-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Tesla Software updates - Australia
- Thread starter Grenadine
- Start date
I enjoyed your videos and your experiences mirror mine. Of the five Enhanced Autopilot features (Auto Lane Change, Navigate on Autopilot, Autopark, Summon, Smart Summon) only one feature works (Auto Lane Change). The other four are broken or pointless. Let me elaborate:
- Autopark - doesn't work
- Summon - pointless but some might find it useful on occasion.
- Smart Summon - too slow to be useful and failed completely when I tried.
- Navigate on Autopilot - seems to be short sighted and merges behind slower moving cars before jumping back into overtaking lane. Failed to exit the highway and aborted.
Charging $5k for that package is a joke. I *might* consider paying $1k for auto lane change but definitely not the rest.
Summon is very useful when someone blocks the drivers door, or if there is a car space that you can fit into but cannot get out of the drivers door….so yeah pointless until you need it, but when you need it summon is brilliant, if it works at that time.
sulrich70
Member
To be honest, with the likes of BYD and other manufacturers catching up - this should be thrown in for people at no cost.
And having to pay extra for auto park on a car in this price range is not good. And it doesnt work as you have said.
Kia on their ICE vehicles have summon. Alot of manufacturers also have lane assist and lane changing. My 2016 commodore had auto park. These should be free
Maximillan
Member
I drove a Peugeot 508 rental for a few months
I discovered that it was steering around corners for me, a bit of a surprise
Priit
Active Member
Well, I have been away on holidays for the past month, so have not been able to try any of that but should be back this week and will take a look. I wonder how close the sides the summon allows.
My garage has really narrow door, so when I back my Tesla in there it leaves about 5cm on both sides between the mirrors and the garage door, also my garage is really short so I need to get it to about 10cm from the back to allow the garage door to close. Once in the garage there is enough room to open the driver door to the first position and get out of the car as long as I leave 5-10cm on the other side to the wall, so has never been an issue for me and by now is mostly automatic action for me to back into the garage and end up in right spot. For me summon would not be useful for this as I want to end up in the garage anyways to enter my house through garage but would be interesting to see if it even attempts this.
I’ve used it to fit in a commercial carpark with a wall one side and a column the other. I was around 15cm off the wall. You can also set it to allow tigher gaps. Having said that, if I’m standing too close to the car when using summon it freaks out and stops. Same problem if too far away.
I suspect it might struggle with your garage setup.
Priit
Active Member
So, I had a go with the new EAP features and have come to pretty similar conclusion as most others have here. The auto lane change should be part of basic AP and works amazing. the NoA on the other hand is not ready for public consumption. I had 3-4 times it tried to change lanes, it started indicating and then took way too long indicating so that 1-2 seconds after mine started indicating the car behind me figured I was just indicating and not going so they indicated and pulled beside me and then my car could not go, so it backed off and then tried again almost turning into the car next to me, I had to take over so many times. the basic AP with the auto lane change is awesome though. NoA took an exit for me and it did not feel or look natural at all, so I ended up disabling the NoA. The auto park is way too slow and even worse than me at parking and the summon does not even attempt to enter my garage, just stops at the door. From this software stack it looks like its a long way from being on the level with human drivers. I am a bit surprised as the basic AP on my Tesla has been rather good so the I was expecting the additional features to be similarly functional and stable, but they seem to be way early alpha release stage at least here in AU. I would give it another try once it reaches beta stage, but in its current form, its not worth the 5k it costs.
I have been using NoA. Doesn't seem to add much of use though. Was nearing an exit on the freeway and I was anticipating seeing it plan and take the exit. Then it started to rain lightly and "NoA not available due to reduced visibility" message. So it reverted to EAP meaning I had to do the exit manually. The overtaking feature is nice but takes a bit of getting used to. I use Autosteer most of the time anyhow, so NoA really does not add much for my use. But I will keep using it while it lasts for the novelty. Autopark I have tried a few times and it works for me about 1/4 of the time. Rest of time it borks for one reason or another making me look like a goose.
Priit
Active Member
I gave NoA another chance this weekend and did 2 days of testing. For me it appears NoA does not really look at the road information as much as it does some form of internal map information that is not visible to user. On Mad Max mode it was flat out dangerous and tried to turn into the cars next to me. on the medium option it is much better as it does not try to run into things but there are a few reasons I am still not using it.
1. It appears to drive like average QLD driver and tries its hardest to stay in middle lane regardless if the left lane is empty.
2. very inconsistent when passing other cars (i.e. sometimes goes back to original lane right after passing the car, sometimes waits a minute or a few km-s and sometimes does not appear to have any interest going back to left lane)
3. Appears to hate catching up to a car that is in the left lane. When I catch up, it hits the brakes rather suddenly and annoyingly. where it does not appear to have same effect when catching up to a car in the right lane.
4. exit taking is not as smooth as changing lanes on AP, it appears to start normally then do a rather sudden and jerky turn to center on the exit lane, that is rather unexpected and feels off.
I also noticed that areas that used to have roadworks on M1 here that are now really good and unrestricted road now appear to have a reaction from the car. Every time I approach a place where roadworks had speed reductions 2 years ago, the car slammed on brakes even though there was not speed limit change and the speed limit sign on screen still showed 100. Also at every location the road narrowed during roadworks few years ago the car would indicate on NoA and change to center lane. If I cancelled changing lane, it would try again every few seconds, so after 10th time it hit the brakes on clear 3 lane road with no obstructions, so I took over.
This leads me to suspect that the speed limit and map shown to the user may not be the one NoA uses on background and they may be using some really od data in my area. They talk about using vision based response, but that does not appear teh case here, the only thing they react to based on vision is other cars.
@Priit I agree with all your points, except that no Tesla will drive into another car on its own, Mad Max or not. It may just feel closer than what you're used to - if you're a conservative driver that is.
In 3 years of having no AP improvements to speak of in Australia, I think it's fair to say we can stop testing/reporting until they declare they're working on FSD for our market. Which will happen sometime in early 2030. Maybe.
For effs sake, our cars still set the speed limit based on a 40 sign on a bus, or think a lane open/closed indicator in a tunnel / on a bridge is a traffic light.
In 3 years of having no AP improvements to speak of in Australia, I think it's fair to say we can stop testing/reporting until they declare they're working on FSD for our market. Which will happen sometime in early 2030. Maybe.
For effs sake, our cars still set the speed limit based on a 40 sign on a bus, or think a lane open/closed indicator in a tunnel / on a bridge is a traffic light.
Priit
Active Member
I dont consider myself a conservative driver and to be fair most of my mates consider me to be a rather Mad Max driver most of the time. However I do agree with you that the car may not actually crash into others, however it acts too erratically on the road. For me in a 30km stretch of M1 it had 4-5 instances where it started indicating, took way too long indicating 3-4 seconds, so the car behind me pulled next to me, then my car starts swerving toward the car that just managed to accelerate next to me, even though the swerve action stops at the edge of my lane, it causes the car now next to me to also swerve to avoid staying too close to me and honk their horn to draw attention to them being there. Then the NoA backs off for a moment and tries the swerve action again even though the car is still there next to them. After this happened with 3-4 different vehicles behind me and getting exactly same reaction, I adjusted the NoA from Mad Max to medium as it was making me look like a drunk driver or an idiot. If the car indicates, the time before it starts action should be halved to avoid the car behind you getting fed up with it and pulling up next to you while you still indicate. If you see a car pulling next to you, do not swerve towards them. Those should be the basics the NoA should get. When driving on AP and I want to change lanes, I do the decision making, I ensure there is space to change lane, then indicate and within 1-2 sec I give the wheel a nudge and the car so beautifully completes the lane change, so that tells me it is fully capable of doing a beautiful lane change, just that its decision process is somewhat off, specially when it tries to start the swerve action and then cancels it a few times, that should not happen on the road.
NoA drives like an idiot. No doubt about that. It lacks foresight, planning, and pretty much the only thing it's doing well is monitoring the blind spots.
The long delays between indicating and executing probably are due to Tesla's legal department advising them to delay so they can claim the driver would have had time to intervene *should* something go wrong.
The long delays between indicating and executing probably are due to Tesla's legal department advising them to delay so they can claim the driver would have had time to intervene *should* something go wrong.
I did some NoA during the week on a busy highway (with an eye in the rear view mirror and foot near pedals) to try to work out what was happening with the overtaking problems.
It deciding to change lanes is quite slow, and I agree with @ZeeDoktor about it giving the driver time to check things are safe. If I start a lane change with the indicator then it is quite prompt.
Despite overtaking several large trucks and the occasional car I had none of those profoundly irritating episodes of phantom braking.
I’m wondering if it was because there was in fact several cars around showing that the way was clear and I was not about to crash into some stationary traffic queue.
I’m basing my thoughts on a couple of the previous notes and Tesla experience. First, that the car would include the concept of traffic queues into its model when driving in multi lane traffic. Second, that when the stop light recognition while driving it will follow a car ahead through a green light, but need a prompt not to stop if the light was green in the absence of traffic. Third the driving model seems to be strongly influenced by past critical incidents - eg the Tesla that drove into a truck crossing a highway on a crest and now my car is petrified of crossing traffic even a long way ahead.
Not that this is any particular mitigation for the gap between raised expectations and reality for FSD, but instead to try to understand the car behaviour better so I can know how best to exploit it.
It deciding to change lanes is quite slow, and I agree with @ZeeDoktor about it giving the driver time to check things are safe. If I start a lane change with the indicator then it is quite prompt.
Despite overtaking several large trucks and the occasional car I had none of those profoundly irritating episodes of phantom braking.
I’m wondering if it was because there was in fact several cars around showing that the way was clear and I was not about to crash into some stationary traffic queue.
I’m basing my thoughts on a couple of the previous notes and Tesla experience. First, that the car would include the concept of traffic queues into its model when driving in multi lane traffic. Second, that when the stop light recognition while driving it will follow a car ahead through a green light, but need a prompt not to stop if the light was green in the absence of traffic. Third the driving model seems to be strongly influenced by past critical incidents - eg the Tesla that drove into a truck crossing a highway on a crest and now my car is petrified of crossing traffic even a long way ahead.
Not that this is any particular mitigation for the gap between raised expectations and reality for FSD, but instead to try to understand the car behaviour better so I can know how best to exploit it.
Similar threads
