The only things I can think of as phantom braking from my side are things like the car slamming on the brakes when approaching certain manhole covers, asphalt patches which are black while surround asphalt is gray, and various other cases. Some of these areas happen at the same spots regularly.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
FSD Beta 10.13
- Thread starter Todd Burch
- Start date
The only things I can think of as phantom braking from my side are things like the car slamming on the brakes when approaching certain manhole covers, asphalt patches which are black while surround asphalt is gray, and various other cases. Some of these areas happen at the same spots regularly.
sleepydoc
Well-Known Member
The term ‘phantom braking’ includes false AEB activations as well as the random 5-10 MPH slowdowns that occur on a regular basis. I suspect the NHTSA is far more concerned with false AEB activations as these are much more likely to be a safety issue. Fortunately they are also much less common.
This is not to justify or excused the other PB events that continue to make Tesla’s TACC worst in class, just to clarify that in general they are not safety concerns and likely not of interest to the NHTSA.
On the return segment of a recent road trip I had repeated Autopilot PB problems while heading westbound on I-80 through the gentile mountain grades of Nevada.
The problem appeared to be due to so-called “inferior mirages” on the hot baked highway which appeared, due to an optical illusion, to be flooded with water in the distance.
This would commonly happen as I was driving through a mildly hilly area which caused the mirage to reliably appear as I began descending into an area with another upcoming hill.
The problem appeared to be due to so-called “inferior mirages” on the hot baked highway which appeared, due to an optical illusion, to be flooded with water in the distance.
This would commonly happen as I was driving through a mildly hilly area which caused the mirage to reliably appear as I began descending into an area with another upcoming hill.
AlanSubie4Life
Efficiency Obsessed Member
What was the magnitude of the speed loss and over what timeframe? Everyone really needs to start running GoPros I guess! Very hard to know in the moment. In my experience, passengers (other than my wife) are actually less bothered by the braking than the driver is. That suggests the actual velocity change is relatively small (assuming the driver intervened after less than 1 second of course). But maybe it depends on the passenger!
Does anyone know of any good YouTube demonstrations of actual phantom braking? I haven’t seen any (false AEB events due to the firmware problem in FSBb excluded).
Add shadows cast onto the roadway by trees in the median of interstates to the list
diamond.g
Active Member
I think they stopped doing rapid updates after NTHSA got involved with the AEB bug. Or at least it seems like that is when the updates slowed down.
I don't count the infuriating speed stutters that current FSDb & AP suffer from as Phantom Braking.
However, when they occur with monotonous regularity for no apparent reason they turn AP from being a useful driver assistant into a useless gimmick.
Yes, you could drive along with your foot covering the go pedal - but you shouldn't have to.
Then when it does decide to lift off, the sudden deceleration from 75 to 65 throws occupant heads forward, spills drinks and anything not strapped down slides forward.
To avoid being rear ended by the car behind who is still doing 75mph, you put your foot on the pedal to get back up to speed and all those occupant heads go bouncing back again.
Repeat every 5-10 miles and said occupants will beat you around the head until you stop using AP with them in the car.
Also don't feel the need to video these things - the fact that the family is back to not wanting me use AP is enough for me.
Current FSDb is worse than any previous release than I can remember for this stuff and they don't seem to be focussing on it at all.
I don't care about Chucks unprotected left - they need to fix basic stuff first. The inability to drive down a road at speed without constantly making passengers car sick makes it unsuitable for use.
that is an insane amount of phantom braking. I have an 80 mile round trip commute, most of which is on a major interstate artery. I recently had a phantom braking event. The one before that was around FSDb 10.5. Thousands of miles without an issue.
The amount of variability with PB is really strange, esp since interstates are somewhat similar across states.
Prior to beta there were several consistent sections of freeway where I had inexplicable slowdowns. The interesting thing is that after getting on beta these same areas that I classified as freeway are now switching into the city streets for a brief moment. Maybe when we get on the full stack these will get better.
And it may even be map data. This is the section I mentioned. They show this as an exit on the map but in reality, this is just a split. It is 2 lanes going east not one as listed. When driving there is not even a jog in the lanes it is straight all the way through.
sleepydoc
Well-Known Member
Others have hypothesized that PB events may be due to GPS/mapping errors. The problem with this theory is that, at least with FSD, when it enters an area with a lower speed limit the slow down is so gradual as to be imperceptible. It’s gradual enough that I will actually disengage and manually slow the car down so I don’t get tagged for speeding.
But, Tesla's data collection is so massive they could train a neural network to tell a driver to stop picking their nose.
What is your route through that area? I drive through there at least monthly but I'm usually eastbound and in heavy evening traffic.
What's so amazing is he can't fathom that there could be a huge engineering team working on state of the art voice recognition and voice to text (DeepMind, Google Brain, etc) using the data provided that has nothing to do with the different team that works on using abstracted data for marketing. Like if Apple, Google, Amazon, MS just wanted your data to market to you, they wouldn't send raw audio, just transcripts.
What's even more funny is that there are over 3 billion android devices. Over 50 million Google home devices sold. But ofcourse Hail Tesla!
yep 281 from Hico, TX to Lampassas - 130 mile total drive one way, averaging one slowdown every 5-10 miles.
its a clear, well made road, almost no traffic
to be clear - its not braking, its slowing down quickly, like lifting your foot of the gas when going uphill.
Slowing down from 75mph to 60 or 55 when it isn't gradual is really uncomfortable. If it isn't a comfortable ride then what's the point of letting it manage speed?
My AP average has been dropping consistently over the last few months as I'm finding it easier and more relaxing than using AP, which is just crazy.
I take 694 eastbound through that all the way to 35W. It always had a hicup for me in that area.
sleepydoc
Well-Known Member
This is what I experience as well with the same frequency. We drove our 2020 Forester up to North Dakota a few weeks ago and it was so nice to have adaptive cruise that just worked! I also rented a Camry a few weeks ago and the adaptive cruise worked perfectly on that, too. The subaru uses vision for its cruise but I believe the Camry used radar.
I'll try and pay attention the next time I go through there and try to update if there's any problem with my car.
KArnold
Active Member
Not sure why, but I'm an outlier on PB.
I was in the "June" deployments and only have 1,500 miles (mostly city traffic) so far using FSDB on every drive but I have not had a single PB incident. I've been on all kinds of roads and see many manual interventions - lane choice, wrong turn signal, UPL, UPR, and so on - but it's beyond me why no PB at all. I've even gone out of my way to find 2-lane twisty roads that 18-wheelers tend to use (OH-315 north of I-270 for those of you familiar with central Ohio), but even that did not produce any hits. Our roads in the area for the most part are well maintained/marked, FWIW. I'll also admit I don't often drive after dark or in the rain though. Very curious as to why some seem to get it constantly and I'm having trouble finding one. Thank God for small miracles.
I was in the "June" deployments and only have 1,500 miles (mostly city traffic) so far using FSDB on every drive but I have not had a single PB incident. I've been on all kinds of roads and see many manual interventions - lane choice, wrong turn signal, UPL, UPR, and so on - but it's beyond me why no PB at all. I've even gone out of my way to find 2-lane twisty roads that 18-wheelers tend to use (OH-315 north of I-270 for those of you familiar with central Ohio), but even that did not produce any hits. Our roads in the area for the most part are well maintained/marked, FWIW. I'll also admit I don't often drive after dark or in the rain though. Very curious as to why some seem to get it constantly and I'm having trouble finding one. Thank God for small miracles.
Having gone all year without PB events until a recent cross country trip, I was surprised at how many PBs I experienced. The only thing I could see on my trip that could cause the PBs was that most, if not all, of the events occurred when there was no lead car anywhere close and most events occurred at the crest of a hill. My suspicion is that the PBs were a result of heat reflections that might have been misinterpreted as another vehicle, or large obstruction. If the reflection patch is moving toward the car, it might think that there is something rapidly approaching.
Forget about the old bridge shadow claim. On my route there were almost no bridges crossing above the roads.
Driving around the Dallas area, I rarely get isolated from other vehicles, so maybe the heat reflections are less frequently seen? Just a hypothesis.
Similar threads
