NHTSA Vehicle Speed Control Complaints

[tahdah793]

We're fairly deep into our exploration of ordering an M3 LR. We still have some reservations related to various threads concerning manufacture defects. But here's the real issue holding us up right now. I saw a couple of articles about run-away acceleration and other speed control problems. So I looked at the NHTSA site and found about 12 complaints for 2019 cars and another 12 on 2020 cars. That in itself isn't a cause for serious concern. But what is of concern is the lack of Tesla's cooperation (reported in the press) with the investigation and, in the absence of a NTHSA ruling what does it mean? My understanding is that Tesla has data on everything that happens in the million cars on the road. If that's right and they are not helping the NHTSA investigators determine user error vs manufacture flaw why is that?

I won't go into the specifics of the cases here but there are some that involved crashes (no lives lost), some that could have involved crashes if the driver hadn't been alert and some were at high speeds (e.g. unexpected, unneeded braking) and some in slow speed turns or parking lot situations.

So, questions:
1. Anyone here know anything about these cases or have had any similar situations where there was sudden unexpected acceleration or deceleration?
2. Any insights on why Tesla might not be cooperating other than the obvious concern above?
3. Any other related thoughts on the subject?

Thanks, potential & otherwise positively inclined owner and forum newbie.

 

[Saghost]

Phantom braking is a real thing on the freeway, though not common and slowly getting rarer. It's not caused by any sort of fault in the system per se - it happens because the Autopilot sensors think they are seeing something dangerous coming up and are deliberately slowing the car to keep you safe - but the sensors are misunderstanding what is happening (often associated with metallic overhead structure that also have shadows on the road.) I expect this to progressively happen less often as Tesla continues improving the neural networks that interpret the visual and radar data.

Almost all complaints about sudden acceleration in low speed turns and parking situations in all car brands are the result of the driver hitting the wrong pedal. There's not much Tesla can do about this, but they did add the optional "obstacle aware acceleration" setting that will greatly reduce the car's response to the accelerator if the car thinks there's something right in front of you. If you hit both pedals at the same time, the car will ignore the accelerator input and beep at you while putting a warning message about pressing both pedals in the instrument cluster.

 

[MasterC17]

Phantom braking is a real thing on the freeway, though not common and slowly getting rarer. It's not caused by any sort of fault in the system per se - it happens because the Autopilot sensors think they are seeing something dangerous coming up and are deliberately slowing the car to keep you safe - but the sensors are misunderstanding what is happening (often associated with metallic overhead structure that also have shadows on the road.) I expect this to progressively happen less often as Tesla continues improving the neural networks that interpret the visual and radar data.

A̶l̶m̶o̶s̶t̶ all complaints about sudden acceleration in low speed turns and parking situations in all car brands are the result of the driver hitting the wrong pedal. There's not much Tesla can do about this, but they did add the optional "obstacle aware acceleration" setting that will greatly reduce the car's response to the accelerator if the car thinks there's something right in front of you. If you hit both pedals at the same time, the car will ignore the accelerator input and beep at you while putting a warning message about pressing both pedals in the instrument cluster.

Fixed it for ya

But in all seriousness a Tesla cannot accelerate on its own without user input. If you do some further research there is a multitude of documentation covering this. The recent allegations brought against Tesla were by a short-seller who was incentivized to see the stock drop and does not own a Tesla; not exactly a reliable source.

If you press the brake pedal and also try to accelerate the car will not move.

 

[Orwell]

Good read here regarding unintended acceleration.

Tesla Hacker Says Unintended Acceleration Is Impossible In Teslas

Phantom braking is real, but it's rare and with each software update there seems to be less occurrence. There have been some scary moments for me, but even with that I would much rather have Autopilot.

 

[ewoodrick]

Fixed it for ya

But in all seriousness a Tesla cannot accelerate on its own without user input. If you do some further research there is a multitude of documentation covering this. The recent allegations brought against Tesla were by a short-seller who was incentivized to see the stock drop and does not own a Tesla; not exactly a reliable source.

If you press the brake pedal and also try to accelerate the car will not move.

You really have to be careful with statements like this, as they are just plain wrong.

If I'm at a stop and I've got TACC on, the car will absolutely accelerate on it's on.
Now, in these cases, it was indeed something else.

 

[ewoodrick]

We're fairly deep into our exploration of ordering an M3 LR. We still have some reservations related to various threads concerning manufacture defects. But here's the real issue holding us up right now. I saw a couple of articles about run-away acceleration and other speed control problems. So I looked at the NHTSA site and found about 12 complaints for 2019 cars and another 12 on 2020 cars. That in itself isn't a cause for serious concern. But what is of concern is the lack of Tesla's cooperation (reported in the press) with the investigation and, in the absence of a NTHSA ruling what does it mean? My understanding is that Tesla has data on everything that happens in the million cars on the road. If that's right and they are not helping the NHTSA investigators determine user error vs manufacture flaw why is that?

I won't go into the specifics of the cases here but there are some that involved crashes (no lives lost), some that could have involved crashes if the driver hadn't been alert and some were at high speeds (e.g. unexpected, unneeded braking) and some in slow speed turns or parking lot situations.

So, questions:
1. Anyone here know anything about these cases or have had any similar situations where there was sudden unexpected acceleration or deceleration?
2. Any insights on why Tesla might not be cooperating other than the obvious concern above?
3. Any other related thoughts on the subject?

Thanks, potential & otherwise positively inclined owner and forum newbie.

My understanding is that Tesla is working with the NHTSA because the NHTSA hasn't actually opened an investigation. Tesla hasn't been asked to investigate.
I always enjoy watching some of the various "news" outlets during the day, Somebody breaks the story, then someone copies it, but has to change the wording to avoid plagiarizing it. Then the next outlet, and then the next. After 4 or 5 outlets copy it, so much of the message has been changed.

More importantly, how many store fronts have been run into in your area this week? It is a regular occurrence for all cars. And these are just about always determined to be user error. Someone is using their accelerator to move slowly, another car appears and they hit their brakes, whoops, the foot slipped of the brake onto the accelerator.

The situation was widely reported, but the reality of what was happening was barely reported. As have been said, it was a short seller trying to manipulate the stock to make some money. You can see Tesla's reaction on their blog.

 

[derotam]

But in all seriousness a Tesla cannot accelerate on its own without user input

You really have to be careful with statements like this, as they are just plain wrong.

If I'm at a stop and I've got TACC on, the car will absolutely accelerate on it's on.

@ewoodrick But setting TACC is a user input, so the car doesn't accelerate on it's own in error, it accelerates as directed to by past user input.

An unintended acceleration complaint has to be based on the car accelerating in error and in contradiction to the users input.

But what is of concern is the lack of Tesla's cooperation

Tesla didn't cooperate with an agreement about talking to the media and making public statements about investigations where they signed an agreement with NHTSA. I believe they did this in response to NHTSA making a public statement that Tesla believed was in error. Tesla still cooperated with NHTSA in regards to data that NHTSA requested under their authority.

 

[tahdah793]

I appreciate the responses so far but dismissing the complaints as a bunch of short sellers doesn't work at nhtsa.gov. To file a complaint you need to provide a VIN. So if it's a short-seller, they own a Tesla. Maybe shelling out +- 50K is a small price to try to cover a short-sale but it doesn't seem to work very well so it might be more effective to use a more complicated option than just sell a call or buy a put.

And now related of my original questions: I have no doubt that the documentation indicates that it's impossible for the car to accelerate without petal input. The question at hand is: is it possible for there to be an undocumented bug...of course it is. I've been in the computer industry for years and bugs exist. I'm not a AI expert but it's clear to me that machine learning can produce wrong behaviors aka a bug. Witness the phantom deceleration response above. So, nhtsa's job is to find out if these are user errors as the responses so far indicate or is it a system problem. So, unless Tesla cooperates with nhtsa by providing data what conclusion should I come to in the absence of a ruling from nhtsa? I woiuld like to make my purchase decision sooner rather later so I'm wondering if there's a legitimate reason why Tesla would not be cooperating with nhtsa?

Thanks again, prospective owner, not a short seller or troll.

 

[Msjulie]

Nothing built by humans will be 100% perfect. I still believe the report is unfounded but I can't give you any specific guarantee. It's easy to find VINs online BTW so I don't see much credit in that as the guy who filed it was indeed a documented aggressive short-seller

I'm in the bay area so Tesla's presence here is pretty high. Aside from the oft-reported issues related to some aspects of build quality (which are improving and even my 2018 is fine) consider how popular the car is before being swayed too heavily by potential click bait.

 

[derotam]

To file a complaint you need to provide a VIN. So if it's a short-seller, they own a Tesla.

Not that hard to get a random VIN from a salvage vehicle up for auction, used car sales ad, etc, etc.

I'm not a AI expert but it's clear to me that machine learning can produce wrong behaviors aka a bug. Witness the phantom deceleration response above.

"Phantom deceleration" isn't necessarily a "bug" per se.

I'm wondering if there's a legitimate reason why Tesla would not be cooperating with nhtsa?

Define cooperating....Are you trying to say that Tesla is refusing to provide ANY data with NHTSA? That would be unlikely. Usually what happens with investigations is that NHTSA attempts to sign in to an agreement with the manufacture in regards to the investigation. Part of this agreement is that the manufacturer does not publicly release certain information related to the investigation during the investigation. I believe the manufacture, whether there is an agreement or not, is REQUIRED to provide NHTSA with certain investigation related data.

 

[Cycle11111]

Careful for sure. A significant number of the complaints were filed by a short seller on behalf of multiple Tesla owners even though they do not own a Tesla themselves. Tesla did co-operate with NHTSA wrt the investigation and none of the incidents could be traced to a vehicle fault. Teslas are not perfect for sure but they have many layers of protection - I for one have tried the press the brake and accelerator and the brake will overpower the Go peddle and if you press both while at a stand still the car simple doesn't move and a warning shows both are being pressed. Also note the brake pedal (like other mftgs) is closer to the driver than the accelerator to help avoid unintended application of the accelerator.

 

[MasterC17]

You really have to be careful with statements like this, as they are just plain wrong.

If I'm at a stop and I've got TACC on, the car will absolutely accelerate on it's on.
Now, in these cases, it was indeed something else.

You contradicted yourself. If you input TACC, the output is that it will accelerate on its own. I stated the vehicle will not accelerate without user input. TACC is user input.

 

[MasterC17]

I appreciate the responses so far but dismissing the complaints as a bunch of short sellers doesn't work at nhtsa.gov. To file a complaint you need to provide a VIN. So if it's a short-seller, they own a Tesla. Maybe shelling out +- 50K is a small price to try to cover a short-sale but it doesn't seem to work very well so it might be more effective to use a more complicated option than just sell a call or buy a put.

And now related of my original questions: I have no doubt that the documentation indicates that it's impossible for the car to accelerate without petal input. The question at hand is: is it possible for there to be an undocumented bug...of course it is. I've been in the computer industry for years and bugs exist. I'm not a AI expert but it's clear to me that machine learning can produce wrong behaviors aka a bug. Witness the phantom deceleration response above. So, nhtsa's job is to find out if these are user errors as the responses so far indicate or is it a system problem. So, unless Tesla cooperates with nhtsa by providing data what conclusion should I come to in the absence of a ruling from nhtsa? I woiuld like to make my purchase decision sooner rather later so I'm wondering if there's a legitimate reason why Tesla would not be cooperating with nhtsa?

Thanks again, prospective owner, not a short seller or troll.

That NHTSA Tesla "Sudden Unintended Acceleration" Petition? Created By A TSLA Short Seller Who Doesn't Own A Tesla | CleanTechnica

 

[tahdah793]

Not that hard to get a random VIN from a salvage vehicle up for auction, used car sales ad, etc, etc.


"Phantom deceleration" isn't necessarily a "bug" per se.



Define cooperating....Are you trying to say that Tesla is refusing to provide ANY data with NHTSA? That would be unlikely. Usually what happens with investigations is that NHTSA attempts to sign in to an agreement with the manufacture in regards to the investigation. Part of this agreement is that the manufacturer does not publicly release certain information related to the investigation during the investigation. I believe the manufacture, whether there is an agreement or not, is REQUIRED to provide NHTSA with certain investigation related data.

Nothing built by humans will be 100% perfect. I still believe the report is unfounded but I can't give you any specific guarantee. It's easy to find VINs online BTW so I don't see much credit in that as the guy who filed it was indeed a documented aggressive short-seller

I'm in the bay area so Tesla's presence here is pretty high. Aside from the oft-reported issues related to some aspects of build quality (which are improving and even my 2018 is fine) consider how popular the car is before being swayed too heavily by potential click bait.

Thanks for the reply. I agree there's no perfect software out of the gate and machine learning is still in it's infancy. But there's no such thing as a random VIN. It's a structured string that includes manufacturer ID, model, other attributes, e.g.AWD, model year, etc. and a sequence number assigned by the manufacturer. If you're suggesting counterfeit VINs that would be a problem for Tesla that they would no doubt address.

I don't have a definition of "cooperating" it's just what the article said. Maybe it's dragging their feet, maybe nhtsa has asked for a lot of information that isn't as readily available, or, just maybe they are counterfeit VINs and they haven't told nhtsa yet. You'd think they'd want that to be known.

And, unless all of the 20+ individual complaints are all fraudulent and submitted by this documented, infamous short-seller (my trades aren't public information), it seems to me that there are real people driving real Tesla 3 cars that they perceive to have a problem. It could be user error in all of the cases if so I would expect Tesla to be eager to provide the data. In any case I don't know that yet. So, maybe the right answer for me is to wait and see what nhtsa says 6mo to a year from now but I was hoping to get some insights that would help us to decide to order sooner than later.

 

[tahdah793]

@ewoodrick But setting TACC is a user input, so the car doesn't accelerate on it's own in error, it accelerates as directed to by past user input.

An unintended acceleration complaint has to be based on the car accelerating in error and in contradiction to the users input.



Tesla didn't cooperate with an agreement about talking to the media and making public statements about investigations where they signed an agreement with NHTSA. I believe they did this in response to NHTSA making a public statement that Tesla believed was in error. Tesla still cooperated with NHTSA in regards to data that NHTSA requested under their authority.

This is helpful. I'll chase it down. Thanks.

 

[MasterC17]

Thanks for the reply. I agree there's no perfect software out of the gate and machine learning is still in it's infancy. But there's no such thing as a random VIN. It's a structured string that includes manufacturer ID, model, other attributes, e.g.AWD, model year, etc. and a sequence number assigned by the manufacturer. If you're suggesting counterfeit VINs that would be a problem for Tesla that they would no doubt address.

I don't have a definition of "cooperating" it's just what the article said. Maybe it's dragging their feet, maybe nhtsa has asked for a lot of information that isn't as readily available, or, just maybe they are counterfeit VINs and they haven't told nhtsa yet. You'd think they'd want that to be known.

And, unless all of the 20+ individual complaints are all fraudulent and submitted by this documented, infamous short-seller (my trades aren't public information), it seems to me that there are real people driving real Tesla 3 cars that they perceive to have a problem. It could be user error in all of the cases if so I would expect Tesla to be eager to provide the data. In any case I don't know that yet. So, maybe the right answer for me is to wait and see what nhtsa says 6mo to a year from now but I was hoping to get some insights that would help us to decide to order sooner than later.

I don't think you are going to get anything better than this: Tesla Hacker Says Unintended Acceleration Is Impossible In Teslas

“The drive units contain at least three distinct pieces of hardware that all simultaneously crosscheck any pedal input. The pedal hardware itself contains two independent sensors that must agree as well. This way, only a genuine pedal input will cause torque to be commanded. There are additional crosschecks of this in other modules as well, such as the ABS.

While Tesla is far from perfect, the powertrain control setup, specifically related to these sudden acceleration claims, is something they deserve a lot of credit on. It's a solid system, with a ton of thought put into how to make it as safe as possible. There's plenty of things to call out where Tesla has done something outright stupid, or dare I say it unethical, over the years... But this just isn't one of them. They've done this right.”

Tesla has sold about 500,000 Model 3's. Using 20 "possible" examples of a defect to influence your buying decision seems a bit strange. People claim unintended acceleration across every manufacturer and every model. This has been going on for decades and will likely continue. Pedal confusion is real. When people press the accelerator thinking it is the brake their natural instinct is to press it harder.

 

[tahdah793]

That NHTSA Tesla "Sudden Unintended Acceleration" Petition? Created By A TSLA Short Seller Who Doesn't Own A Tesla | CleanTechnica

I'm not sure what a nhtsa petition is. What I see are "complaints" that are tied to a specific vehicle and a specific situation. The article does have an apology associated with it so it's discarded input as far as I'm concerned.

Thanks.

 

[derotam]

Thanks for the reply. I agree there's no perfect software out of the gate and machine learning is still in it's infancy. But there's no such thing as a random VIN. It's a structured string that includes manufacturer ID, model, other attributes, e.g.AWD, model year, etc. and a sequence number assigned by the manufacturer. If you're suggesting counterfeit VINs that would be a problem for Tesla that they would no doubt address.

I don't have a definition of "cooperating" it's just what the article said. Maybe it's dragging their feet, maybe nhtsa has asked for a lot of information that isn't as readily available, or, just maybe they are counterfeit VINs and they haven't told nhtsa yet. You'd think they'd want that to be known.

And, unless all of the 20+ individual complaints are all fraudulent and submitted by this documented, infamous short-seller (my trades aren't public information), it seems to me that there are real people driving real Tesla 3 cars that they perceive to have a problem. It could be user error in all of the cases if so I would expect Tesla to be eager to provide the data. In any case I don't know that yet. So, maybe the right answer for me is to wait and see what nhtsa says 6mo to a year from now but I was hoping to get some insights that would help us to decide to order sooner than later.

Don't get the car, it's not for you.

You didn't read the rest of the line where I said "random VIN". Please don't mis-represent what I said. Random did not equal counterfeit in my context.

 

[lbowroom]

If you’re worried about people’s perception of problems that aren’t real.... don’t buy any car

 

[Msjulie]

I googled "Tesla Model 3 VIN" and got this: Car report for 5YJ3E1EA5KF328931 VIN number

I have no idea whose car it is nor do I care

A VIN is a VIN

Thanks for the reply. I agree there's no perfect software out of the gate and machine learning is still in it's infancy. But there's no such thing as a random VIN.