Any hack to remove the autopilot nag?

[aaron0k]

I dont have to provide anything or prove anything. It was you who made an unsubstantiated claim that "there were better ways" to monitor drivers, and its for you to provide supporting evidence. The VTI quote is interesting, but is just opinion. Has the guy done any actual research to quantify the different attention systems? I couldn't find anything to suggest this on his web site. So your claim remains, in fact, just your opinion.

I provided information showing that subject matter experts (more than one) all consider wheel torque sensors inferior to camera based systems or a combination of both. Which is what you "tried" calling me out on...

"You have nothing to prove"... I'll take that as a "no" then; that you can't provide any information supporting your position... and can only submit 'your opinion'. You may not want to concede the point, however I provided concrete, third-party information supporting 'my opinion' that cameras are better than torque sensors.

I think most reasonable people would consider these "expert" opinions; at the very least more credible than yours... or mine.

As for the NTSB quote, its not relevant.

Really? I'll quote it again for you...

"monitoring of driver-applied steering wheel torque is an ineffective surrogate measure of driver engagement" - NTSB​

That statement pretty much negates your entire argument. It's from the NTSB. But I guess you're saying TMC forum member opinions are more relevant than statements from the NTSB. Weird.

By the way Doctor; you must have missed my other question.... Would that driver in the AP crash still be with us if Tesla used cameras instead of torque? (pssst... answer is yes).

Anyway, this is silly... made my point. Substantiated it thoroughly. Moving on...

 

[Fernand]

$8k, exactly. We paid $5k for EAP in Dec 2018, then $3k in 2019 for the FSD, in full awareness of it being for the future. You can buy a roughly similar config nowadays for almost $8k less, though I like the medium range battery, the premium interior/speakers. Bottom line, I got no life-hours to allocate to spilt milk over changing prices, or grumbling about imperfections. I bought and I'm driving what I like right now, and it's been a unique experience watching it evolve. If I'm buying this great a car, I'm all in.
.

 

[Knightshade]

But they did pay for EAP+FSD

Sure- but EAP has been providing the same full fucntionality everyone post 3/19 had to pay $7000 to get.

So that's better

as well as the generally higher price of the cars in 2018.

Given how often this has been debunked it's baffling people keep repeating it.

With the tax credits considered (and the HW that's been removed as standard) prices on the actual cars are virtually identical today as in 2018 (in some cases they're higher now)


For example in 2018 a model 3 LR AWD was $59,000 with EAP in blue before destination/tax.

And I got a $7500 federal tax credit.

So $51,500 net price.


Today LR AWD is $48,990, plus $1000 for blue, plus $7000 for FSD (to get all the stuff EAP used to have)- that's $56,990... it'll be another roughly $425 to get all the HW the 2018 came with free (homelink, 14-50, mats, phone cables, etc)... and destination is $275 higher than 2018... so $57,690 today. $0.00 federal tax credit.

The 2018 car was over $6000 cheaper than a 2020.

Even if you add FSD to the 2018 for $3000 (even though as of today it still adds 0 functionality) the 2018 is still over $3000 cheaper than a new 2020.


Now if you wanna bring up folks who bought a P100D with Ludicrous in 2018 you've got a good case for paid a lot more... but 3 buyers? Not so much.

$8k, exactly. We paid $5k for EAP in Dec 2018, then $3k in 2019 for the FSD, in full awareness of it being for the future. You can buy a roughly similar config nowadays for almost $8k less

Net of tax credit, you really can't. See math above.

 

[Fernand]

Yeah, you're probably right, if we include the changing rebate landscape. Crunching numbers isn't my forte, in the end, for me, it's all about the car.

 

[Fernand]

I'm curious. If they say

monitoring of driver-applied steering wheel torque is an ineffective surrogate measure of driver engagement" - NTSB

then what exactly do they suggest as an effective non-surrogate type monitoring? a professional NTSB companion in the car?

I think they're just trying to say something "expert" for this unfortunate case. Because a distracted driver is not 100% detectable, short of wearing an immobilizing harness, hand restraints, and wiring into the frontal cortex, eyeball and eyelid muscles. Like in Clockwork Orange.

But what are they doing to prevent inattention and accidents in all the millions of old cruise-controlled ICE cars that just plough ahead?
.

 

[S4WRXTTCS]

I can't see using TACC. Who's using THAT? Without auto-steer and without ability to change lanes by tapping the turn signal, it's just a cruise control, and not very useful.

I use TACC largely because of the shortcomings with AP. I imagine basic AP owners mostly use TACC because they have no way to change lanes without stopping lane-steering, and then starting it back up after the lane change. There is also probably a small subset of people who simply feel more comfortable with TACC.

Probably the number one thing I despise about AP is the tendency for it to recenter itself in the right lane during a merge point. I'm certainly not the only one that's had this issue. Every time there is a new firmware upgrade people claim it's been solved, but I've never seen it disappear completely.

As to what frustrates me the most about TACC/AP/NoA is not frustration that it can't do more, but frustration that what it does today isn't more consistent.

I certainly have enjoyable drives, but I also have drives where it fails at very basic things. Now in a way I might have to temper my expectation because most of my TACC/AP/NoA driving is on road trips. It's much more difficult to go 400+ miles of different variables versus 50 miles a day over the same variables every day.

There is also the potential that there might be something wrong with an individuals car. So that kind of thing does add variability to the fleet.

Here is a good example of that possibly being a factor.
Autopilot performance after upgrading to FSD : teslamotors

This individual complained of the EXACT same thing that I despise. But, they claim they don't have any issues with a Model S with EAP. They didn't check the firmware versions so I don't know what role that played.

I'm not very confident of that because I've seen plenty of youtube reviews (who are pretty "it is what it is" types) from people doing extensive NoA testing where it's not able to score a good grade. In fact the last one I watched the reviewer didn't even bother testing it any further because NoA failed to get out of the passing lane, and I think there were some other issues as well. So they were just going to wait for those things to be fixed.

In my own testing I'm just flabbergasted by what doesn't meet my expectations which I don't feel are all that high.

TACC isn't smooth on P3D. The bench mark is the AP1 Model S I had previously.
Lots of people including myself report false braking with TACC or AP (I don't think false braking events are any different between the two). The bench mark again is AP1 Model S.
NoA is highly dependent on maps so that one I'm a little more forgiving on, but some of the decisions it makes is downright mind-boggling
Smart Summons is a complete joke. I even went through the effort of updating the OpenStreet maps for the location I was testing it at work. The update vastly improved it, but it still failed to consistently work properly (10 times was the goal). I gave up on it, and haven't tested it since then. This one might be worth checking since it's been months.
Smart Park we're still waiting on, and in the meantime all we have is autopark which can be really frustrating.
Unconfirmed lane changes is broken as reported by a lot of people (excessive delay compared to what it used to be)
Pedestrian detection for AEB failed to get the top grade from IIHS
Pedestrian detection didn't pass the AAA test, and didn't get the best score either.

Don't take my word for any of this. Just spend some time watching videos of people doing extensive testing. It doesn't meet even moderate expectations.

It's also important to point out that progress on FSD has stalled out. It's currently undergoing a rewrite of sorts. So I could easily be complaining about something that will be completely different in 6 months.

 

[S4WRXTTCS]

I'm curious. If they say

monitoring of driver-applied steering wheel torque is an ineffective surrogate measure of driver engagement" - NTSB

then what exactly do they suggest as an effective non-surrogate type monitoring? a professional NTSB companion in the car?

I think they're just trying to say something "expert" for this unfortunate case. Because a distracted driver is not 100% detectable, short of wearing an immobilizing harness, hand restraints, and wiring into the frontal cortex, eyeball and eyelid muscles. Like in Clockwork Orange.

But what are they doing to prevent inattention and accidents in all the millions of old cruise-controlled ICE cars that just plough ahead?
.

They're trying to push the NHTSA to force new cars to have driver monitoring systems similar to what some of the 2020 Subaru's have. So instead of a torque sensor it's going to be watching your eyes, and can tell if you're looking away from the road.

It won't be just for L2 vehicles like the Telsa, but probably all new cars after a certain year.

Now I don't think the NTSB will be successful as the NHTSA is historically a very car manufacture friendly agency. If they do anything they'll do just what they did with AEB where they get voluntary agreements in place for some way out date.

But, you will see Europe require driver monitoring systems in all new cars much more quickly.

Driver monitoring is also necessary to go from L2 to L3. The car needs to be able to go "Oh, hey you. Don't fall asleep on me. I might need to call on you." I think Tesla plans on skipping L3 where they go straight to L4 so that might not really matter.

 

[S4WRXTTCS]

Wow, that's like Orwell's 1984 in a Tesla.

Subaru, BMW, Cadillac, and maybe some others already have it on certain models.

It's not nearly as Orwell'ing as people think. It's a completely closed loop system. It's better than Tesla revealing to the world how many seconds your hands weren't detected on the wheel after you were involved in some accident that got media attention. Where they reveal that data KNOWING that it doesn't mean your hands weren't on the wheel.

I do find it kinda funny to worry about driver monitoring on a car that's connected to the internet. Heck mine is connected to the internet, and Teslafi records exactly where I drive everyday, and how fast I go including numerous other things.

So I think I already traded my soul for convenience, and data. Lots and lots of data. Yummy scrunchy data.

 

[ElectricIAC]

Coming from BMW’s system of head-eye monitoring using DA+, I can say it’s still on the buggy side and has its moments as well. That being said, the pressure sensors on the wheel used in that system aren’t the definition of perfection either and the nags can be worse with assistance feature disconnecting without much fanfare (oh, so you thought you were still lane keeping: NOPE)

But at least it can read PSL signs...

 

[drtimhill]

I provided information showing that subject matter experts (more than one) all consider wheel torque sensors inferior to camera based systems or a combination of both. Which is what you "tried" calling me out on...

"You have nothing to prove"... I'll take that as a "no" then; that you can't provide any information supporting your position... and can only submit 'your opinion'. You may not want to concede the point, however I provided concrete, third-party information supporting 'my opinion' that cameras are better than torque sensors.

I think most reasonable people would consider these "expert" opinions; at the very least more credible than yours... or mine.

Really? I'll quote it again for you...

"monitoring of driver-applied steering wheel torque is an ineffective surrogate measure of driver engagement" - NTSB​

That statement pretty much negates your entire argument. It's from the NTSB. But I guess you're saying TMC forum member opinions are more relevant than statements from the NTSB. Weird.

By the way Doctor; you must have missed my other question.... Would that driver in the AP crash still be with us if Tesla used cameras instead of torque? (pssst... answer is yes).

Anyway, this is silly... made my point. Substantiated it thoroughly. Moving on...

You claimed, quite categorically, that there were other systems that were better than that used by Tesla. You provided no specifics, no references, and no further information. In a follow-up, you provided a quote from a (proclaimed) consultant who provided his opinion. Again, no factual evidence was supplied by you or him .. none. Therefore your claim remains unsubstantiated opinion, and nothing else. You can quote as many other opinions as you like; they are just more opinions. And on what do you base the idea that such people are "experts"? Because they were quoted in the press? Or on the internet? Seriously? Are you aware that in actual, real, science, "expert" opinions carry no weight whatsoever?

As for the NTSB, where do they make any comparison to other systems? I repeat, you claim other systems are better. Where do they say that, or in any way imply that? "The Tesla system is not good" does in no way imply anything about any other systems.

As for your hypothetical "would he have survived if xxx". Well, he might perhaps have survived if he had not engaged Autopilot, or had not bought a Tesla, or had not gone to work that day, or if Apple had disabled playing games when the iPhone was in motion, or if he had bothered to read the Tesla manual/warnings and not acted like an idiot, or if Caltrans had maintained the junction properly. But neither you nor I nor anyone else can state categorically he would/would not have survived if the car had a camera-based driver attentiveness system.

And yes, this is indeed silly, and a bit sad that you have to claim to have "substantiated" something when in fact you just re-stated an opinion again and again as if that makes it fact. Fake news, anyone?

 

[Fernand]

@S4WRXTTCS, thanks truly for the explicit list. I must be easy to please. That right lane centering when the lane widens doesn't frazzle me. The car has no way to be sure whether the lane will narrow, or widen, or split, and what it does might not be what I'd do, but I don't care, it's always done it safely.

I have no way to know about hitting pedestrians statistics. It sure detects and reacts to them during Smart Summon. Smart Summon is in my eyes a test bed. It's too slow to be very practical, but it does its slow motion moves, so I don't think of it as "a complete joke".

Autopark truly and fully sucks in detecting and announcing the space, but once it does, it's always worked OK for me, and it's helpful given my back problems.

But the important thing for me is NOA, or, barring NOA mode, AS/AP on highways and city boulevards. The way my car works is just plenty good enough. I did just have an indecision slow-down yesterday, what some people call phantom braking I guess, that was unexpected, but it's like the third in well over a year, it's neither a regular event nor hard to deal with.

I'm not coming from a Model S, maybe if I did I'd be more frustrated. Self-initiated lane changes in mad max, the way my car is set up, are generally way faster and smoother than mine. Whether it waits for the best place/time for 2 or 5 seconds doesn't matter if it's not blocking anyone. I have seen it hold up a lane maybe 2-3 times over the last 14 months. I'm much worse.

I'm not expecting or demanding a professional driver, as yet. If I let it drive, it does it on average much better than me, and it takes the stress level down immensely compared to actually doing all the work. If it needs intervention, I do it, no problem.

Yes, we know the passing lane surrender has been disabled on HW2. In California, anyway, it's not mandatory like it is in Europe. I'm waiting for my HW3.

I'm having a great time with this car, my glass is half full for sure; I can't see fuming and feeling disappointed over AP details, life's too short for that! I sincerely hope that as the system evolves you'll get more joy out of it.

As to intrusive detectors, it's not about data in this case. I just reject the whole premise of tight social control. I think we're on opposite sides of this issue.
.

 

[smatthew]

Smart Summons is a complete joke. I even went through the effort of updating the OpenStreet maps for the location I was testing it at work. The update vastly improved it, but it still failed to consistently work properly (10 times was the goal). I gave up on it, and haven't tested it since then. This one might be worth checking since it's been months.

How can you say Smart Summon is a joke, when you haven't tried it in months?

 

[Joshan]

So instead of a torque sensor it's going to be watching your eyes, and can tell if you're looking away from the road.

And if I need to wear sunglasses?

 

[pscarroll]

How can you say Smart Summon is a joke, when you haven't tried it in months?

I try it every day and it's getting worse or at best no better than it was before, it's barely useable.

 

[Joshan]

I try it every day and it's getting worse or at best no better than it was before, it's barely useable.

Are you using "come to me" it seems to work great for me now. The other option of plotting a path? not so much.

 

[Knightshade]

And if I need to wear sunglasses?

Then, assuming the use the right type of camera/ir/sensors, it'll still work fine.

Caddys 1st gen system worked "most of the time" with sunglasses, but making it work the rest of the time was something they specifically cited as a planned improvement for the 2nd gen version GM is going to roll out to a lot more vehicles than just the CT6.

I'd expect anybody deploying such a system in the future to have learned the same lesson.


(again though I don't see Tesla going to this type of system AT ALL unless forced to by the NHTSA or, more likely, the European equivalent)

 

[trm2]

And if I need to wear sunglasses?

Cadillac's system works even if you are wearing sunglasses.

 

[Joshan]

Thanks for the replies, never researched the Cadillac system. Learned something new today

 

[Joshan]

will this be the new hack?

 

[DopeGhoti]

Maybe the best system will be educating people to take driving as seriously as they should. Maneuvering a several-thousand-kilogram metal box down a three meter-wide path at a hundred kilometers per hour is a task that conveys a high degree of responsibility, and the gravity of this responsibility is in very large part not taken seriously in America at all.