Sudden Unintended Acceleration

[D.E.]

I have big feet. My Honda minivan engine has raced when I intended to stop.

The problem was that my shoe bridged both pedals so I was applying the brake but was unintentionally pressing the accelerator at the same time.

I think this may be he mechanism for some unintentional acceleration cases, maybe all. It was disturbing the first time it happened. It has happened again since bit now I know exactly what’s going on so it isn’t nearly as disturbing as it was the first time.

With the Honda, the engine makes a lot of noise (engine racing) but the brake wins. If the pedal heights are different and/or if the car is more powerful, maybe the brake doesn’t win. If the accelerator can be depressed without much braking, the car might move aggressively forward while the operator thinks he’s only pressing the brake pedal. That would also explain the car’s data recordings.

I should try it in the Tesla on a deserted street.

Best,
David

 

[croman]

Tesla will notice both pedals being pressed and issue a warning error message.

 

[dennis_d]

This here is where Tesla proving the actual logs would help in any case. A verbal explanation alone can be more easily misunderstood and misreported onwards.

I agree it is a curious co-incidence that the number happened to be 18%, the same as the cruise speed limit of 18 mph. But of course it may be just a co-incidence. (It did not originally occur to me to look at this number, since the equivalent number in kph is different, 30.)



If the driver was slowing down by regen (very common, yes?), enabling cruise certainly can result in feeling symptoms similar to SUA. I don't know if this happened in @Ying's case, but it has IMO happened in TMC reported cases in general.

If he wants logs he need to act pretty fast because car keep only two weeks of logs you can pull.

 

[D.E.]

Tesla will notice both pedals being pressed and issue a warning error message.

So are you saying the car will not accelerate, or that it will but displays a message?

 

[Chaserr]

You get a moment of acceleration and then I think it cuts off, or the brakes simply overpower it. I tested it long ago on purpose, but can't recall exactly. The three-chirp alert and message popup in the IC are obvious though.

 

[croman]

You get a moment of acceleration and then I think it cuts off, or the brakes simply overpower it. I tested it long ago on purpose, but can't recall exactly. The three-chirp alert and message popup in the IC are obvious though.

I've never tested it accidentally or on purpose but I believe it cuts off power as described above.

 

[Az_Rael]

Agreed. I've never ever seen anyone drive a non-manual where they used their left foot to brake except in some very rare instances (e.g. to prevent rollback on a hill).

I remember there was some thread here on TMC where people were arguing about the "merits" of driving an automatic w/two feet (left foot braking, right foot accelerating). Two foot driving I don't think was it.

Until I saw the debate thread, it hardly registered for me that there are actual people who do that. I suspect the reaction from most automatic drivers is of bewilderment and shaking their head, which I've witnessed.

On this note, I think i've posted about this several times. Some automakers (e.g. Nissan) won't even let you resume (CC) unless your speed is above 25 mph.

And, IIRC, the behavior on Toyotas (e.g. my 06 Prius) is even more conservative/annoying. If you drop below 25 mph, the cruise control intentionally forgets the set speed, so resume won't work anyway.

If Tesla implemented these, it seems that could prevent some of these incidents.

Only downside of this is it will prevent engaging TACC in stop and go traffic. I sometimes need to disengage TACC in stop and go to let folks merge in from a closing lane (the car won’t leave extra space). But then I do want to resume TACC even if I am driving less than 25.

 

[wk057]

I would find it amusing if Tesla released the raw logs.

Literally no one outside of Tesla that I'm aware of, besides myself, has made any significant progress decoding any actual information from their proprietary log format. It's a very packed full binary log, but contains loads of data. At ~66k miles my car only has collected a few GB of logs, but those logs go back to mile 0 and are very detailed. I still have not deciphered 100% of the data in them, but I managed to correlate a lot of things once I realized how exactly they were handling the capture and logging itself.

My point is that the logs are basically useless by themselves. People are questioning Tesla's interpretation of the logs, for some reason, yet you're not going to want the raw log from Tesla... you're going to want a parsed/interpreted human-manageable version, not the binary blob of data the car actually saves and distributes to Tesla.

So either way, you're relying on Tesla to provide the relevant data (which they seem to already have done).

 

[BigD0g]

I would find it amusing if Tesla released the raw logs.

Literally no one outside of Tesla that I'm aware of, besides myself, has made any significant progress decoding any actual information from their proprietary log format. It's a very packed full binary log, but contains loads of data. At ~66k miles my car only has collected a few GB of logs, but those logs go back to mile 0 and are very detailed. I still have not deciphered 100% of the data in them, but I managed to correlate a lot of things once I realized how exactly they were handling the capture and logging itself.

My point is that the logs are basically useless by themselves. People are questioning Tesla's interpretation of the logs, for some reason, yet you're not going to want the raw log from Tesla... you're going to want a parsed/interpreted human-manageable version, not the binary blob of data the car actually saves and distributes to Tesla.

So either way, you're relying on Tesla to provide the relevant data (which they seem to already have done).

Can open source your parser so the community can add to it, and we can all benefit and mitigate the need for Tesla to interpret the logs, along the same lines of the reverse engineering work done for the HPWC for solar installations.

 

[RAW84]

I've made offers to people several times over cases where people were trying to blame Tesla for user error issues or felt that an autopilot airbag triggered recording would exonerate them from something.

My offer is always the same: You provide me access to the vehicle, I'll pull the logs and parse the relevant data. I'll then share the results, regardless of what they end up revealing, with you, Tesla, your insurance company, and I reserve the right to post the data wherever else it may be relevant.

Thus far only one person decided to take me up on this, then got cold feet at the 11th hour thinking it'd be best not to do so. I'm reasonably certain that these people just didn't want to have confirmation that they'd screwed up.

Every case of "sudden unintended acceleration" with a Tesla is driver error. Period. There is no way for the vehicle to accelerate on its own like people claim. It's also always pedal misapplication, too, where the driver presses the accelerator when they should be braking. Almost always cases where the car is slowing to a stop, then "suddenly" accelerates (because the driver hit the accelerator instead of the brake at the time they would be hitting the brake to stop the car).

I've pulled logs from at least two cars that, when looking back at news and posts, claimed unintended acceleration. The logs in both cases clearly showed the driver applying the accelerator pedal at the time of the accident. Electrek did an article about this: Several Tesla owners claim Model X accelerated/crashed on its own but everything points to user error

The vehicle logs the outputs of both hall effect sensors in the accelerator pedal independently. They both must match their respective output curves during a go-pedal press in order for the car to respond to a request for acceleration. If anything is off, the car doesn't move. If one sensor goes out, the car will operate in limp mode with drastically reduced torque.

Suffice it to say, there quite literally is no way for a Tesla Model S/X/3 to do what people claim without the driver pressing the accelerator pedal.

I had a case of unintended acceleration once, but I wasn't parking and I know I hit the wrong pedal, just not with the amount of force that should have caused the car to accelerate as much as it did.

I was stopped at a red light a couple of cars back. This was after TACC but prior to hill hold and autosteer. During that time I used to use TACC to hold the brake for me at red lights. When there light would turn green I'd tap the brake to disengage TACC so I could be prepared to go when the car in front of me moved. But I tapped the accelerator by mistake and this caused the car to accelerate violently as if the car ahead of me was no longer there and it was trying to get up to 30 MPH (or whatever it was set at) asap. Thankfully, I had left enough space and was able to stop the car before it hit the car in front of me, but I've always wondered if the logs would've been able to show how much I depressed the accelerator in that instance. Do you think the logs would've provided clarity in my case?

 

[bmah]

My point is that the logs are basically useless by themselves. People are questioning Tesla's interpretation of the logs, for some reason, yet you're not going to want the raw log from Tesla... you're going to want a parsed/interpreted human-manageable version, not the binary blob of data the car actually saves and distributes to Tesla.

Thanks for that. I would speculate that "people" were expecting that the car saved something like semi-human-readable yet low-level (think "syslog") messages which they could interpret themselves, to avoid any particular viewpoint that Tesla might intentionally or unintentionally inject into whatever they released.

Admittedly that's what I envisioned, although I'm interested just for curiosity's sake.

Bruce.

PS. Sorry for straying off-topic, sigh!

 

[AnxietyRanger]

I would find it amusing if Tesla released the raw logs.

Literally no one outside of Tesla that I'm aware of, besides myself, has made any significant progress decoding any actual information from their proprietary log format. It's a very packed full binary log, but contains loads of data. At ~66k miles my car only has collected a few GB of logs, but those logs go back to mile 0 and are very detailed. I still have not deciphered 100% of the data in them, but I managed to correlate a lot of things once I realized how exactly they were handling the capture and logging itself.

My point is that the logs are basically useless by themselves. People are questioning Tesla's interpretation of the logs, for some reason, yet you're not going to want the raw log from Tesla... you're going to want a parsed/interpreted human-manageable version, not the binary blob of data the car actually saves and distributes to Tesla.

So either way, you're relying on Tesla to provide the relevant data (which they seem to already have done).

A printout of a machine-parsed version of the log (which Tesla is bound to have a piece of software to do) would still be better than nothing.

Of course a third-party black box would be even better, but that is for the future.

I mean it's not like having access to the logs of a device you own is somehow too much to ask anyway?

I wonder if GDPR will actually force Tesla to do this in Europe?

 

[AnxietyRanger]

Thanks for that. I would speculate that "people" were expecting that the car saved something like semi-human-readable yet low-level (think "syslog") messages which they could interpret themselves, to avoid any particular viewpoint that Tesla might intentionally or unintentionally inject into whatever they released.

A binary log is nothing special, though.

All it means is somewhere at mothership there is a software/server that deciphers it into a printout. Something that a Service Center certainly could likely get their hands on and provide a customer.

 

[pedriscoll]

I had a similar incident last June. I was pulling into a parking space and the car “unexpectedly” accelerated and when I tried to brake, accelerated faster, jumped over the concrete barrier and hit a large bush. No building thankfully. As I ultimately reconstructed the accident in my mind later, I realized that I was utilizing 1 pedal driving and had my foot on the accelerator and not the brake. When I pushed the pedal on “final approach” to the parking space, the car accelerated and when I hit the pedal again to stop (thinking it was the brake), it accelerated even faster with the resulting accident. At the time I swore that the car did this on its own, but in retrospect realized it was driver error.

It was a $4000 repair job to fix the barely visible damage to the front undercarriage. Took 3 months to get some of the parts, but luckily the body shop was able to do temporary repairs to make it drivable while waiting for the other parts. It was an expensive lesson to me, but I now make absolutely sure I know which pedal I am using at all times.
I suspect that the OP had a similar course of events.

 

[SucreTease]

I had a similar incident last June....

Can we now admit that this problem is common enough with Teslas (due to their potential instant high acceleration) that some kind of technological protection against it ought to be made available as an option to reduce the likelihood of this problem?

(jump ahead to the 50-second mark)

I propose that it be made a configurable option setting to greatly lessen the instant acceleration available while moving below a certain speed when an obstruction is detected (by appropriate sensors) in front of the the vehicle. One possibility is sounding the same three-beep alarm that precedes a potential collision, followed by approximately a two-second delay before allowing a gentle acceleration.

 

[Zythryn]

Can we now admit that this problem is common enough with Teslas (due to their potential instant high acceleration) that some kind of technological protection against it ought to be made available as an option to reduce the likelihood of this problem?
...

No, "we" can't.
We have anecdotal evidence of some people doing this.
Some people also do this with other cars.

For me to agree, you would have to show me that the rate of Tesla drivers doing this is substantially higher than that of other vehicle drivers.

I'm not saying your conclusion is wrong, I am just saying it is unsupported by the evidence.

Here is an article suggesting cars without creep may be safer than those with.
Why pedal mixups are dangerously common

 

[croman]

@SucreTease - I don't mind an option that users who fear they will have issues can enable but I'd rather Tesla focus on bigger issues like EAP/FSD failures thus far.

 

[SucreTease]

@SucreTease - I don't mind an option that users who fear they will have issues can enable but I'd rather Tesla focus on bigger issues like EAP/FSD failures thus far.

That response can be used to exclude implementation of any feature or improvement except EAP/FSD. Is that what you are arguing for?

 

[croman]

That response can be used to exclude implementation of any feature or improvement except EAP/FSD. Is that what you are arguing for?

Pretty much.

 

[Chaserr]

There are slower EVs out there, and there are slower modes available in tesla cars. That instant responsive throttle is what sold me on my car, and many others like me. Crippling the car on purpose seems as ridiculous to me as pouring gasoline into it.

If I make a mistake, I own up to it. I definitely don't blame the world for my errors.