Ironically Elon and FSD team claimed radar was the cause of phantom brakes…
But yea agreed, I have way more freeway phantom braking since having my radar disabled
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Navigate on Autopilot is Useless (2018.42.3)
- Thread starter wk057
- Start date
But yea agreed, I have way more freeway phantom braking since having my radar disabled
While the Waymo Butler Bot will always deliver the drink correctly, but only to one geofenced room in the houseI'm looking forward to a time when people can complain about Tesla Butler Bot Beta spilling coffee on them or taking the drink to the wrong room
Unfortunately unlikely, I have FSD beta with radar disabled and phantom braking on freeway is even worse. Most people without radar will say the same
EVNow
Well-Known Member
Within 3 ft of the refrigerator only.While the Waymo Butler Bot will always deliver the drink correctly, but only to one geofenced room in the house
This is definitely not my experience, and definitely not the experience of anyone else I've personally spoken to about these.
I currently personally own.... hmm... 3 Model S, an X, and two Model 3 .... that are legally drivable, at least (way more if you count not usable). The X and two S are AP1, the other S is AP2.5, and one 3 is AP2.5 and other is AP3. The one AP1 S I use the most has AP1 hacked to never need hands on the wheel and never limit speed (up to the 90 MPH limit at least).
All of the AP1 vehicles far outperform any of the AP2.5 or AP3 vehicles on highway use. Hands down. I know it, the other people who use my vehicles know it, my passengers know it, etc. There's a clear distinction.
To be blunt: AP2.5 and AP3 try to kill us regularly and randomly. Phantom braking is out of control, and their propensity to do dumb things that are completely unpredictable is very high. The phantom braking is so bad on the highway that my wife refuses to even use AP in her Model 3 anymore (which is a clean title, in-warranty, AP2.5, fully unmodified vehicle, for the record). The phantom braking is unpredictable. The dumb decisions AP2+ makes are completely unpredictable. Sometimes it will do just fine in a section, other times it wants to drive off into the median or into a truck. It's barely usable at all. NoA? Forget it. Just adds additional stupidity to the mix.
I'm honestly completely amazed that anyone has anything good to say about AP2/3 given my terrible experience with it on many different vehicles over the years. Like, when people like the above claim it works "perfectly" I'm very inclined to just believe they're lying/trolling/whatever. Like, it just doesn't match reality as far as all of the evidence I have available.
AP1 is far from perfect, especially outside of divided highways. It's not great at tight curves or hills on back roads and such, and honestly, it was never intended to be anyway. On interstate highways though, it will beat AP2 every time on usability. Even in the few places it does get it wrong, those places are completely predictable. It's not that it will fail in a spot or situation occasionally or unpredictably. Nope, if you do a route with AP1 a few times you'll know exactly where it's going to screw up, and it will screw up there every time. AP2 also has situations it fails at every single time, but many many more situations where it can work sometimes, but not always.
People keep bringing up hills and curves. As noted, hills and curves on some back roads definitely mess with AP1. This is well known. But, I've never encountered a curve or hill on an interstate highway that caused AP1 a problem. Do they exist? Maybe. But I live in the foothills of the Appalachian Mountains, hills everywhere... and have not encountered such a problem with AP1. AP2 on the other hand.... heh!
For example, I drove a Model 3 just 40 miles yesterday on the highway (80 total round trip). I had to disengage at least a dozen times for various reasons. That same route AP1 does so repeatable and so flawlessly every single time (like, literally hundreds of times) that if I had to in some emergency situation I'd probably bet my life on it. I would almost be comfortable getting in the car, getting on the highway, engaging AP1, and going to sleep with a timer to wake me up before the exit, and be extremely certain that I would get there without any issue. That's how much better AP1 is on the highway vs AP2. Like in a dire emergency, I know I could get on the highway and let AP1 do its thing without me even needing to monitor if I had to, especially if I know the route.
AP2? Not a chance in hell. I'd be dead for sure. Like, it's not even a dice roll. If so, AP2 somehow rolls a critical miss every time. There's zero chance AP2 would succeed. Heck, even if it was following the AP1 car on the same route I wouldn't trust AP2 to not kill me.
They're just far too different. AP1 is perfect for highway use. AP2/3 just isn't there. It may do better on some non-highway edge cases, but it does nothing better on the highway than AP1.
Edit: Oh, and yeah... the yoke is garbage also... pretty sure anyone who says it's better is just selling themselves on it so they don't seem like such a fool for falling for it. It's objectively worse in many ways, and at best equal in others... then if you add the clunky touch controls and other changes, it's way worse by far. I tried it for several hundred miles of on and off highway driving. I could never own and regularly use a car with that interface. If they made the steering ratio what it should be a for a yoke (180-ish degrees lock to lock) then it'd be GREAT.... but a yoke with a standard lock to lock? That's by far one of the dumbest things Tesla has done... and that's saying a lot.
Last edited:
And yet other people, myself included, have far different experiences. I have HW3 model 3 both without and now with FSD and NoA/AP performance on freeways is pretty much flawless except for the occasional spurious lane change (which it does safety, albeit slowly). I drove my car from Seattle to Portland a few months ago (about 200 miles), and 90% of the time the car was driving on I-5 on NoA, and I did not get a single phantom braking event.
I'm not saying your experiences are not real, but clearly there is some factor that makes your drives so much worse than mine, though I have no idea what it is.
See, you say these things, and yet 15 minutes ago I just drove my AP3 Model 3 to Home Depot and back, which involves 6 miles each way of interstate driving and got no less than THREE panic braking phantom braking events in different spots, with nearly no other traffic on the road. One outbound, two inbound... 12 miles of AP use. Nothing on the visualization to suggest why.
So yeah, I just don't buy it. If it were one car or something, ok maybe something wrong on my end... but it's every single AP2+ vehicle I've ever driven (which would be dozens at this point), including the Plaid.
But, you can't make it from Seattle to Portland though Tacoma without phantom braking while on AP let alone NoA.
It's simply not possible.
Tacoma will mess it up as that's where Navigation related phantom braking occurs.
Maybe that's the 10% of the time it's not on NoA
I've been thinking of creating a $1K challenge.
Where I'll donate $1K to charity of choice of the first person that can go from Lynnwood, WA to Portand, OR and back with NoA without an intervention. The only allowance would be turning it off to take an exit to charge a single time in both directions.
I'd probably have a window of time like 3 months.
Maybe from Jan 1st to March 31st.
Obviously a camera needs to record the drive from the dashboard, and another camera showing any foot action.
Last edited:
There was some major road changes going on at that point and yes I drove manually as neither the car nor I really knew what was going on .. roads were way variant from the maps so I had to follow sign-posted directions carefully. But that's different in my mind to phantom braking, which doesnt occur for me when the car is on well marked freeways.
Again, there seems to be a large variation in the experiences of different cars and/or individuals, and it's not clear to me what is causing that.
Again, I dont see these things. I've used AP to drive me to/from 5 or 6 different locations up and down the freeway this week. At no point did I get any phantom braking, either in quiet or dense traffic. Only this morning the car drove me to/from a mall that was 10 miles away by freeway .. NO phantom braking events at all. In fact it was a boring uneventful drive.
I like the challenge for multiple reasons.
It's during the winter time in the PNW during what is likely going to be the rainiest year on record.
It involves both directions through the cesspool that is I5 on Tacoma
It involves numerous encounters with Semi Trucks (some phantom braking and canceled NoA lane changes are confusion over where the semi's are in the lane)
It involves going north on I5 though Seattle where the nav gets tripped up. NoA is likely going to take you into downtown Seattle when you want to go north.
It involves over 400 miles which is typically enough to trigger at least one or two phantom braking events even on the best version of firmware I've used (the safety score Vision+Radar before the mess that is Pure Vision).
There used to be a glitch north of Seattle where NoA would constantly try to force you out of the left lane even if you clearly wanted to go faster than other people.
I might limit the challenge to only FSD beta people with Pure Vision as pure vision is known to have more phantom braking events. This is part of why we're seeing large variations. The other difference is regional differences where some areas having a lot of navigation related errors, and Tesla has no mechanism for reporting them and getting them resolved. There is probably also person differences on expectations. Like to me 90% is still a failure.
There is also more forgiveness for some behaviors than others. The whole phantom braking really wrecks a persons impression of a system because it can be so alarming. The example of Tacoma I mentioned isn't alarming as I know approximately where it will do it, but I still tense up under overpasses because one time a thousand miles ago it suddenly braked fairly hard while going under one. It was one I went under 100's of times before without issue, but for some reason that that day it decided that the overpass I out to kill it.
To me the biggest problem with AP2/FSD is the inconsistency.
AP1 was fairly consistent. It just did what it did and there wasn't that much variability. There will always be some variability with ADAS systems because the real world is messy. But, AP2/FSD needs to be a lot more consistent.
I generally don't pay too much attention to individual stories, but to rolling averages. Like my experience isn't as bad as WK057 and I think he's a bit of cheat because he hacked his AP1. But, I trust that he experienced what he experienced with AP2 despite the fact that I've never had anything that bad.
With the FSD Beta however my experiences seem far worse than the rolling average so I worry that something is wrong with my car. It's like its not even a beta, but some strange from of therapy where I get out all my bad energy on it.
If I enable TACC in my Model 3 with my fiancée in the car, even for a second to stretch my leg, she gives me a look like "This POS better not hit the brakes". By far, unnecessary braking on the highway is one of the worst behaviors in this system because it's so frequent and so abrupt.
That's the kind of thinking that gets you banned from some places. Point out objective facts, get shouted down, and then 6-12 months later you get to see Andrej confirm what you said on stage at a revolutionary event.
But now you can clearly see the binnacle display. Which was already below the top of the wheel anyway, and which Tesla said was unnecessary when the 3/Y were released.
Yeah, at this point these cars are being designed for Elon, and I'm pretty uninterested in buying another. They almost had me on the Y, but then the heat pump problems happened. Then they hatch problems happened. Then the lack of radar happened. Now, I'm just waiting for the EQE to be released. As more "normal" customers buy Teslas, they're going to learn really fast that people don't have patience for spending $60k on an experiment.
I'm probably just going to have to setup some GoPros in the cars one day and have at it.
First a highway drive on AP1 in my hacked car, then the same leg in a 100% stock AP1 car (will be the same result, I'll just have to hold the wheel), then the same leg in multiple AP2/3 vehicles, then for fun the same in AP2/3 vehicles with NoA enabled.
There will be a CinemaSins style sin counter in the corner each for AP1 and AP2+ for every time I need to do something to prevent it from doing something dumb or to recover from it doing something dumb. If AP1 gets a single ding I'd be surprised.
Here's the route: Google Maps ... a pretty mundane stretch of interstate highway. There is a construction zone around the I-77 junction, and another around mile 150-158 where they've been resurfacing one side, occasionally cattleshooting to the other side of the road. The markings in all case are good and unambiguous.
Who wants to place some wagers who the winner will be? Welcome to bring your own vehicle!
Spoiler: I do this weekly or more, and AP1 has no chance of being defeated.
First a highway drive on AP1 in my hacked car, then the same leg in a 100% stock AP1 car (will be the same result, I'll just have to hold the wheel), then the same leg in multiple AP2/3 vehicles, then for fun the same in AP2/3 vehicles with NoA enabled.
There will be a CinemaSins style sin counter in the corner each for AP1 and AP2+ for every time I need to do something to prevent it from doing something dumb or to recover from it doing something dumb. If AP1 gets a single ding I'd be surprised.
Here's the route: Google Maps ... a pretty mundane stretch of interstate highway. There is a construction zone around the I-77 junction, and another around mile 150-158 where they've been resurfacing one side, occasionally cattleshooting to the other side of the road. The markings in all case are good and unambiguous.
Who wants to place some wagers who the winner will be? Welcome to bring your own vehicle!
Spoiler: I do this weekly or more, and AP1 has no chance of being defeated.
I'm not sure I see any purpose in this challenge. No-one is disputing that phantom braking is a thing, so what does your long drive, which is carefully chosen to maximize the probability of phantom braking occurring, achieve?I like the challenge for multiple reasons.
It's during the winter time in the PNW during what is likely going to be the rainiest year on record.
It involves both directions through the cesspool that is I5 on Tacoma
It involves numerous encounters with Semi Trucks (some phantom braking and canceled NoA lane changes are confusion over where the semi's are in the lane)
It involves going north on I5 though Seattle where the nav gets tripped up. NoA is likely going to take you into downtown Seattle when you want to go north.
It involves over 400 miles which is typically enough to trigger at least one or two phantom braking events even on the best version of firmware I've used (the safety score Vision+Radar before the mess that is Pure Vision).
There used to be a glitch north of Seattle where NoA would constantly try to force you out of the left lane even if you clearly wanted to go faster than other people.
I might limit the challenge to only FSD beta people with Pure Vision as pure vision is known to have more phantom braking events. This is part of why we're seeing large variations. The other difference is regional differences where some areas having a lot of navigation related errors, and Tesla has no mechanism for reporting them and getting them resolved. There is probably also person differences on expectations. Like to me 90% is still a failure.
There is also more forgiveness for some behaviors than others. The whole phantom braking really wrecks a persons impression of a system because it can be so alarming. The example of Tacoma I mentioned isn't alarming as I know approximately where it will do it, but I still tense up under overpasses because one time a thousand miles ago it suddenly braked fairly hard while going under one. It was one I went under 100's of times before without issue, but for some reason that that day it decided that the overpass I out to kill it.
To me the biggest problem with AP2/FSD is the inconsistency.
AP1 was fairly consistent. It just did what it did and there wasn't that much variability. There will always be some variability with ADAS systems because the real world is messy. But, AP2/FSD needs to be a lot more consistent.
I generally don't pay too much attention to individual stories, but to rolling averages. Like my experience isn't as bad as WK057 and I think he's a bit of cheat because he hacked his AP1. But, I trust that he experienced what he experienced with AP2 despite the fact that I've never had anything that bad.
With the FSD Beta however my experiences seem far worse than the rolling average so I worry that something is wrong with my car. It's like its not even a beta, but some strange from of therapy where I get out all my bad energy on it.
A number of people have posted here that they see PB events in much calmer, non-challenging conditions, which is not my experience, nor of others here also. The issue then, is not to prove that it can happen (that is not in dispute), but to try to understand what variable or variables cause such disparate experiences.
Demonstrating Phantom braking isn't the objective. Instead the objective is to shine publicity on the need for Tesla to straighten up Navigation because NoA will never work without Navigation being fixed.
There needs to be a mechanism with feedback for customers to be fully involved in alerting Tesla of issues with the navigation.
I have no idea why Tesla hasn't implemented this mechanism.
With Apple I can report something, and a few days later they send me a notification saying its fixed. It's almost like they're inducing me to fix every problem I see because there is that satisfaction of solving a problem.
Where with Tesla I don't bother because what's the point. It just goes into the ether.
Having what should be a fairly easy task be unachievable would be a good way to shine light on it.
It's like the opposite of the Kobayashi Maru.
I'm happy to go make a 50 mile drive on AP2. Provided I head south instead of north from here, I expect no issues. If it's north, I know exactly where the issues will be.
Unless it's night and raining, then **** AP2 because it will suddenly red hands me for no reason.
Okay maybe it's not great.
It’s very likely that the phantom braking is a function of certain flavors of AP2 (depending on what hardware and model) and the more remote geographic region (e.g., certain areas of North Carolina and Tacoma).
I’m in the DC area which is likely one of the highest concentration of Teslas outside of California. In the old days of AP1 I routinely did bug reports on many phantom braking locations (e.g., that last overpass on 66 in virginia just before the bridge to DC was a reliable phantom brake event) and eventually the phantom braking stopped on AP1. MAybe all those places are mapped out of existance in the higher density areas.
In my use of AP2 in two different Model 3s in the family (one with FSD but not the beta button, and one with EAP), and now mostly in my Model S refresh, I never experience phantom braking in the usual highway routes around the DC area Maryland and Virginia. I’ve also already driven to and in Manhattan and used AP up and down 95 in between with zero phantom braking.
Maybe phantom braking is now an artifact mostly in the less Tesla dense areas. Most of my driving is up and down the 95 corridor so I’ll not be able to test this very well but even plenty of country driving in Maryland and Virginia I just haven’t experienced any phantom braking.
I will agree with the very original post is that I don’t have a use case for NOA — it is just too slow and I’m too impatient. I sometimes do turn it on if I think I might not pay attention and miss an exit and it will reduce that risk.
And I rather miss my ability to do highway AP with one hand on the wheel controlling all functions of AP with my left hand on the old stalks on the model S. Moving the speed to right scroll wheel and following distance to the screen is a great annoyance compared to at lefthand fingertips on the AP ministalk.
Although I recommend not defeating the hand on wheel test as it seems safer to keep one hand on the wheel to both sense any wheel movement directly through your hand being on the wheel and also being able to instantly respond as needed with any steering input needed.
I’m in the DC area which is likely one of the highest concentration of Teslas outside of California. In the old days of AP1 I routinely did bug reports on many phantom braking locations (e.g., that last overpass on 66 in virginia just before the bridge to DC was a reliable phantom brake event) and eventually the phantom braking stopped on AP1. MAybe all those places are mapped out of existance in the higher density areas.
In my use of AP2 in two different Model 3s in the family (one with FSD but not the beta button, and one with EAP), and now mostly in my Model S refresh, I never experience phantom braking in the usual highway routes around the DC area Maryland and Virginia. I’ve also already driven to and in Manhattan and used AP up and down 95 in between with zero phantom braking.
Maybe phantom braking is now an artifact mostly in the less Tesla dense areas. Most of my driving is up and down the 95 corridor so I’ll not be able to test this very well but even plenty of country driving in Maryland and Virginia I just haven’t experienced any phantom braking.
I will agree with the very original post is that I don’t have a use case for NOA — it is just too slow and I’m too impatient. I sometimes do turn it on if I think I might not pay attention and miss an exit and it will reduce that risk.
And I rather miss my ability to do highway AP with one hand on the wheel controlling all functions of AP with my left hand on the old stalks on the model S. Moving the speed to right scroll wheel and following distance to the screen is a great annoyance compared to at lefthand fingertips on the AP ministalk.
Although I recommend not defeating the hand on wheel test as it seems safer to keep one hand on the wheel to both sense any wheel movement directly through your hand being on the wheel and also being able to instantly respond as needed with any steering input needed.
Last edited:
But the point is that NoA is working for many people .. myself included. It's not clear to me what is causing this disparity, nor what percentage of people who use NoA experience signifcanr phantom braking. (You can't judge it by posting frequency here, for obvious reasons.) Until someone can figure out that percentage, nothing much is going to get accomplished.
Again, last night I drove back from the grocery store (Metropolitan Market in Kirkland), which had FSD beta drive me to the freeway, NoA drive me up I-405 to 522, and then FSD drive me to my house. The only intervention was on the horrid on-ramp to I-405 which is so badly designed I would not expect any car to manage it (humans have trouble too), especially at night. I got one instance of medium-aggressive braking (maybe 60 mph to 50 mph) during the merge from I-405 to 522 when an SUV started to merge over and the car slowed down fast to make sure it wasn't in a collision path .. not what I would call a "phantom" brake.
I'm not saying any of this to sound smug, or to say that others are imagining things, but the title of this thread is "NoA is useless", which is clearly not true for many people.
Similar threads
