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

FSD is unavailable now due to error in defeat device detected

This site may earn commission on affiliate links.
Searched TMC as soon as I got home from a 1000 mile trip to California and back.
I got 3 strikes for non existent defeat device. It's hard to "Remove defeat device" when it is your left hand holding the wheel at about 8:30.
Looks like I'm not alone.
FSD 2022.44.30.5, M3p, steering set to Sport.
 
  • Like
Reactions: MTOman and Chisale
More details on exact conditions (vehicle, light, hand position, Road scenario, etc.) would be interesting. Safely taken video too since it seem reproducible.
I don't put my hands at 10 -2 like you're showing in your picture. I've found that balances the wheel, pulling down evenly and then you get more nags since there's no net torque. I usually have some combination of left or right hand at 4 or 8 o'clock and gripping with just enough tension for it to detect weight. Sometimes that works for tens of miles and then out of the blue, nags and cheat device detected warnings. I have never attempted to use a cheat device nor will I ever do something like that.
 
I don't put my hands at 10 -2 like you're showing in your picture. I've found that balances the wheel, pulling down evenly and then you get more nags since there's no net torque.

I used to use both hands with them braced such that one was pulling steadily and the other lightly rested on the wheel.

However, I have found that this does not seem to eliminate nags (it may have at some point but certainly does not now).

Specifically, if you are already applying torque, applying torque does not always clear the nags. Sometimes (but not always), it seems to require torque in the other direction. In any case the nag will often come up and then clear on its own, and at other times you have to wiggle the other way.

Anyway, requires more study to confirm how it actually works. I’ve done some looking and recording but tough to figure out exact pattern.

But in the end I find just driving normally, gently torquing back and forth according to road curves, results in no more nags than my prior approach (and it seems easier to clear them since I am free to torque either way).

It’s silly that nags are still required as the cabin camera data can definitely be used to ensure that hands are on the wheel (even though it can see neither). Some day.

I think the defeat detection may have changed how constant torque works in terms of clearing nags (which, if so, would be because they use an extremely crude method of defection).
 
Last edited:
  • Like
Reactions: Chisale
I think the defeat detection may have changed how constant torque works in terms of clearing nags (which, if so, would be because they use an extremely crude method of defection).
The only non-crude method of detection would be to put proper capacitance sensors in the steering wheel.

Back when folks first claimed that they would be able to detect the difference between a weight attached to the wheel and an arm attached to the wheel, I said that it would never work correctly. What we're seeing now has pretty much proven me correct. Most people drive with their thumbs hanging on the cross bar, give or take, which is exactly where steering wheel weights would have to attach. The torque on the wheel from either one will always be basically the same.

About the only thing you could maybe do would be detecting that the weight shifts to pulling in the opposite direction when some specific point shifts past the midline, except that people's arms also do that while driving up to the point where they move the arm. So unless you go for ten minutes without ever seeing any torque at all except in the direction that would indicate a weight on a particular side, *maybe* you could conclude that there is probably a defeat device, but even then, you'd have to establish a consistent pattern of the excess mass going much farther past the center than is reasonable, while behaving in a manner that makes no sense based on the road conditions repeatedly (e.g. pulling the wheel in a direction that tugs the car closer to a barrier instead of away from it like a human would do). And if you get to that level of accuracy in knowing what an actual human would do, at that point, you likely have already achieved at least L4 self driving, making the defeat device detection largely moot. :(

As for me, a long time ago (non-FSD), I got to the point where if I get annnoyed enough at it flashing at me while driving with both hands on the wheel, I jerk the wheel hard enough to kick out Autopilot, then reengage. That usually pacifies it for a while. I assume it still will.
 
The only non-crude method of detection would be to put proper capacitance sensors in the steering wheel.

There are plenty of other ways, like analog detection of torque as you describe - and using the cabin camera to figure it out (which it could do, without being able to see the hands or the wheel). A combination of these things has a very distinct pattern.

The torque on the wheel from either one will always be basically the same.

They are not the same at all. As long as they can measure different amounts of torque it should be easy to distinguish (pattern match).

Another easy way is to look for both hands and also look for sustained torque on the wheel when both hands are on the wheel. That is a very good indicator of a high likelihood of a defeat device. Except for those who are able to put sustained torque on the wheel with their knee (a bit tough I think, though possible, and basically unnecessary in a Tesla). Won’t see this situation with most users (seeing both hands would be pretty rare), but it is just one way to get people on a naughty list.
 
There are plenty of other ways, like analog detection of torque as you describe - and using the cabin camera to figure it out (which it could do, without being able to see the hands or the wheel). A combination of these things has a very distinct pattern.



They are not the same at all. As long as they can measure different amounts of torque it should be easy to distinguish (pattern match).

Another easy way is to look for both hands and also look for sustained torque on the wheel when both hands are on the wheel. That is a very good indicator of a high likelihood of a defeat device. Except for those who are able to put sustained torque on the wheel with their knee (a bit tough I think, though possible, and basically unnecessary in a Tesla). Won’t see this situation with most users (seeing both hands would be pretty rare), but it is just one way to get people on a naughty list.
You could calculate the dominant frequency response of the steering wheel to vertical impulses imparted by the road to the car body. With a non-symmetrical steering wheel, the wheel will twist back and forth slightly with every bump in the road. The frequency spectrum of this movement can be calculated and the standard deviation of the values determined. A wheel weight should have a very dominant frequency resonance with low standard deviation. A human will almost certainly have a much more varied effect, spreading out the frequency response. It would be interesting to see the data in histogram form, but I suspect that a person could easily detect the difference between the wheel weight and human.

I can think of a few other methods, including using impulse response magnitude or estimate of rotational inertia. Obviously, I don't know how Tesla is doing it but obviously the idea is to find a method with as little overlap to the human. From numerous reports here, it appears that the current method has too much overlap.

As far as capacitance wheel sensors go, I suspect there are ways to trick those as well and maybe harder to detect the cheat devices.
 
There are plenty of other ways, like analog detection of torque as you describe - and using the cabin camera to figure it out (which it could do, without being able to see the hands or the wheel). A combination of these things has a very distinct pattern.
The position of your arms provides at best a ballpark guess. Some people drive with their hands practically in their laps, which would be hard to distinguish from actually being in their laps without knowing the length of their forearms.


They are not the same at all. As long as they can measure different amounts of torque it should be easy to distinguish (pattern match).
Amounts of torque caused by an attached mass, whether a hand or a weight, are continually affected by outside factors — bumps, turning, acceleration, deceleration, inclines, etc. — all of which would have to be somehow factored in/canceled out, because otherwise they will all cause wide variations in the amount of torque applied by that mass.

And I'm not convinced there's that much difference in the rotational torque caused by the arm having pivots compared with a free-hanging weight of similar mass.


Another easy way is to look for both hands and also look for sustained torque on the wheel when both hands are on the wheel. That is a very good indicator of a high likelihood of a defeat device. Except for those who are able to put sustained torque on the wheel with their knee (a bit tough I think, though possible, and basically unnecessary in a Tesla). Won’t see this situation with most users (seeing both hands would be pretty rare), but it is just one way to get people on a naughty list.
It took me a third reading to realize that you must have meant "when both hands are off the wheel". :cool:

But yes, in that particular situation, assuming you have cabin cameras, seeing significant torque while your hands are visibly away from the wheel would be easily detectable, assuming that it never confuses something else for your hands, such as a passenger's hands, patterns on your shirt or pants, a chicken sandwich that your kid threw at you from the back seat, etc. 😁


You could calculate the dominant frequency response of the steering wheel to vertical impulses imparted by the road to the car body. With a non-symmetrical steering wheel, the wheel will twist back and forth slightly with every bump in the road. The frequency spectrum of this movement can be calculated and the standard deviation of the values determined. A wheel weight should have a very dominant frequency resonance with low standard deviation.
On average, I'd expect a human to have more of a bounce-back after an impact, but it depends on how tightly you're gripping the wheel, arm weight, how tense your arms are, etc., not to mention whether you let go of the wheel at those moments or hold on. You can't count on that. If there is any bounce-back, the frequency of that bounce-back would be basically dependent on the length of the person's arms and the wheel angle, and so probably won't vary much at all. But there may not be any bounce-back if the person drives more rigidly.


A human will almost certainly have a much more varied effect, spreading out the frequency response. It would be interesting to see the data in histogram form, but I suspect that a person could easily detect the difference between the wheel weight and human.

But the question isn't whether you could detect it or not, but rather whether you could detect it without false positives, which by their nature, are trust-destroying in both directions, in the single-digit minutes before the drive ends.
 
Some people drive with their hands practically in their laps, which would be hard to distinguish from actually being in their laps without knowing the length of their forearms.

You should check out the cabin cam. As a human it is actually pretty easy to figure out. Just watch for a couple minutes, see what happens when going around corners, etc. - remember the idea is to figure out whether the driver is engaged and generally in the habit of holding the wheel.

Watch some video and you’ll see it is pretty easy.

when both hands are off the wh

Yes that is what I meant. Sorry.

assuming that it never confuses something else for your hands, such as a passenger's hands, patterns on your shirt or pants, a chicken sandwich that your kid threw at you from the back seat, etc.
If it does this for a simple scene, we’ve got much bigger problems with FSD. The ability to discern all objects in a scene has to be truly remarkable. Otherwise it is kind of pointless and dangerous.

And remember it can do this over many minutes, it’s a video, not a static image.
 
Last edited:
On average, I'd expect a human to have more of a bounce-back after an impact, but it depends on how tightly you're gripping the wheel, arm weight, how tense your arms are, etc., not to mention whether you let go of the wheel at those moments or hold on. You can't count on that. If there is any bounce-back, the frequency of that bounce-back would be basically dependent on the length of the person's arms and the wheel angle, and so probably won't vary much at all. But there may not be any bounce-back if the person drives more rigidly.
The wheel is going to move as a result of road-induced vibration. The only issue is whether the car can resolve this normal movement. If the position encoder does not have sufficient resolution, then this method would not work. But, if anything, I would expect a human would greatly dampen any impulse, or vibration, effect and remove high frequency components. A wheel weight applies no damping, so the wheel is more likely to find a natural resonance.

I'd love to have raw wheel data to play with some techniques in Matlab. But, I'm suspecting that the raw data is not passed over the CAN bus. I have a feeling that Tesla programmed the steering controller to do whatever processing is required.
 
I used to use both hands with them braced such that one was pulling steadily and the other lightly rested on the wheel.
FWIW, my wife's Honda CR-V hydrid has LKAS and ACC and as of right now just factoring in those features, it works way better than Tesla's versions. Now, it is incredibly strict with having hands on the wheel. If you remove then, within about 2 seconds a countdown timer starts at 5 seconds and goes down. If you've still not put your hands on the wheel, then it shuts off. BUT..... it is incredibly good at detecting when your hands are on the wheel. Two hands at 10 - 2. One hand at 6. It doesn't matter. If any hand is firmly on the wheel, then it will detect it and you're good to go. Can't comment on other manufacturers "hands on the wheel" detection. But clearly Honda is doing far better than Tesla at this particular issue.
 
FWIW, my wife's Honda CR-V hydrid has LKAS and ACC and as of right now just factoring in those features, it works way better than Tesla's versions. Now, it is incredibly strict with having hands on the wheel. If you remove then, within about 2 seconds a countdown timer starts at 5 seconds and goes down. If you've still not put your hands on the wheel, then it shuts off. BUT..... it is incredibly good at detecting when your hands are on the wheel. Two hands at 10 - 2. One hand at 6. It doesn't matter. If any hand is firmly on the wheel, then it will detect it and you're good to go. Can't comment on other manufacturers "hands on the wheel" detection. But clearly Honda is doing far better than Tesla at this particular issue.
What happens when it shuts off? Will the car just slam into something?
 
FWIW, my wife's Honda CR-V hydrid has LKAS and ACC and as of right now just factoring in those features, it works way better than Tesla's versions. Now, it is incredibly strict with having hands on the wheel. If you remove then, within about 2 seconds a countdown timer starts at 5 seconds and goes down. If you've still not put your hands on the wheel, then it shuts off. BUT..... it is incredibly good at detecting when your hands are on the wheel. Two hands at 10 - 2. One hand at 6. It doesn't matter. If any hand is firmly on the wheel, then it will detect it and you're good to go. Can't comment on other manufacturers "hands on the wheel" detection. But clearly Honda is doing far better than Tesla at this particular issue.
Have to disagree just a bit after owning the Accord premium touring edition. Yes it does monitor better but to say it’s even compatible with LKAS is a huge stretch. Basically a lane pinball system and if there was Any road curve left or right it failed. Loved the car but the AP type component was more so gimmick over value.
 
Have to disagree just a bit after owning the Accord premium touring edition. Yes it does monitor better but to say it’s even compatible with LKAS is a huge stretch. Basically a lane pinball system and if there was Any road curve left or right it failed. Loved the car but the AP type component was more so gimmick over value.
I keep reading about other manufacturers' lane keeping systems that disconnect on curves. That must be quite the surprise to have the car suddenly head for the oncoming lane or ditch. One nice thing about my Tesla is that it has never disconnected while taking a curve, no matter how sharp.
 
  • Like
Reactions: Yelobird
Have to disagree just a bit after owning the Accord premium touring edition. Yes it does monitor better but to say it’s even compatible with LKAS is a huge stretch. Basically a lane pinball system and if there was Any road curve left or right it failed. Loved the car but the AP type component was more so gimmick over value.
I must admit that the Tesla is smoother going around curves than the Honda which is a little jerky (though safe). But I still stand by my statement that the Honda is better. The Honda I can use. If I turn on the Tesla right now, it's going to give me a strike due to an erroneous "cheat device detected" warning and then give me the red steering wheel and then shut off.
 
  • Like
Reactions: MTOman
Under "Customize Navigate on Autopilot",
Full Self-Driving (Beta) is greyed out.
the greyed out FSD says my FSD will be restored at the next software update.
 

Attachments

  • Full Self Driving.JPG
    Full Self Driving.JPG
    683 KB · Views: 119
  • Love
Reactions: AlanSubie4Life
I received cheat device alert twice and fsd disabled for the rest of the drive
My fsd section shows improper ap usage detected twice , if I receive 5 more disengagement fsd will be disabled

Does this mean I have two strikes out of 5

In the previous update , the qty 5 used to decrease, can some one confirm, did they stop updating it
 
I received cheat device alert twice and fsd disabled for the rest of the drive
My fsd section shows improper ap usage detected twice , if I receive 5 more disengagement fsd will be disabled

Does this mean I have two strikes out of 5

In the previous update , the qty 5 used to decrease, can some one confirm, did they stop updating it
Instead of asking people who don't have access to your car to know why not just look at the Autopilot menu yourself?
 
Last edited: