Phantom Braking Measured

[animorph]

Had mine only 2 weeks, but have had this happen once while exiting the freeway on NOAP. The car gave a few sudden brakes with nothing in front of me but plenty of road. Not sure why. I suspect the computer errs on the side of paranoid braking rather than ignoring a signal that could be noise. Until the AP computer is able to improve signal to noise, this is probably the preferred behavior.

When taking freeway exits NoA may indeed slow down intentionally. You can see the Max Speed indicator on the display actually change be 5 MPH at a time. If it's a transition to another freeway, the speed should automatically return to your setting when you have merged onto the new freeway. If it's an exit to city streets NoA will slow down quite a bit and turn itself off, leaving just standard AP.

 

[Daniel in SD]

Generally it's why if someone hits you from behind it's going to be their fault- because they did not leave enough room to be able to brake their own car in time if the one in front of them braked for whatever reason.

If I slam on my brakes on the highway, in traffic, for absolutely no reason with max summer performance tires on my car I'm pretty sure I would be rear ended 90% of the time. I would argue that it's "reasonable and prudent" to expect cars in front of you not to do hard braking for no reason. However, with the type of braking that OP measured I would agree that if someone hit him it would be their fault.

 

[Knightshade]

"Reasonable and prudent" is a horrendously ambiguous term that is not going to stand up in any court of law when a vehicle brake checks for no reason, whatsoever.

It'll stand up fine, because the guy who hit his brakes in front of you isn't the one on trial- the person who was following too closely and rear-ended him was.


"Too closely to stop" is not reasonable and prudent.

Defining "too closely" requires quantifiable measures (i.e. distance in units such as feet or car lengths), not ambiguous ones that are only qualitative and judgement calls.

I mean- pretty clearly it does not- since "reasonable and prudent" is the law in most US states, and has been for years or decades depending on the state. And it's widely and generally enforced as such.

Thus leaving it to the judgement of the officer issuing the citation, and the prosecutor handling the case, and the judge deciding it, seems to suffice just fine without more specific quantifiable measurements.


A great way to not have to worry about this is- don't tailgate people. Follow at a distance that would allow your car to stop without hitting the guy in front of you if he slamed on his brakes.

Especially in states where deer are known to just sprint across even interstates and the guy in front of you might slam on his brakes at any time.

And the perspective that leaving enough space to always avoid an accident if being brake checked is reasonable and prudent does not come from those who live and drive in bigger cities such as Atlanta (where I learned to drive).

That's adorable.

I learned to drive in NY.

Still have managed to not rear-end anybody, either when I lived/drove there.... or LA... or any of the other bigger-than-atlanta cities I've driven in.....because I don't follow too closely.

Keeping the car length for every 10MPH rule of thumb will keep a driver constantly slowing as others merge into the space. And more and more studies are showing that the slower drivers are the ones who cause more accidents on freeways, though they often are not even involved.

And yet, again, the facts we actually have disagree with your opinion.

For example-

In the 1st quarter of 2019 Tesla registered one accident for every 2.87 million miles driven in which drivers had Autopilot engaged.

For those driving without Autopilot, they registered one accident for every 1.76 million miles driven.

Meaning a higher decel rate, then add that 0.3 seconds for the car behind it, and it brakes even harder. It is the accordian affect, and if traffic is heavy, there WILL be a collision.

The data shown in the original post is not a "high decel rate though" so blaming AP for something it's not doing seems odd.

Anybody hits you with the decel rate shown- that's their fault and they were following way too closely.

I've been pulled over for following too closely, actually.

That ain't surprising

At the end of the day, this all boils down to one thing that drivers depend on from other drivers - predictability.

That'll be the nice thing with FSD- especially as it comes to a lot of cars... it'll be way more predictable than humans who can't even agree among each other how to drive.

The TACC/AP often behaves in ways that are very unpredictable.

So do humans.

From the (admittedly thin) data we have so far though- AP is doing a better job than humans are avoiding accidents.

However, with the type of braking that OP measured I would agree that if someone hit him it would be their fault.

Exactly.

If the data from the OP is what they mean by "phantom braking" it's an utter non-issue unless someone is directly tailgating you at insanely dangerously close distances.

 

[anticowboyism]

If the data from the OP is what they mean by "phantom braking" it's an utter non-issue unless someone is directly tailgating you at insanely dangerously close distances.

But a human driver could take into account that someone is tailgating and adjust their stopping speed to avoid a collision. This is the definition of "defensive driving". An advanced auto-pilot should be able to do so as well.

 

[Daniel in SD]

But a human driver could take into account that someone is tailgating and adjust their stopping speed to avoid a collision. This is the definition of "defensive driving". An advanced auto-pilot should be able to do so as well.

I agree but they also have to behave in a predictable fashion. I think that autonomous vehicles should reduce the total number of accidents not just get in fewer "at fault" accidents.
Why People Keep Rear-Ending Self-Driving Cars
Apparently 57% of autonomous vehicle accidents in California were rear end accidents last year. Though they don't say what the rates are relative to human drivers.

 

[Msjulie]

"But part of being a good driver is behaving in a way others expect, which doesn’t include constantly stamping on the brakes."

Predictability is a key.. The inattentive driver following an AI car that random brake checks, safe following distance or not, can react inappropriately to the condition out of surprise. Nothing good comes from that..

 

[Wingsy]

I don't understand why some of you are willing to give Tesla a pass on this. Phantom braking is dangerous and abnormal. When my car brakes hard (and yes, 0.15G IS hard braking) for absolutely no reason (daytime, clear road) then the obvious conclusion is it shouldn't have done that. It isn't being conservative, safety-wise, it's being confused. I know it can do much better and that's my beef -- Tesla isn't putting enough resources in getting this problem fixed. They are wasting their time with fart apps, games and dog comfort. How do I know it can do better? Because I drove a car with adaptive cruise for 7 years and the only times it phantom braked was when I was overtaking and beside a semi, about once every 6 months or so. Very seldom. Not every 30 minutes. Not only that, it was smooth when changing speeds. I don't have that any more. I don't like it and I'm saying so.

 

[Wingsy]

This example isn't hard braking but it is abnormal braking. It appears that TACC locked onto the van on my right and matched its speed as the van exited the interstate. It slowed from 70 to 60 as the van went up the exit ramp, only to resume speed once TACC lost track of the van (as you can see by the disappearing ghost image of the van). WAY off to my right, as I tried to show in the 2nd attachment below.

 

[HopeToGolf]

Phantom braking is a problem.

With 15K miles logged in the last year and a daily commute of only about 15 miles round trip, I have taken many long trips with EAP engaged. Phantom braking is so scary for me at highway speeds that I change my driving behavior. If traffic is tailing me too closely, I will either pull over to the next lane or disengage EAP. I am very concerned about getting rear ended at 79 MPH when the Phantom Braking happens when a tailgating driver doesn’t expect the vehicle to reduce its speed by a bunch (15 mph?).

I report it EVERY SINGLE TIME by engaging the voice command and pushing that right scroll wheel and saying “Bug Report - Phantom Braking.” I encourage everyone to do the same.

This issue is probably my biggest gripe with my Tesla.

 

[Enginerd]

The thing I find interesting about the plot in post #1 is the recovery behavior. I've noticed the same characteristic when TACC slows for someone pulling out of your driving lane. Once the offending car clear (and I mean really clear), the car gives an initial acceleration pulse, then pauses, then slowly accelerates back up to the set speed. Why does it do this two-pulse acceleration? It's clearly 2 facets of the control logic that aren't well blended. But why hasn't Tesla blended them yet?

 

[Knightshade]

I don't understand why some of you are willing to give Tesla a pass on this. Phantom braking is dangerous and abnormal. When my car brakes hard (and yes, 0.15G IS hard braking).

I don't think it is.

Standard regen alone can do 0.2gs and this isn't even that much.

That's without using the friction brakes at all

So 0.15g is basically 25% less braking than the amount of braking the car does simply by taking your foot off the accelerator at speed- not touching the actual brake pedal at all

If that was "dangerously hard braking" then regen would cause a ton of accidents, but that's simply not the case.


Stand on the brake pedal and it'll brake around 1.0g so I'm not sure how 0.15 is "hard braking" to anybody who's ever actually used the brake pedal.

 

[Msjulie]

If that was "dangerously hard braking" then regen would cause a ton of accidents, but that's simply not the case.

Say I walk up to you, facing you, in a moderately crowded room with people talking all around and say Hey, How are you? Will you be startled (outside the WTF who is this person)?

Now say I approach you from behind in a quiet library, you're reading a book. I tap you on the shoulder and say the same Hey.. would you be startled?

The worst bit about unpredictable behavior is you don't see it coming...

 

[Knightshade]

Say I walk up to you, facing you, in a moderately crowded room with people talking all around and say Hey, How are you? Will you be startled (outside the WTF who is this person)?

Now say I approach you from behind in a quiet library, you're reading a book. I tap you on the shoulder and say the same Hey.. would you be startled?

The worst bit about unpredictable behavior is you don't see it coming...

.... that is... a fairly terrible analogy....

When I'm reading in a library by myself I'm actively reading a book and since I'm by myself not expecting to be interrupted by someone. Something both unexpected and entirely outside that focus might, indeed, surprise me.



When I'm driving a car I'm actively paying attention to the car in front of me (and all the other cars/people/whatever on the road).

Not only that, the car in front of me changing speed (or lanes) is an entirely possible and reasonable thing that might happen that I'm looking out for.



And lastly, you appear to have totally glossed over the point that the braking described here is actually less than the amount you get with no AP engaged at all if you simply take your foot off the accelerator and regen. It's nowhere near "hard braking"

If someone hits you when slowing at 0.15g they were dangerously tailgating you, and their insurance will be covering your repairs and your rental car

 

[Msjulie]

When I'm reading in a library by myself I'm actively reading a book and since I'm by myself not expecting to be interrupted by someone. Something both unexpected and entirely outside that focus might, indeed, surprise me.

You proved my point thanks

 

[Daniel in SD]

I, for one, welcome our new robot overlords.
I now expect the unexpected when driving my Tesla (also from other drivers when driving any car). What makes me the most nervous is when the car doesn’t brake as soon as I would like. The random taps of the brakes make me think it’s paying attention and just being overly cautious

 

[Knightshade]

You proved my point thanks

Was your point that your analogy is terrible and no bearing on driving? Because that's the only part of it I proved


unlike in a library when I am driving a car I'm actively focusing on other cars, not a book and I am expecting other cars to potentially change speed/lane/etc.

So one doing so is not "a surprise" at all. It's explicitly behavior I expect might happen and am actively monitoring for.

So 100% unlike your library example.


Again- the braking actually measured here is less than simple standard regen braking.

It's basically the same as a manual transmission car (or some more modern autos) coming off the gas pedal and engine braking.

If that (or the example in the measured post) cause a rear-end- that's the fault of the guy driving dangerously close behind, not the guy who barely slowed down.

 

[Msjulie]

You substantiated my claim that behaviors out of the norm are startling regardless of what they are... phantom braking, cars steering of their of volition, people surprising you in a library..

Regen vs braking force irrelevant.

 

[Knightshade]

You substantiated my claim that behaviors out of the norm are startling regardless of what they are

Yes, and a car in front of you while driving slowing down a little bit is not out of the norm

hence why your analogy does not work at all.

Regen vs braking force irrelevant.

It's extremely relevant- it debunked the idea this is "hard braking"

it's actually less braking than simple standard regen can do- and nobody is complaining that is some kind of safety risk and demanding they get a setting to turn it off for safety.

 

[Msjulie]

Yes, and a car in front of you while driving slowing down a little bit is not out of the norm

hence why your analogy does not work at all.

We shall have to agree to disagree since what I am describing, PHATOM braking, is not a car in front of you... so yeah it is very much out of the norm, hence the name and I think my analogy of unexpected things was great if I do say so myself.

 

[Lunares]

We shall have to agree to disagree since what I am describing, PHATOM braking, is not a car in front of you... so yeah it is very much out of the norm, hence the name and I think my analogy of unexpected things was great if I do say so myself.

Your analogy is awful since it isnt even from the right perspective. The analogy should be from the tailgating car perspective, not the phantom braking car.

@Knightshade is completely right. 0.15G braking is nothing. It is completely normal for any car on the road to have a braking speed of that much. If I were to rear end a car that brakes like that, it is completely my fault for being too close. Because again, many cars on the road (not just Tesla or autonomous cars) have normal reasons to do that.

If the car was doing full 1G braking of a complete pedal stomp, that's abnormal. It's not.