Software Update 2018.21.9 75bdbc11

[Joe F]

They have an unfixable design flaw but they will never admit it. They will try to smooth it over as well as they can, forcing drivers into the less safe option of having only one hand on the wheel instead of two, because the torque sensor can't detect two hands on the wheel because each hand cancels out the torque of the other.

Not true if done correctly.

Statements like this by some here underscores the false idea that the torque sensor is somehow a failed method. It's not. It's a failure by some to understand how things work as currently designed.

Take a non Tesla car with no steering automation at all, and you have to steer manually, using both hands on the wheel. A curve in the road means you have to turn the wheel in the direction of the turn. Your arms move in opposite directions. If they did not, you would not be able to turn the wheel as your own strength would cancel each out. This is common sense and not something we humans don't have to think about.

Now take that example to a Tesla AP system. Your argument that your two hands cancel each others movement out, thereby nullifying the torque sensor from doing its job, makes absolutely no sense whatsoever. For that to be true, you would not be able to turn the wheel at all, and AutoSteer would disengage.

 

[realvvk]

BTW, whenever I read the articles about a crash and Tesla comes out proclaiming that "hands were not detected on the wheel for 6 seconds leading up to the crash", I think about my own experience of having my hands on the wheel constantly but still getting nagged -- just because Tesla did not detect hands on the wheel leading up to the crash does not mean hands were not on the wheel. Not detected for 6 seconds tells me hands were probably on the wheel but undetected. Not detected for 60 seconds is a different story.

1000x this. I find it very, very concerning that Tesla even uses this tactic to justify problems with the Autopilot. I get what they are doing, I do. They want to show that the driver was not paying attention and that lead to the crash. But currently Tesla has NO WAY of proving that the driver was not paying attention. Because even if you hold the wheel, the car does detect it AT ALL. The fact that the logs says that driver's hands were not detected does not mean that he was not holding the wheel. It just means that he was not putting rotational torque on the wheel AT THAT EXACT TIME. It is BS and Tesla knows it.

 

[ahkahn]

1000x this. I find it very, very concerning that Tesla even uses this tactic to justify problems with the Autopilot. I get what they are doing, I do. They want to show that the driver was not paying attention and that lead to the crash. But currently Tesla has NO WAY of proving that the driver was not paying attention. Because even if you hold the wheel, the car does detect it AT ALL. The fact that the logs says that driver's hands were not detected does not mean that he was not holding the wheel. It just means that he was not putting rotational torque on the wheel AT THAT EXACT TIME. It is BS and Tesla knows it.

So you're telling me that the guy in the 101 crash was so preoccupied trying to properly torque his steering wheel that he ran head on into a concrete barrier?

I'm sorry, but common sense tells you the guy wasn't paying attention... especially in an area he allegedly knew was a faulty spot for AP.

 

[Abqnative]

I agree with Joe F and find the hands on feedback to be natural. I just got the latest update and am planning to try out the exit feature. I am concerned however about following the exit ramp path. In the previous version of the software when I request a lane change, it may take a while or not do it at all even if the path is clear. For example, I request a lane change to the left and the car remains in the current lane. I see in the display that the left lane exists but I also see the left side sensor white arc which means it sees an object (which is not there). BTW. The arc is from the side of the car and not the left quarter panel.

 

[realvvk]

So you're telling me that the guy in the 101 crash was so preoccupied trying to properly torque his steering wheel that he ran head on into a concrete barrier?

I'm sorry, but common sense tells you the guy wasn't paying attention... especially in an area he allegedly knew was a faulty spot for AP.

All I am saying is that using the logs to PROVE he was not paying attention is very misleading. We all know that. Any Tesla driver who uses Autopilot knows that. Of course he was not paying attention. However, if you look at the second by second account of what happened, you can see that he only had a split second to react. I think in this particular case it is not all that clear that the autopilot did not misinterpret the lane markings. But despite whatever the autopilot did, it was the driver's responsibility to pay attention and to use good judgement to predict what would happen.

 

[rnortman]

Not true if done correctly.

Statements like this by some here underscores the false idea that the torque sensor is somehow a failed method. It's not. It's a failure by some to understand how things work as currently designed.

Take a non Tesla car with no steering automation at all, and you have to steer manually, using both hands on the wheel. A curve in the road means you have to turn the wheel in the direction of the turn. Your arms move in opposite directions. If they did not, you would not be able to turn the wheel as your own strength would cancel each out. This is common sense and not something we humans don't have to think about.

Now take that example to a Tesla AP system. Your argument that your two hands cancel each others movement out, thereby nullifying the torque sensor from doing its job, makes absolutely no sense whatsoever. For that to be true, you would not be able to turn the wheel at all, and AutoSteer would disengage.

I honestly am having a lot of difficulty parsing what you wrote here into anything that makes sense. I'm sorry, it's probably my stupidity.

If I apply torque as if I were steering the car, that would be sensed. But the point of autopilot is to not steer the car. I am talking about having two hands passively on the wheel. The system does not detect this. I am not making this up, I own a HW2.5 Model S. It does not detect both hands on the wheel passively in anything like a comfortable position, unless you're on a pretty curvy road -- when the wheel turns more than a few degrees you end up applying torque, but this doesn't help on long highway stretches. One hand on the wheel is the only viable option for long straight highways.

 

[Pentium2004]

Not true if done correctly.

Statements like this by some here underscores the false idea that the torque sensor is somehow a failed method. It's not. It's a failure by some to understand how things work as currently designed.

Take a non Tesla car with no steering automation at all, and you have to steer manually, using both hands on the wheel. A curve in the road means you have to turn the wheel in the direction of the turn. Your arms move in opposite directions. If they did not, you would not be able to turn the wheel as your own strength would cancel each out. This is common sense and not something we humans don't have to think about.

Now take that example to a Tesla AP system. Your argument that your two hands cancel each others movement out, thereby nullifying the torque sensor from doing its job, makes absolutely no sense whatsoever. For that to be true, you would not be able to turn the wheel at all, and AutoSteer would disengage.

I just drive down the road shaking the wheel like I'm a 5 year old pretending to drive. Autopilot seems to be OK with that since I don't get any warnings. JK lol. I typically just hold on with my left hand somewhere around 7-8 o'clock and rest my elbow on my leg. It works very well but you will have to switch out arms because it gets tiring.

 

[Joe F]

1000x this. I find it very, very concerning that Tesla even uses this tactic to justify problems with the Autopilot. I get what they are doing, I do. They want to show that the driver was not paying attention and that lead to the crash. But currently Tesla has NO WAY of proving that the driver was not paying attention. Because even if you hold the wheel, the car does detect it AT ALL. The fact that the logs says that driver's hands were not detected does not mean that he was not holding the wheel. It just means that he was not putting rotational torque on the wheel AT THAT EXACT TIME. It is BS and Tesla knows it.

Hi there,

I see that you're in Philly. Is that the Philly in PA (there are others).

If so, I'm kind of your neighbor in Willow Grove and would like to meet up so we can compare how our Tesla AP's differ, given that your Tesla does not detect your hands are on the wheel at all. Perhaps there's something wrong with your car that needs attention by the good folks in Devon.

Let me know if interested and we can pick someplace neutral! I'm not available this Saturday though...

-Joe

 

[rnortman]

I just drive down the road shaking the wheel like I'm a 5 year old pretending to drive. Autopilot seems to be OK with that since I don't get any warnings. JK lol. I typically just hold on with my left hand somewhere around 7-8 o'clock and rest my elbow on my leg. It works very well but you will have to switch out arms because it gets tiring.

Agreed. This works well, but is tiring. The more comfortable position with two hands on the wheel is both more comfortable over a long drive and safer over any distance, but works poorly at avoiding the nags.

 

[Pentium2004]

I agree with this and it also worries me about false reports from Tesla. I've seen others comment with the same concern. That Tesla log will be like the law when you crash on AP. If you crash while on AP with your hands on the wheel and it just happens to not detect at the time of impact, it'll be your word against that log. Nobody will believe your word.

 

[NerdUno]

Part of the problem is that AP and AP2 are all-or-nothing systems. Either AutoPilot is driving the car, or you are. Never both. If you don't provide sufficient torque on the steering wheel, you set off the alarm. If you put too much torque on the wheel, you turn off AutoPilot. Personally, I prefer the Nissan approach with the Leaf and Rogue that lets you steer normally without disabling AutoPilot. It only turns off if the car loses the lane markings. And it turns back on when the lane markings reappear. There's also no detectable torque requirement even though the car correctly detects when your hands are on the wheel.

 

[NOLA_Mike]

I agree with this and it also worries me about false reports from Tesla. I've seen others comment with the same concern. That Tesla log will be like the law when you crash on AP. If you crash while on AP with your hands on the wheel and it just happens to not detect at the time of impact, it'll be your word against that log. Nobody will believe your word.

I would think that the videos now surfacing showing "Hold Steering Wheel" warning while people's hands are on the wheel would be enough to discredit Tesla's assertion that you didn't have your hands on the wheel.

But I'm no lawyer and YMMV...

 

[Joe F]

I honestly am having a lot of difficulty parsing what you wrote here into anything that makes sense. I'm sorry, it's probably my stupidity.

If I apply torque as if I were steering the car, that would be sensed. But the point of autopilot is to not steer the car. I am talking about having two hands passively on the wheel. The system does not detect this. I am not making this up, I own a HW2.5 Model S. It does not detect both hands on the wheel passively in anything like a comfortable position, unless you're on a pretty curvy road -- when the wheel turns more than a few degrees you end up applying torque, but this doesn't help on long highway stretches. One hand on the wheel is the only viable option for long straight highways.

The bottom line is, you do not need to "steer" an AP Tesla when using AutoSteer. If someone chooses to keep both hands on the wheel (I personally don't unless in challenging road conditions, including where AP is known to screw up) all one need do is leave one hand "loose" and ready to take control if need be. Neither hand needs to "steer" regardless. All one need do is feel the wheel as it moves, and yes it still does move slightly in straight road segments, and periodically tighten your grip to provide resistance to the movements. That is enough to be sensed by the torque sensor, proving your are still awake, alive and alert enough for Ap to carry on. And there is no "tugging" involved as some have suggested they need to do. If anything, all that's requires is a very gentle attempt to turn the wheel until lit feels as if the movement is being resisted, in practice the wheel doesn't even move.

People who believe the need to constantly provide "torque" for AP to work without nags are simply doing it wrong.

 

[Cirrus MS100D]

Received the update on Monday night on our late summer 2017 MS100D with AP2 MCU 1 and I honestly have to put myself in the “what’s the big deal?” camp about the AP nags. I just wasn’t seeing or experiencing it. In fact, I only got the white bar 1 time during a 40 minute drive, mix of highway and rural (not technically where I shouldn’t be using AP, I know, but it’s fun to see how it handles it.)

I found the back roads to be very good, with nice anticipation slowing down for sharp curves. This was coming from .18, which was also pretty good.

Highway felt solid, but like others, I’ve detected some “flittiness” like it’s more on-edge. My suspicion is that some of the “false-positive / ignore” knobs have been turned down a bit to mitigate some of the problems of the recent headline-inducing crashes. No evidence for this, just gut feel.

Overall I’m happy and like seeing cars in adjacent lanes (even though I can see them with my eyes) as I have a feeling it’s only a short while before they start showing up BEHIND us too.

 

[kavyboy]

I honestly am having a lot of difficulty parsing what you wrote here into anything that makes sense. I'm sorry, it's probably my stupidity.

If I apply torque as if I were steering the car, that would be sensed. But the point of autopilot is to not steer the car. I am talking about having two hands passively on the wheel. The system does not detect this. I am not making this up, I own a HW2.5 Model S. It does not detect both hands on the wheel passively in anything like a comfortable position, unless you're on a pretty curvy road -- when the wheel turns more than a few degrees you end up applying torque, but this doesn't help on long highway stretches. One hand on the wheel is the only viable option for long straight highways.

I can confirm (on all counts).
The only time I have ever gotten a nag (which doesn't include the latest version, which I won't be installing) was when I had both hands on the wheel. I've done about 15,000 miles on AP1 with one hand on the wheel and only gotten one or two nags, if any, that way. When traffic gets crazy in an unfamiliar city or in any other kind of situation that's uncomfortable for some reason, the music goes off, conversations end, and I'll put both hands on the wheel and keep AP on as a "backup driver" kind of thing, to help keep me centered while I shift a bit of attention to whatever needs it. That's when I get nags - both hands on the wheel, hyper vigilant, and totally aware. And my only distraction is the nags.

 

[rnortman]

The bottom line is, you do not need to "steer" an AP Tesla when using AutoSteer. If someone chooses to keep both hands on the wheel (I personally don't unless in challenging road conditions, including where AP is known to screw up) all one need do is leave one hand "loose" and ready to take control if need be. Neither hand needs to "steer" regardless. All one need do is feel the wheel as it moves, and yes it still does move slightly in straight road segments, and periodically tighten your grip to provide resistance to the movements. That is enough to be sensed by the torque sensor, proving your are still awake, alive and alert enough for Ap to carry on. And there is no "tugging" involved as some have suggested they need to do. If anything, all that's requires is a very gentle attempt to turn the wheel until lit feels as if the movement is being resisted, in practice the wheel doesn't even move.

People who believe the need to constantly provide "torque" for AP to work without nags are simply doing it wrong.

Can you develop a two-handed style that avoids the nags? Yes, it is possible. But it is enough less comfortable than one-handed nag-avoidance that 99% of drivers are going to go one-handed 99% of the time.

As for your specific recommendation, one hand loose and one hand ready, that is (a) really uncomfortable for a long period, and (b) pretty much equivalent to just using one hand and having the other resting ready in your lap.

 

[croman]

Lawyers, I would guess.

Yes, you can blame lawyers for the nags but you can also thank us for holding the Nation together and enforcing Rule of Law when the executive is determined to undermine both.

 

[alcibiades]

Yes, you can blame lawyers for the nags but you can also thank us for holding the Nation together and enforcing Rule of Law when the executive is determined to undermine both.

Oh stoppit. My wife's a lawyer! And I love her!

Anyway, I didn't mean to be blaming lawyers for the nags, I was just saying I think that lawyers are why we are about to see a wording change in the AP warning. You know, so it accurately reflects what Tesla is detecting because Tesla is going to be sued/or is being sued for publicly claiming that accident victims/bad drivers hands weren't on the wheel.

 

[Joe F]

Can you develop a two-handed style that avoids the nags? Yes, it is possible. But it is enough less comfortable than one-handed nag-avoidance that 99% of drivers are going to go one-handed 99% of the time.

As for your specific recommendation, one hand loose and one hand ready, that is (a) really uncomfortable for a long period, and (b) pretty much equivalent to just using one hand and having the other resting ready in your lap.

Some people should just buy a non Tesla vehicle I guess.

 

[BigD0g]

Yes, you can blame lawyers for the nags but you can also thank us for holding the Nation together and enforcing Rule of Law when the executive is determined to undermine both.

We can also count on the lawyers to get us what we bought if Tesla fails to deliver on the FSD dream bought by many!
Too soon for that class action, however I think an EAP CA is loooooong overdue and no $288 doesn't cover it, because
they still have not delivered 1 EAP feature 1.5 years later!