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

I was amazed by autopilot/fsd yesterday

This site may earn commission on affiliate links.
I am not sure we can infer that AP/FSD braked on signal. AP/FSD may have detected change in direction which driver did not and took appropriate action. I too take issue with people who ride their brakes on the motorway.
OP believes the indicator triggered the event and also stated that the lorry didn't move. That sounds like the car slowed down at a point where it didn't need to, a driver in control wouldn't have slowed down and a crash most likely would not have occurred if the car didn't slow down.
 
It wouldn't take long for you to realise that you've fallen asleep with the amount of rage going on around you plus you'd be back up at motorway speed in <5s so no issues

I haven't tried it :) but I'm pretty sure that it wouldn't get to the "just park it" stage as all the beeping and screeching would wake you up long before then :). Medical emergency would be a different thing ... and hopefully some passerby seeing me parked up neatly in lane would stop and summon some help. If I was off the road in a ditch might be that no one would see me ... of just mutter "That needs a Police Aware sticker" :(

AP/FSD may have detected change in direction which driver did not and took appropriate action

I'm still on AP1 ... would AP2 do that? or does it (still) need "significantly intruded into lane" before it reacts and yields?
 
I would be surprised if the current version of AP is reacting to indicators. I've noticed similar behaviour whilst on NoAP, when in the overtaking lane the car will slow to match the speed of the car/lorry on the inside lane, even when the car in front is pulling away. I have never quite found a reason for it doing this myself or on any of the forums.

However, I have just been out to do a full test on 28.2 for my youtube channel and whilst I did see the same behaviour a couple of time, this time the car waited for a bit and then speeded back up to max setting without me needing to press the accelerator pedal. I think this is new for 28.2?
 
I would be surprised if the current version of AP is reacting to indicators. I've noticed similar behaviour whilst on NoAP, when in the overtaking lane the car will slow to match the speed of the car/lorry on the inside lane, even when the car in front is pulling away. I have never quite found a reason for it doing this myself or on any of the forums.

However, I have just been out to do a full test on 28.2 for my youtube channel and whilst I did see the same behaviour a couple of time, this time the car waited for a bit and then speeded back up to max setting without me needing to press the accelerator pedal. I think this is new for 28.2?

Mine is in the max max setting mode and yesterday it gave us a 40 - 70 pedal down experience after the car in front pulled over. Thought to myself at the time, "Its just showing off now".
 
I don't think it really does react to indicators, that would be big news on the forums if it did.

I have experienced gentle slowing when in the outside lane for no apparent reason though. Sometimes it gets spooked by trucks on the inside lane if you are on a gradual right hand bend, as if it thinks the truck is moving across into your lane. Then it usually speeds back up again once it realises its mistake.

Basically it's a little unpredictable in speed control and it pays to hover your foot over the accelerator pedal rather than the brake pedal when in AP.
 
  • Like
Reactions: WannabeOwner
I might actually have dashcam footage of it and if I do, I'll post it and you can all judge for yourselves. Bear in mind that 2019.28.2 is fairly new and Tesla don't really give many notes about autopilot/fsd improvements, but I do agree that this would mostly likely be considered a larger incremental "ability". The fascinating thing with AI though is that sometimes you really don't know how it makes decisions. I trained a fair deal of neural networks myself and sometimes the re-inforcement learning process will make connections that aren't always initially very obvious, they alluded to that in the autonomy day presentation.
 
Ran some tests on 28.2, saw one example of slowing whilst in overtaking lane but it was very slight and very brief, it quickly worked out that it didn't need to do anything.
See NoAP section of video
)

Overall I would say that 28.2 is pretty well sorted.
Got 28.3 this morning, but not expecting anything other than bug fixes.