can be potentially be very dangerous not with standing the comfort factor, as on more than one occasion I have the car following me start to accelerate into the space I had partially vacated. So soon learned not to use auto lane change if someone too close behind/motorway too busy which really restricts it’s usefulness.

It's not poor implementation. The UNECE rules that apply in this country say that the lane change manouver has to complete within (IIRC) 9 seconds of the indicator going on. If it can't, it has to abort. If the car doesn't detect you giving the confirmation through wheel pressure within whatever the time limit is, it has to abort to comply with the law, even if the delayed manouver would be perfectly safe.

 

[monkeypig]

Has anyone else lost their seatbelt warning chimes when stationary/travelling at low speed? Not sure if mine disappeared after the spring update of 2024.20. You can hear them but only just, they operate normally at approximately 20mph. I’m in 24 m3lr

 

[FastLaneJB]

Is that not because there is a roundabout a bit further up the road?

Could be. I think it turns off quite a good few miles before it but it’s new to me so will see when I next head towards the big smoke.

 

[FastLaneJB]

I think with dual carriageways it’s more that there is uncontrolled access to that section of road, which would include your roundabout.

One example, which is a bit of an outlier in its own right as it allows NoA on dual carriageways and not just auto lane change, is A31 Hoggs Back heading west.

Auto lane change (and towards the end of that section, NoA functionality) is available all the while there is no way to access the road. The moment you get centre breaks, track access, farm gates, property access or lay-bys etc, auto lane chance ceases to be available. Sometimes it’s only for a short distance, 00 yards or so, and you might not know when/if it’s active - that said it’s very noticeable when NoA additionally becomes available when the double/single blue does a bit of a dance as it becomes active or not due to property access or centre divider gap. But NoA availability on uncontrolled roads is quite unusual.

Once you start spotting these uncontrolled access points, behaviour becomes quite predictable.

This makes sense, hadn’t considered this. Thinking about if then it wouldn’t be the roundabout but there is other roads with no real run up that people can enter quite slow into the road (When they aren’t driving a decently fast car).

 

[JIBYC]

Has anyone else lost their seatbelt warning chimes when stationary/travelling at low speed? Not sure if mine disappeared after the spring update of 2024.20. You can hear them but only just, they operate normally at approximately 20mph. I’m in 24 m3lr

I think the chimes are bugged in *.20. I'm getting them when I am in park.

 

[Drew57]

It's fascinating reading all these posts as I don't have EAP/FSD and had no requirements to drive in the dark since the .14 updates or .20.1 which I got earlier this week.

So far my car seems to behave exactly as it did prior to .14 so the updates have been basically irrelevant until I have need to drive early or late once the nights draw in. However I have noticed my neighbour (unusually) driving out and back soon after in his M3P on a couple of late evenings recently so presumably that's on the days he received the two latest updates!

 

[MrBadger]

It's not poor implementation. The UNECE rules that apply in this country say that the lane change manouver has to complete within (IIRC) 9 seconds of the indicator going on. If it can't, it has to abort. If the car doesn't detect you giving the confirmation through wheel pressure within whatever the time limit is, it has to abort to comply with the law, even if the delayed manouver would be perfectly safe.

I totally disagree. There is a window of opportunity where the car can start its manoeuvre and not complete it (it only has to get a set of wheels over the line and not actually have to fully complete the manoeuvre) within the time frame then unsubtly abort the manoeuvre. This should be totally predictable but the Tesla implementation does not predict this during this window.

Earlier in the sequence the car will detect that the manoeuvre will not complete and gracefully timeout, albeit probably to the annoyance of surrounding vehicles that see the car indicate but then not doing anything.

But the way that the car can start the manoeuvre with no chance of ever being able to complete it (ie get a wheel set over the line) within the time frame so rapidly swerves back from whence it came is quite frankly piss poor implementation. It should correctly predict that the manoeuvre will not complete in time and timeout gracefully in the same way it handles the situation earlier in the manoeuvre.

I can see an argument whereby something unpredictable occurs, a car suddenly appearing in the lane, but you can reproduce this abort quite reliably even with no other vehicles present.

I suspect that Tesla simply do not allow enough margin for error in this relatively common scenario. Nothing to do with the regs, just poor implementation in this scenario.

 

[Overworked]

It is the violent swerve back into the original lane when it aborts due to timeout that I find disconcerting. Everything is nice and smooth and relaxed up until that happens!

 

[FastLaneJB]

It is the violent swerve back into the original lane when it aborts due to timeout that I find disconcerting. Everything is nice and smooth and relaxed up until that happens!

Yes agree and MrBadger is right on this, it shouldn’t even start the manoeuvre when clearly it should be able to workout it won’t do it in time and also aborts.

Also frankly the rules on this should be a suggested time to change a lane so as not to be too slow and too fast but not force a car to abort if it doesn’t quite meet them.

 

[Tony Hoyle]

The abort provision is for something unexpected like a fast car undertaking that you can't predict.

It's not to make up for tesla's prediction of traffic behaviour being terrible so it gets taken by surprise constantly.

 

[FastLaneJB]

The abort provision is for something unexpected like a fast car undertaking that you can't predict.

It's not to make up for tesla's prediction of traffic behaviour being terrible so it gets taken by surprise constantly.

Yes obviously it should abort then but it’s definitely aborting when that’s not happening either.

 

[Aliomenti]

Automatic cooled seats seems to work properly now on 2024.20.1, ventilating the seats as needed.

On 2024.14.9, the seats only auto heated, but had to be manually set to ventilate.