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

Phantom Braking

This site may earn commission on affiliate links.
The above completely fallacious argument can be distilled to "other automakers have performance issues so it's ok to give Tesla a pass on it."

Also, this statement is patently wrong: "There's nothing for Mazda to "fix"-- occasional false braking is inherent to a basic active cruise control system. That's why this issue can be found in every brand of such a system"
 
One thing to keep in mind with phantom braking is none of this is AEB braking.
I believe what I experienced in my Tesla was indeed phantom AEB. The car slammed on the brakes, locked up the wheels and veered left as it slid on the road, then suddenly righted itself as if nothing happened. This was all caused by a truck passing in the opposite direction. No AP or TACC was active. It was frightening indeed.

So if you didn't believe teslas suffered from phantom AEB, now you have heard at least one story where one did.

Some people want to classify phantom AEB and phantom braking into separate categories, but I'm not sure why. I would say both are examples of phantom braking. In actuality, likely only the former involves the physical friction brake system.
 
I STILL say it's the inability (unwillingness of the programmers) to vary the regen. If the regen was variable and the physical brakes were used some, the Tesla could BEAT the Mazda in smoothness. The Mazda doesn't jerk because when it lets off the gas, it just coasts. Heaven forbid a Tesla could just coast! This is one of the areas where Tesla's stubborness and hubris is on full display.

So what i don't get - how are these super awesome - $250k value - robotaxis going to be remotely successful? What robotaxi rider is going to want anything but a super smooth ride with no jerking 100% of the time in every ride? IMO a computer will never drive better than a person. We live in an analog world that the computer will never understand.
i tend to agree with the first part, i think people feel the braking in tesla's more due to the regen, hell even if you just stop accelerating, not decelerating, it kind of feels like your applying the brakes. as you said in ICE cars when this happens you coast, but even tho its basically the same thing in a Tesla it "feels" different.

the second part tho, once the programing is sufficiently complex/advanced a computer will be able to out perform a human in every category no contest, how soon we get there is really the only debate imho.
 
I think what the regulators are really gonna be concerned about is the trend and the insane uptick in complaints

According to the NHTSA website, just filtering on complaints tagged to Forward Collision Avoidance so might be missing some data if they're classified differently:
  • 2020 Model Y has 17 complaints.
  • 2021 Model Y has 147 complaints.
  • 2022 Model Y has 46 complaints and the year has barely started
People here would likely have suspicions about what's going on
 
I believe what I experienced in my Tesla was indeed phantom AEB. The car slammed on the brakes, locked up the wheels and veered left as it slid on the road, then suddenly righted itself as if nothing happened. This was all caused by a truck passing in the opposite direction. No AP or TACC was active. It was frightening indeed.

So if you didn't believe teslas suffered from phantom AEB, now you have heard at least one story where one did.

Some people want to classify phantom AEB and phantom braking into separate categories, but I'm not sure why. I would say both are examples of phantom braking. In actuality, likely only the former involves the physical friction brake system.

I think they should be classified as completely different things because false AEB activation can happen at ANY time while driving the vehicle. The only way to turn it off is to turn it off every time you get into the car to drive it. The other reason is the design intent is different. AEB is supposed to only activate when a collision is determined to be unavoidable.

The NHTSA has worked with automakers to make AEB a standard feature so all new vehicles have them. As owners of these vehicles we need to do our due diligence to report false AEB activations when we feel as if there was no justification for the activation.

Phantom braking while on TACC/AP only happens while in those modes. Whether its a NHTSA worthy event depends on the severity. How often it phantom brakes is a combination of the design intent, and how well designed/tested it was.

When reporting an event to the NHTSA I don't expect a customer to know the source of the false braking, but the report should include whether they were in TACC/AP mode (or the equivalent for other vehicles).

The overall expectation should be almost zero chance of an false AEB activation event. I've driven a lot of cars, and I've only experienced it once. I felt like it was understandable in the moment so I didn't worry too much about it.

Phantom braking while on TACC/AP is something that I would expect to happen from time to time especially if the manufacture valued safety over convenience. It's a challenging balance to get right. One of the reasons my Jeep never has phantom braking is the system seems quite limited. It doesn't have numerous sources of possible false activation like Tesla does.

The biggest issue I have with Tesla is the lack of customer communication. These things happen and there is no way for us to even know what's going on. The lack of this customer communication means that these things never get fixed until they blow up into big issues.

I immensely enjoy my Tesla, but the company doesn't prioritize fixing glitches. I tend to be someone that finds problems, and reports them. I can be the worlds most annoying customer because I'll go out of my way to prove to a company that their product does not work. I can't do that with Tesla because there is no way to contact them in a way that's appropriate for root cause analysis.

Even getting a map/nav issue fixed is difficult for a customer to do with Tesla, but with Apple or Google its easy.

All these OCD geeks driving Tesla's and we don't really benefit from any of them.
 
  • Like
Reactions: Sporty and Stach
One thing to keep in mind with phantom braking is none of this is AEB braking.

I've never had a false AEB braking in either Tesla I've had even during that "fun" time with 10.3 where it got a recalled for AEB braking.

The average Tesla phantom braking event is a nuisance event.
I think you are mistaken. Like you, I have never experienced AEB phantom braking. I have only experienced the "nuisance" phantom slowdowns that occur under TACC. But it is clear that some Tesla owners have experienced phantom AEB braking that is far more severe that the annoying slowdowns that affect pretty much all of us. I can't tell you how common it is, but it does seem to happen. Or at least it has happened in the past. If the problem has been resolved, that would be great, but I think the jury is still out.
 
Having never experienced Phantom Braking in my vision only Model 3, I would like to get a feel from anyone still not dug into one camp or the other about one thing...Is there a chance that a good percentage of these phantom events are the car registering a lower speed limit from a sign (or a ghost sign)? Additionally, I am assuming everyone who experiences a phantom event is a) reporting it to Tesla service, B) cleaning and calibrating their cameras, and C) Saving their dash cam footage right? Because if I felt my car was doing that, those would be the first three things I would do. Freaking out online would likely fall down the list.

I hear it always described the same way I have seen the car sometimes address a change down in speed limit...not well. But calling it horrendous and dangerous is a hilarious misrepresentation as well.

The other thing I would like to know, is are there any actual documented accidents caused by this?

People slam on their brakes every single day and they don't cause horrific crashes, which is the way people seem to be reacting to this issue. Like it's out there murdering people on the daily.

I'm not saying this isn't an issue, nor it shouldn't be addressed, but I have seen a very small handful of video examples showing this, and the ones I watch I have to watch a few times to really even understand what this horrific stopping event even was, it's almost always a 5-10 mph regen slowdown.
 
  • Like
Reactions: Knightshade
I think you are mistaken. Like you, I have never experienced AEB phantom braking. I have only experienced the "nuisance" phantom slowdowns that occur under TACC. But it is clear that some Tesla owners have experienced phantom AEB braking that is far more severe that the annoying slowdowns that affect pretty much all of us. I can't tell you how common it is, but it does seem to happen. Or at least it has happened in the past. If the problem has been resolved, that would be great, but I think the jury is still out.
When I wrote that post I knew that there was a non-zero chance of false AEB activation, but felt like the conversation was mostly about phantom braking while on TACC/AP.

It's tough to differentiate between the two in a Tesla because lots of us including myself predominantly drive on the highway/freeway with TACC or AP. So we have no idea what would have happened if we were driving manually.

I wish the car had a counter of AEB activations so we would know.
 
So what i don't get - how are these super awesome - $250k value - robotaxis going to be remotely successful? What robotaxi rider is going to want anything but a super smooth ride with no jerking 100% of the time in every ride? IMO a computer will never drive better than a person. We live in an analog world that the computer will never understand.
SURELY you dont think lack of smooth ride is the ONLY thing preventing Robotaxis from being a thing ANYTIME in the coming years. We might want to get past almost running over cyclists and running into green poles in broad daylight first. 🤣
 
  • Like
Reactions: Stach and hatch
Having never experienced Phantom Braking in my vision only Model 3, I would like to get a feel from anyone still not dug into one camp or the other about one thing...Is there a chance that a good percentage of these phantom events are the car registering a lower speed limit from a sign (or a ghost sign)? Additionally, I am assuming everyone who experiences a phantom event is a) reporting it to Tesla service, B) cleaning and calibrating their cameras, and C) Saving their dash cam footage right? Because if I felt my car was doing that, those would be the first three things I would do. Freaking out online would likely fall down the list.

I hear it always described the same way I have seen the car sometimes address a change down in speed limit...not well. But calling it horrendous and dangerous is a hilarious misrepresentation as well.

The other thing I would like to know, is are there any actual documented accidents caused by this?

People slam on their brakes every single day and they don't cause horrific crashes, which is the way people seem to be reacting to this issue. Like it's out there murdering people on the daily.

I'm not saying this isn't an issue, nor it shouldn't be addressed, but I have seen a very small handful of video examples showing this, and the ones I watch I have to watch a few times to really even understand what this horrific stopping event even was, it's almost always a 5-10 mph regen slowdown.

The only way I can see being able to get an idea of the severity of the braking event is if Tesla added either a g-meter or a mph readout to the dashcam.

Video itself doesn't really convey it.
 
  • Like
Reactions: Stach and rjpjnk
The only way I can see being able to get an idea of the severity of the braking event is if Tesla added either a g-meter or a mph readout to the dashcam.

Video itself doesn't really convey it.



FWIW some user a year or more ago in one of the threads actually tested this with a g-meter, and found all his phantom braking was about 0.2g of deceleration.

Which is....basically just regen- no actual "braking"

Which seemed to confirm the issue is less OMG IT IS DANGEROUS and more "People unused to EVs are constantly surprised by any slowdown OR speedup they didn't know was coming because the usual clues from an ICE vehicle aren't present"

Sadly I meant to bookmark the post and didn't, and the search feature is....not great.... so wasn't able to find it again. If anyone does happen to have a link it'd be great to repost it though.


Disclaimer: that doesn't mean this is the value in 100% of cases, a false AEB could well "slam" on the brakes, but that seems vastly more rare on Teslas.
 
  • Disagree
  • Like
Reactions: VanGuy and JustJohn
Having never experienced Phantom Braking in my vision only Model 3, I would like to get a feel from anyone still not dug into one camp or the other about one thing...Is there a chance that a good percentage of these phantom events are the car registering a lower speed limit from a sign (or a ghost sign)? Additionally, I am assuming everyone who experiences a phantom event is a) reporting it to Tesla service, B) cleaning and calibrating their cameras, and C) Saving their dash cam footage right? Because if I felt my car was doing that, those would be the first three things I would do. Freaking out online would likely fall down the list.

I hear it always described the same way I have seen the car sometimes address a change down in speed limit...not well. But calling it horrendous and dangerous is a hilarious misrepresentation as well.

The other thing I would like to know, is are there any actual documented accidents caused by this?

People slam on their brakes every single day and they don't cause horrific crashes, which is the way people seem to be reacting to this issue. Like it's out there murdering people on the daily.

I'm not saying this isn't an issue, nor it shouldn't be addressed, but I have seen a very small handful of video examples showing this, and the ones I watch I have to watch a few times to really even understand what this horrific stopping event even was, it's almost always a 5-10 mph regen slowdown.

Have experience PB regen many times on TACC, but not sure what camp I'm supposed to be in! (personally not 'freaking out', not concerned that it is 'dangerous'; OTOH, da' wife certainly is and does, so I rarely use TACC when she's riding in the car)

a) No way. (Tesla knows about PB and will change it IFF Elon tells them to.)
b) Only when I wash the car, but it is in a closed garage most days
c) No way. (Normally on a highway doing 55+, and have no intention of taking my eyes off the road and fumbling with the bouncing screen to save footage. Tesla can do its own R&D. see a) above)
 
FWIW some user a year or more ago in one of the threads actually tested this with a g-meter, and found all his phantom braking was about 0.2g of deceleration.

Which is....basically just regen- no actual "braking"

Which seemed to confirm the issue is less OMG IT IS DANGEROUS and more "People unused to EVs are constantly surprised by any slowdown OR speedup they didn't know was coming because the usual clues from an ICE vehicle aren't present"

Sadly I meant to bookmark the post and didn't, and the search feature is....not great.... so wasn't able to find it again. If anyone does happen to have a link it'd be great to repost it though.


Disclaimer: that doesn't mean this is the value in 100% of cases, a false AEB could well "slam" on the brakes, but that seems vastly more rare on Teslas.


HA I FOUND IT!

MY KUNG FU IS THE BEST!


the hardest braking measured was under 0.2g, mostly around 0.15g

That's not "dangerous"-- indeed the acceleration force back up to speed right after is actually slightly greater than the braking force.

But as I say could see how it might be annoying when it's not expected.
 
  • Disagree
  • Like
Reactions: VanGuy and JustJohn
Having never experienced Phantom Braking in my vision only Model 3, I would like to get a feel from anyone still not dug into one camp or the other about one thing...Is there a chance that a good percentage of these phantom events are the car registering a lower speed limit from a sign (or a ghost sign)? Additionally, I am assuming everyone who experiences a phantom event is a) reporting it to Tesla service, B) cleaning and calibrating their cameras, and C) Saving their dash cam footage right? Because if I felt my car was doing that, those would be the first three things I would do. Freaking out online would likely fall down the list.

I hear it always described the same way I have seen the car sometimes address a change down in speed limit...not well. But calling it horrendous and dangerous is a hilarious misrepresentation as well.

The other thing I would like to know, is are there any actual documented accidents caused by this?

People slam on their brakes every single day and they don't cause horrific crashes, which is the way people seem to be reacting to this issue. Like it's out there murdering people on the daily.

I'm not saying this isn't an issue, nor it shouldn't be addressed, but I have seen a very small handful of video examples showing this, and the ones I watch I have to watch a few times to really even understand what this horrific stopping event even was, it's almost always a 5-10 mph regen slowdown.
Do Vision-only models from 2021> recognize speed limit changes in a way that’s different from <2020 models? What we need to do here is discern what has changed, because something is clearly very different.

People submitting complaints say they feel the braking is dangerous, will likely lead to an accident, and that they often have stopped using the system entirely because of the issue.

Tesla publishes quarterly stats suggesting that Autopilot usage greatly benefits safety on the roads when the system is engaged and working properly. If people are not using it or are using a slightly defective system, that could conceivably result in accidents that would otherwise not have occurred if it were firing on all cylinders.

These things cannot be reasonably reconciled. If Autopilot prevents accidents, we and Tesla should be all the more eager to have the system working properly. If people dont use the system because the experience sucks, those safety benefits aren’t worth the PDFs they’re printed on.
 
There's a difference between "feels dangerous" and "is actually dangerous"

One of the reasons I wanted to dig up that link to someone who used an actual objective measuring device to measure the braking force experienced.

Humans are notoriously terrible measuring devices.


And no, there shouldn't be any difference in how vision only cars handle posted speed limits. How could there be? Radar never had anything to do with that functionality.
 
  • Disagree
  • Like
Reactions: VanGuy and JustJohn
That's not "dangerous"-- indeed the acceleration force back up to speed right after is actually slightly greater than the braking force.

But as I say could see how it might be annoying when it's not expected.
Disclaimer: that doesn't mean this is the value in 100% of cases, a false AEB could well "slam" on the brakes, but that seems vastly more rare on Teslas.
And with that, I think we are approaching consensus, at least from my perspective. Phantom slowdowns in TACC/AP/NOA occur fairly regularly in Teslas. They are not overly dangerous, but they are annoying to the driver, passengers, and the car(s) behind the Tesla. Tesla should work diligently to eliminate this phenomenon as much as possible so that its TACC is in line with that of other manufacturers.

Phantom braking in AEB collision avoidance - the kind that throws the driver forward with enough force to lock the seatbelt and cause physical discomfort - has been reported by some drivers, but most of us have never experienced it. It is dangerous when it occurs, but it is relatively rare. If it has not been corrected yet (and I don't think we know the answer to this), then Tesla should take it very seriously and fix it.

Here is a report that appears to describe the latter:

I've read a lot of posts about phantom braking but it never really concerned me much because up until today I had only experience minor and infrequent events.

Tonight, however, my car really scared me. It slammed on the brakes to the point that the wheels locked up slightly and the car fishtailed a bit to the left. I was not using either cruise control or Autosteer at the time. Just plain old fashioned driving by hand at 35 mph. It was dark, and a pickup truck was approaching in the opposite lane of a hilly two-lane country highway. We were on a straight portion of the road both heading down toward a low point, and just as the truck passed my Model Y slammed on the brakes and the system started beeping and screaming it's very loud alarm signal. Also an indication popped up on the screen that Dashcam footage was being saved.

Fortunately I was driving slowly and it was easy to regain control, but what the heck?!! That was bad! I assume it was the collision avoidance system that was responsible since TACC and AP were not in use. I believe I have only experience phantom braking when AP was active in the past so this seemed unusual. And I have never experienced braking this hard. Is this normal?

Since I know the exact timestamp and have dashcam footage do you think I should contact Tesla? I've watched the footage a few times and there is absolutely no reason I can see that collision avoidance should have been activated.

Car is a 2021 Vision Only Model Y LR.
 
i tend to agree with the first part, i think people feel the braking in tesla's more due to the regen, hell even if you just stop accelerating, not decelerating, it kind of feels like your applying the brakes. as you said in ICE cars when this happens you coast, but even tho its basically the same thing in a Tesla it "feels" different.

the second part tho, once the programing is sufficiently complex/advanced a computer will be able to out perform a human in every category no contest, how soon we get there is really the only debate imho.
I don't know - you really think so? Maybe a computer can out perform a human in every category except one - acting just like a human? I still say driving just like a human is required. I think Tesla is basically on a wild goose chase with computers being able to safely drive in all scenarios. Never gonna happen IMHO.
 
There's a difference between "feels dangerous" and "is actually dangerous"

One of the reasons I wanted to dig up that link to someone who used an actual objective measuring device to measure the braking force experienced.

Humans are notoriously terrible measuring devices.


And no, there shouldn't be any difference in how vision only cars handle posted speed limits. How could there be? Radar never had anything to do with that functionality.
Do we consider added danger because a negative experience is preventing people from using a system that, according to published stats, produces accidents at a far lower rate per mile than average when active?

This is a bit of a facetious argument, but it can't be reconciled with thinking that no action is warranted here. If Autopilot is much safer than a human driver on average per mile, the system needs to be usable or the stats are meaningless.

If Tesla touts the safety of Autopilot as a way of selling vehicles and then the buyers of those vehicles can't use the system because the experience sucks, that opens up a whole other can of worms. This type of situation is where the NHTSA can be there to protect consumers.
 
Last edited by a moderator:
If Tesla touts the safety of Autopilot in their marketing as a way of selling vehicles and then the buyers of those vehicles can't use the system because the experience sucks, that opens a whole other can of worms.
Ofcourse now everyone can sell all the cars they make - but Tesla has to continuously improve on phantom braking - which we know they are trying to. No easy fixes here ... They are now probably chasing "9s" - no "low hanging fruits" anymore.

ps : We should differentiate between actually phantom braking and slowdowns in the presence of other vehicles, VRUs. Everytime there is a quantifiable probability of crash, they will slow down a bit. Mobileye also does it - and explains it in their planning talk.
 
Yours seems to be an unusual case as the vast majority of phantom braking events reported on TMC are of the more moderate kind. Sure there are a handful of harder ones, but full on AEB type braking is rare (except for that FSD Beta 10.3 fiasco)

Do you have just Basic AP? or FSD?

Have you tried recalibrating the cameras?

I'd do the camera recalibration and then start turning off all the stuff I could think of in the autopilot section to see if it resulted in any improvements.

For FSD owners the Traffic Light response one will cause more phantom braking if enabled.
I have FSD but it seems to occur on AP (i.e. freeways where FSD isn't being used yet), on FSD as well as when I'm only using TACC. Of course there's no 'dumb' cruise control option, so there's no way to tell if it's just a fault with the cruise part of the algorithm or with the traffic aware part.

I haven't tried recalibrating the cameras - that's a good suggestion. I'll look up how to do that and see if it helps. Like I said, I will regularly get minor slowdowns in situations where the cameras shouldn't play a role (straight road, no cars, no lights, no shadows, etc), but who knows.
I STILL say it's the inability (unwillingness of the programmers) to vary the regen. If the regen was variable and the physical brakes were used some, the Tesla could BEAT the Mazda in smoothness. The Mazda doesn't jerk because when it lets off the gas, it just coasts. Heaven forbid a Tesla could just coast! This is one of the areas where Tesla's stubborness and hubris is on full display.
You may be right. We really don't know what's going on in the computer. Is it intentionally decelerates or if there's just an interruption in power delivery? People making the argument that "it's not using the brakes, it's using regenerative braking" are just clueless because it doesn't matter. The whole point of cruise control is to maintain speed and of TACC is to adjust the speed based on what's in front of you. If TACC can't distinguish between the two then it's an even bigger failure.
When I wrote that post I knew that there was a non-zero chance of false AEB activation, but felt like the conversation was mostly about phantom braking while on TACC/AP.

It's tough to differentiate between the two in a Tesla because lots of us including myself predominantly drive on the highway/freeway with TACC or AP. So we have no idea what would have happened if we were driving manually.

I wish the car had a counter of AEB activations so we would know.
The end result of phantom braking and auto emergency braking is the same (unwanted/unnecessary deceleration of the car) but they are two different issues. With AEB there is a blaring alarm and TACC is (or should be) deactivated as part of the event. (the dash cam should be activated as well.) With phantom braking there is a sudden deceleration of varying intensity magnitude followed by a resumption of the previously set speed. It may be related to curves, passing vehicles but frequently has no discernible trigger. False AEB activations can clearly be dangerous. Phantom braking potentially so but generally much less likely to be dangerous. (unless you have a hot cup of coffee and it spills in your lap because the car suddenly slows down.)

The last time I drove my MY for a longer period, TACC would randomly slow down by 2-3 MPH every 10 minutes or so. On several occasions it slowed down 10-15 MPH. There was never a good reason an it always resumed the set speed afterwards. In all occasions it was enough to make us sway forward in our seats and on several occasions woke up the dogs in the back seats.

you're right, many people seem to confuse AEB and PB. I quit reading Knightshade's ramblings but from his/her earlier posts the problems he identified in other makes seemed to be AEB more than issues with TACC.

As I've said, I've personally experienced a handful of false AEB activations. They're pretty uncommon and most often have some identifiable trigger but they're not the post of this thread.