Autopilot buddy code is changed?

[Gjy40404]

Did Tesla change the code to avoid steering boosters? So now what is the requirement to keep AP working? I got strike even I focus on the road and constantly keep a force on the wheel (=install a booster)
AP works for 10 miles and forced AP disengagement happened. Any thoughts? Thanks all.

 

[Grabthar]

Yes they did change the code. AP can now tell if you're using a cheat or not. It can detect the cheat so hands on the wheel!

 

[2101Guy]

Key is to randomly move the steering wheel just a tad/bit about once every 8 mins or so, mainly when on long straight stretches. If there are curves in the road, the weight still works effectively for long periods of time.

 

[Gjy40404]

Yes they did change the code. AP can now tell if you're using a cheat or not. It can detect the cheat so hands on the wheel!

Thank you. How does it detect? I don’t think interior camera can see the steering wheel. My eyes are focusing on the road.
If constant force on the wheel is considered as a cheat, does it make sense? If I keep Both hands on the wheel also applying light and constant force. Won’t there be false detection?

 

[Daniel in SD]

Thank you. How does it detect? I don’t think interior camera can see the steering wheel. My eyes are focusing on the road.
If constant force on the wheel is considered as a cheat, does it make sense? If I keep Both hands on the wheel also applying light and constant force. Won’t there be false detection?

My guess is that a human arms holding the wheel do not behave the same way as a fixed weight over bump and around turns.

 

[AlanSubie4Life]

Do we have evidence that this has actually changed? Clearly theoretically easy to detect, but I don’t think there is any evidence for it.

Remember forced disengagement can happen for inattention to the road. Autopilot Buddy by itself is not enough to satisfy the attention monitoring.

We would need a video of the strike and/or more details about the warnings on the screen prior to the strike to know for sure. Was there blue flashing? What vehicle? Looks like 2020 Model S at least, maybe Plaid, based on post history. So has interior camera.

I have not noticed that anything is required other than steady unchanging torque (and attention to the road) but of course hard to know for sure and impossible to emulate a weight.

In general, it is impossible to get a strike, when driving the vehicle.

 

[Tam]

Thank you. How does it detect? I don’t think interior camera can see the steering wheel. My eyes are focusing on the road.
If constant force on the wheel is considered as a cheat, does it make sense? If I keep Both hands on the wheel also applying light and constant force. Won’t there be false detection?

It's a torque monitor, not camera.

Torque produced by imperfect human has a different pattern than a perfect, unchanging weight of an object that you bought.

 

[Tam]

Do we have evidence that this has actually changed? Clearly theoretically easy to detect, but I don’t think there is any evidence for it.

Remember forced disengagement can happen for inattention to the road. Autopilot Buddy by itself is not enough to satisfy the attention monitoring.

We would need a video of the strike and/or more details about the warnings on the screen prior to the strike to know for sure. Was there blue flashing? What vehicle? Looks like 2020 Model S at least, maybe Plaid, based on post history. So has interior camera.

In the beginning, youtubers were showing us how to cheat.

After that, commercial devices were sold with effective cheat.

Then after that, what used to be an effective cheat no longer working.

 

[AlanSubie4Life]

In the beginning, youtubers were showing us how to cheat.

After that, commercial devices were sold with effective cheat.

Then after that, what used to be an effective cheat no longer working.

Lots of reasons for that though completely unrelated to torque requirements.

As I said, we’d have to have video evidence of it. Also on some specific early software releases the torque requirements have been extreme. I’d set those aside.

I’ve seen no such YouTube videos so I assume this is not a thing. And of course we would need to know specific software version to see if it is an exception (see above).

Please note I am not saying it is not a thing. It may well be true. I am just assuming it is not. Extraordinary claims!

 

[flutas]

Do we have evidence that this has actually changed? Clearly theoretically easy to detect, but I don’t think there is any evidence for it.

https://twitter.com/x/status/1591534839240757249

https://twitter.com/x/status/1591535160734154753

 

[AlanSubie4Life]

https://twitter.com/x/status/1591534839240757249

https://twitter.com/x/status/1591535160734154753

Thanks. That is what I was looking for. Somehow missed these Tweets (I do follow him!).

This is wonderful news. I love it.

Now people will have to step up to something resembling a shake weight!

They really are getting ready for true wide release!

 

[Gjy40404]

Sorry I didn’t make that video. What happened to me yesterday:

1. Plaid S with FSD. 36.6 version
2. Install booster and car drive itself with my 0 engagement for 10 miles.
3. Without any prompt no warning for 10 miles.
4 suddenly forced AP disengagement (red steering wheel on screen) prior to that, no warning at all.

I guess I find a solution against this “anti-booster cheat”

 

[clydeiii]

Thank you. How does it detect? I don’t think interior camera can see the steering wheel. My eyes are focusing on the road.
If constant force on the wheel is considered as a cheat, does it make sense? If I keep Both hands on the wheel also applying light and constant force. Won’t there be false detection?

Most likely it just realizes the torque being applied is constant. So the next version of AP Buddy will need to modify its weight constantly I guess. Good luck.

 

[AlanSubie4Life]

Sorry I didn’t make that video. What happened to me yesterday:

1. Plaid S with FSD. 36.6 version
2. Install booster and car drive itself with my 0 engagement for 10 miles.
3. Without any prompt no warning for 10 miles.
4 suddenly forced AP disengagement (red steering wheel on screen) prior to that, no warning at all.

I guess I find a solution against this “anti-booster cheat”

Yes, based on Green’s tweets, sounds like this is expected. This is excellent news.

I have not driven with that version yet so it’ll be interesting to see if using hands on wheel consistently in one direction ever results in a false strike like this. Hope they did a good job of the detection! Can’t have false positives.

It would suck if you have to occasionally torque the wheel rather than constantly. “Constant” torque is the way!

But probably it will be just fine.

 

[brancky3]

Sorry I didn’t make that video. What happened to me yesterday:

1. Plaid S with FSD. 36.6 version
2. Install booster and car drive itself with my 0 engagement for 10 miles.
3. Without any prompt no warning for 10 miles.
4 suddenly forced AP disengagement (red steering wheel on screen) prior to that, no warning at all.

I guess I find a solution against this “anti-booster cheat”

During those 10 miles did you touch the steering wheel at all? Like to change volume, change tracks, adjust speed, or adjust following distance?

 

[Gjy40404]

During those 10 miles did you touch the steering wheel at all? Like to change volume, change tracks, adjust speed, or adjust following distance?

No touch at the wheel at all. No lane change. Asked the manufacturer, they said current mitigation is touch any wheel button to extend the time from being strike

 

[brancky3]

No touch at the wheel at all. No lane change. Asked the manufacturer, they said current mitigation is touch any wheel button to extend the time from being strike

Yea I think it's pretty rare for me to go 10 miles without changing the song or the volume at a bare minimum. Maybe that's why I don't get disengagements as much?

 

[clydeiii]

Yes, based on Green’s tweets, sounds like this is expected. This is excellent news.

I have not driven with that version yet so it’ll be interesting to see if using hands on wheel consistently in one direction ever results in a false strike like this. Hope they did a good job of the detection! Can’t have false positives.

It would suck if you have to occasionally torque the wheel rather than constantly. “Constant” torque is the way!

But probably it will be just fine.

No human can apply constant torque. There will always be slight variations even when you use the technique I demo here:

 

[AlanSubie4Life]

No human can apply constant torque. There will always be slight variations even when you use the technique I demo here:

Yes of course, as I said (“ “). The question is the tolerance on Tesla’s detection.

It’ll probably be just fine; this does not seem all that hard. If I get a strike we’ll know Tesla screwed it up.

Ready for wide release! (Not to just those who ask.)

 

[EVNow]

I guess I find a solution against this “anti-booster cheat”

I hope they ban you from AP/FSD. May be report to NHTSA as well.