Sudden Unexpected Acceleration today

[dsvick]

View attachment 300327 View attachment 300328

Joke #1: Too little coffee... too late...

ohh, come on ... it was right there, you were one letter away, I can't believe you didn't see it ....

Joke #1: Too little coffee... too latte...

 

[scottm]

Regarding the Tesla (looks like a Model S, to me) crash in Los Gatos, Tesla Crashes Into Los Gatos Starbucks says
"The driver is cooperating with a police investigation, she says she thought her car was in park and she does not know what happened."

Forget about having a sound emitter OUTSIDE the car for the benefit of blind pedestrians.... seems like a sound emitter is needed INSIDE the car so driver knows the car is "running" (not in Park).

 

[scottm]

I love thinking funny stuff. Not knowing how/if possible, but couldn't the hack just replace the information coming from the pedal sensors, so any traces/logs would be saved looking like the person was pressing the pedal.

Bingo! Don't need to reverse engineer and rewrite logging files, when it's easier to spoof sensors -- the logs get written correctly for you.

The bonus of spoofing inputs is actually getting the car to do things when it can't discern the physical input as being provided by a person or not... Sure, the car will try to use all available inputs in deducing truth: there might be other necessary conditions, with corresponding sensors...to cross-check. But spoofing a series of inputs, like simulate weight on drivers seat, etc... will eventually get an effect to happen.

Spoofing an input can be done with a physical "fake" of the external stimulus like putting a stick to the go pedal. But maybe also be spoofed electronically by inserting a "man in the middle" attack... taps the sensor wires and mess with the values.

Examining the sensor and seeing what it would take to spoof it reveals a lot. Some important sensors have internal redundancies... drive by wire cars have been around for years. Often, sensors are very simple and simple to spoof, but can contain little complexities to be aware of. Such as there's two potentiometers in the accelerator pedal sensor - ingeniously arranged on the same physical drive axis and mechanism to swipe in tandem but in opposite directions. Therefore, as one value increases the other must decrease at a correspondingly predictable value to be considered a valid input.. any "funny" stuff from one of them and the car determines the pedal is bad, ignores it. Assumes a poor electrical connection, or worn sensor, etc..

You could build a little electronic thing that emulates the sensor but must also consider human factors... like how fast a human can physically act on a sensor. A device can inject legitimate values on the sensor wire that are electrically compatible with the CPU trying to read it. Good software on the reading end of this in the CPU could detect human impossibilities and filter out noise. Such as, sensor values that slide too quickly from one value to another ... faster than humanly possible (e.g. zero to full pedal press in 5 milliseconds).. could cause immediate shutdown, as "untrusted input". Your spoofing device must act smoothly and within the realm of humanly possible.

Once you understand a sensor's design and output you can emulate it. On little sensors there's no intelligence or crypto key protections... it's just raw electrical signaling.

So Jason would have to chuck a brick on the seat as well as jamb a stick to the pedal to get the Tesla to do an unmanned drag race. Or get creative with sensor spoofing.

 

[Runt8]

Bingo! Don't need to reverse engineer and rewrite logging files, when it's easier to spoof sensors -- the logs get written correctly for you.

The bonus of spoofing inputs is actually getting the car to do things when it can't discern the physical input as being provided by a person or not... Sure, the car will try to use all available inputs in deducing truth: there might be other necessary conditions, with corresponding sensors...to cross-check. But spoofing a series of inputs, like simulate weight on drivers seat, etc... will get an effect to happen.

Spoofing an input can be done with a physical "fake" of the external stimulus like that. But maybe also spoofed electronically by inserting a "man in the middle" attack... taps the sensor wires and mess with the values.

Examining the sensor and see what it would take to spoof it. Some important sensors have internal redundancies... drive by wire cars have been around for years. Often, sensors are very simple and simple to spoof, but can contain little complexities to be aware of. Such as there's two potentiometers in the accelerator pedal sensor - ingeniously arranged to swipe in tandem but in opposite directions. Therefore, as one value increases the other must decrease at a correspondingly predictable value to be considered a valid input.. any "funny" stuff from one of them and the car determines the pedal is bad, ignores it. Assumes a poor electrical connection.

But you could build a little electronic thing that emulates the sensor. Injects legit values on the wire that are electrically compatible with the CPU trying to read it. Once you understand a sensor's design and output you can emulate it. On little sensors there's no intelligence or crypto key protections... it's just raw electrical signaling.

So Jason would have to chuck a brick on the seat as well as jamb a stick to the pedal to get the Tesla to do an unmanned drag race. Or get creative with sensors.

If you have physical access then anything is possible...

 

[tpatana]

If you have physical access then anything is possible...

Maybe there was someone else hiding and pressing the accelerator pedal!

Plot thickens...

 

[tpatana]

In the surprise/shock/confusion of the moment she remembers the events differently than they really occurred.

Funny test how bad people are at observing things:

 

[AZM3]

Forget about having a sound emitter OUTSIDE the car for the benefit of blind pedestrians.... seems like a sound emitter is needed INSIDE the car so driver knows the car is "running" (not in Park).

Excellent idea - how about a big block Chevy sound so that a driver can aurally tell the speed of the electric motor when parking. This would definitely reduce pedal misapplication. Heck make it configurable to so that a user can choose what kind of sound they like - Ferrari, Muscle car, F1 car etc.

 

[scottm]

Maybe need a version of this

but needs to work in EV, so would have to key off OBDII port car speed if you want the vroom-vroom of a moving car, instead of microphone because EVs are just too quiet

 

[wk057]

 

[McRat]

Joke #2: Espresso lane

There was no driver. The AutoPilot wanted a double latte:

 

[buttershrimp]

ohh, come on ... it was right there, you were one letter away, I can't believe you didn't see it ....

Joke #1: Too little coffee... too latte...

This particular wreck feels like Deja Brew.

 

[hill]

This is what happens when you have a car that goes full speed too fast for most people to handle.

Operative 2 words; Chill mode
.

 

[Trinity]

An unfortunate incident for sure.

These types of accidents are, unfortunately common - in all cars.

Malcom gladwell did an excellent review of what we know of these crashes and what science has shown about them in the following podcast -

Revisionist History Podcast

Very much worth a listen for all, as this is indeed a safety issue, just likely a different kind than you might be thinking.

 

[rstefani11]

And this is definitely a key point that keeps getting ignored.

The cars have a lot of torque available, and on my bench I actually use the hydraulic brakes as a dummy load when testing motors. However, if the motor isn't moving, it can not break free of a solid brake press. Heck, even the puny pads of the parking brake make it almost impossible for it to break free. The performance motors can break free of the parking brake with extreme difficulty, but not the hydraulic brakes. It just sits there and struggles until the motor throws an error and gives up (few seconds).

Here's a (private) video clip I made a while back of a small rear Model S drive unit on my bench (pretty similar specs to the 3 drive units) trying to break free of just the parking brake, which has wayyyy less force than the hydraulic brakes.

Ramped to full throttle at least once in there, zero wheel movement.

So yeah, sorry... no one's foot was on the brake pedal while the car was quickly accelerating. Just not possible. The accelerator pedal was pressed. User error.

I wish folks would just get over it and move on... but, the internet.

The one experience I can share that will "feel" like an unexpected acceleration is when using regenerative braking at 'normal' level'. As I unfortunately found out driving on some black ice, if regenerative braking is turned on, and a tire detects a slip, it immediately cuts off, which will 'feel' like a surge in speed. Wish I had known this before my car lost control and slid into a police Ford Expedition (this can only happen to me), but that's life. Best recommendation someone had for me was put regen braking in 'low' mode when faced with slippery surfaces

 

[drawfour]

Joke #5: guess he didn't make it to the Espresso-way

Best recommendation someone had for me was put regen braking in 'low' mode when faced with slippery surfaces

Seems like the Tesla should do this automatically... Change regen to low if temperature below a certain threshold, warn the user, and they can override it on the screen by selecting to turn it back to high. Tweet Elon with this safety recommendation, see what he says!

 

[Ciaopec]

After reading @wk057 posts over the years, in my mind there is probably not a more definitive answer to be found including information from Tesla itself. Why this thread continued after the post on results of his private investigation of SUA is beyond me. I am convinced that even if (insert name or your deity, higher being, teen idol, etc, here) posted this information they wouldn't be believed and the argument would go on.

 

[Krazaak]

Excellent idea - how about a big block Chevy sound so that a driver can aurally tell the speed of the electric motor when parking. This would definitely reduce pedal misapplication. Heck make it configurable to so that a user can choose what kind of sound they like - Ferrari, Muscle car, F1 car etc.

Maybe Jaguar wasn't being so stupid adding fake engine noises to the I-Pace.

Tesla could license Nest's smoke alarm warnings... "There is smoke in the entryway... the alarm will sound... the alarm is LOUD"... "You have pressed the accelerator... the car will move forward... the car is QUICK"

 

[jlghertner]

I want to make sure I am understanding your position.

1. Tesla computer malfunctioned and and initiated SUA.
2. Tesla hydraulic brakes malfunctioned and did not slow down the vehicle, but acted as the accelerator, due to computer overriding hydraulic brakes.
3. Tesla needs to drop everything that they are doing so they can prove to you it wasn’t the above issues.
4. NHTSA is going to cover up the facts because they are sick and tired of people complaining about SUA.
5. You refuse to believe operator error happened because, although it happens thousands of times a year, In all types of vehicles, and you weren’t in vehicle when it happened, you would rather believe numbers 1-4 above.

If the above 5 are correct can you please answer who was shooting from the grassy knoll?

Toyota was screwed because of this false
allegation and I assume bias towards US manufacturers.

Check out Malcolm Gladwell’s podcast on this subject for
A nontechnical but excellent psychological understanding of what really happens. Driver error.

 

[Trips]

Funny. It crossed my mind that Tesla should bring you in as an independent subject matter expert - and also get testimony that you already told the plaintiff that it was not possible, but they went ahead with the suit anyway. And then Tesla can attempt to reclaim their costs defending a frivolous suit.

This is would be great especially if we could get video or audio of it. I think we also need wk057 to screen the questions for the quarterly earnings report.

 

[FlatSix911]

The truth will set you free ... restated again for those who missed it the first time

And this is definitely a key point that keeps getting ignored. Here's a (private) video clip I made a while back of a small rear Model S drive unit on my bench (pretty similar specs to the 3 drive units) trying to break free of just the parking brake, which has wayyyy less force than the hydraulic brakes. Ramped to full throttle at least once in there, zero wheel movement.

So yeah, sorry... no one's foot was on the brake pedal while the car was quickly accelerating. Just not possible. The accelerator pedal was pressed. User error.

I wish folks would just get over it and move on... but, the internet.