X owner claims unintended acceleration caused accident

[lklundin]

The final outcome will be determined in a court of law or insurance companies, but if I was a betting person, and I do like watching Forensic Files BTW, I might lean towards siding with Tesla on this one.

Playing the devil's advocate here (and apologies if the question has already been answered elsewhere):
Since it is clearly in Tesla's interest that this comes down to a case of "driver pedal misapplication" and since they quote their sensor logs to support their version, Tesla would have an incentive to falsify these sensor logs.

Are these logs in any way tamper proof so for example NTSB could convince themselves (and the courts) that the sensor logs reflect actual measurements?

 

[Lloyd]

I thought someone had ran into, and tested, that if you apply the brakes and then press the accelerator that the car will accelerate against the brakes. (But if you go the other way around it stop accelerating.)

I have not found that, but I will test that on my way home today.

 

[int32_t]

Isn't it circular reasoning to consult the vehicle logs in a situation like this? Of course the vehicle did what it thought it was told, and of course it records what it thought it was told. The real question is where was the driver's foot at that time? Granted this is very unlikely to happen because there are two throttle position sensors to provide redundancy.

If the logs show some type of ramp to 100%, I would suspect user error. However, if it's just from 0-100% with no ramp, I would suspect the machine.

 

[SR22pilot]

Isn't it circular reasoning to consult the vehicle logs in a situation like this? Of course the vehicle did what it thought it was told, and of course it records what it thought it was told. The real question is where was the driver's foot at that time? Granted this is very unlikely to happen because there are two throttle position sensors to provide redundancy.

There is a big difference between what was happening and what was commanded. The logs should show both. Also, a lot of basic log information can severely limit what might have happened. For example, I can look at a digital data stream and see what a servo was commanded to do. I can look at an encoder on the shaft to see what it did. One is what happened. The other is what was commanded. I suspect Tesla has both pieces of data.

In the early days of Cirrus Aircraft an owner tried to claim Cirrus was at fault for a brake fire that destroyed the plane while taxiing. The owner didn't think about what information could be pulled from the plane. Engine monitor records showed engine RPM well above the max 1000 RPM for taxiing. The owner had stated he was below 1000 RPM. GPS data combined with information on the slope of the runway was used to determine that the pilot had the brakes applied as he taxied. Otherwise the taxi speed would have been much higher considering the excessive RPM. That is why the brakes overheated.

I see this incident as similar to the Cirrus one. The driver can say what she will but data will win out. There won't be just one piece. There will be a consistent set of data that leads to only one conclusion.

 

[alseTrick]

cool story. Don't buy one then.

You can wait till a fully autonomous car comes in in 5 years...till then, this is what you have. If you don't think the technology is good enough, again....don't buy one.

What a weak response.

 

[int32_t]

There is a big difference between what was happening and what was commanded. The logs should show both.

But Tesla is saying that the logs agree: what was commanded did happen. This is why it's circular.

 

[Dr ValueSeeker]

I thought someone had ran into, and tested, that if you apply the brakes and then press the accelerator that the car will accelerate against the brakes. (But if you go the other way around it stop accelerating.)

Happened to David Noland of greencarreports.com, back in 2014. Remember, the last bug in any software is never found.
Life With Tesla Model S: UPDATE On Pedal Placement Problem

 

[animorph]

I assume the logs also show no brake pedal press. That would tend to confirm the accelerator log data was real and not just a glitch. A software or hardware glitch from the accelerator could conceivably cause acceleration. But on top of that should have been data showing a brake pedal application as the car entered the parking space or started acceleration. Highly unlikely both the accelerator and the brake data logging screw up at the same time, in a manner consistent with the accident, without some other noticeable logging corruption.

 

[alseTrick]

What a weak response.

What kind of "active braking assist" feature DOESN'T actually brake when there is an object directly in front of your car? THAT is a problem and one I hope Tesla is working on to fix immediately.

I don't give two craps about "autopilot". Active braking assist has nothing to do with autopilot. It is an entirely separate option that all automakers offer now and I'm pretty sure the auto industry would agree that your ABA is faulty if it lets you smash into a railroad crossing gate.

 

[ecarfan]

Tesla calls it Automatic Emergency Braking, AEB. It is not called "active braking assist". Here is a description of it from the software V6.2 release notes, quote:

Automatic Emergency Braking

Automatic Emergency Braking — a new Collision Avoidance Assist feature — is designed to automatically engage the brakes to reduce the impact of an unavoidable frontal collision.

Automatic Emergency Braking will stop applying the brakes when you press the accelerator pedal, press the brake pedal, or sharply turn the steering wheel.

Automatic Emergency Braking is enabled by default. You can temporarily disable this feature via the AUTOMATIC EMERGENCY BRAKING setting in Controls > Settings > Driver Assistance > COLLISION AVOIDANCE ASSIST. Automatic Emergency Braking will re-enable when you next drive.

Note: Automatic Emergency Braking operates when you are driving at speeds between 5 mph (8 km/h) and 85 mph (140 km/h).
-----------------------------------------------------------------------------------------------------------------------------------------------

So now you have some information about how Tesla's AEB works.

 

[Scannerman]

Someone was late for their manicure appointment.

Your attempt at humor is badly misplaced.

Scannerman

 

[yujsly]

Well, his full quote is:


Which is relevant, because if the accelerator pedal is physically depressed via electronic control (like in my old Toyota), then a log entry which shows whether the accelerator is pressed wouldn't be terribly useful - since cruise control could be the one pressing it.

In that case you would need to know instead whether cruise control wasn't engaged. However, that would be saying: "We have NO log entry showing that cruise control was engaged", rather than saying: "We have a log entry showing the accelerator was pressed".

However, a "NO log entry" is a problem, since a software bug can cause a missing entry. That's why the positive log entry is much better - it would be darn unlikely for a software bug to create phantom log entries.


Since Tesla was stating that they do indeed have a positive log entry showing that the accelerator was moved, the guy had a right to ask if there are non-human forces that can move the accelerator.

I don't think there is (I think cruise control don't physically move the accelerator), but it would be nice if Tesla were to reply officially.

Agree with you! Also, It is a prudent attitude to assume that programmers and drivers are both human and are subject to human errors. So, any analysis should not be based on the log alone.

 

[SR22pilot]

But Tesla is saying that the logs agree: what was commanded did happen. This is why it's circular.

I think calling it circular is misleading. I expect the logs are consistent.

 

[SR22pilot]

Agree with you! Also, It is a prudent attitude to assume that programmers and drivers are both human and are subject to human errors. So, any analysis should not be based on the log alone.

When you interview a group of eye witnesses you look for a common thread. With the logs you would look for consistency. As mentioned in another post I doubt the logs would screw up in a way that was consistent. To have that happen would require an active conspiracy on the part of the programmers.

Some of this depends on how detailed the logs are which we don't know at this time.

 

[SR22pilot]

There have been some comments about the detection of stopped objects. The problem is more difficult than it may appear at first. When parking I think Tesla relies on the ultrasonic detectors. The problem is that they can miss things like low trailers or low overhangs. They can also have issues with poles and mesh items like chain link fences. The radar unit used on the road has its own issues. There are lots of reflections from the road surface. It appears that Tesla eliminates some stationary reflections to remove signal clutter. What follows is my guess at what might happen and should not be taken as fact. If a car in front is moving it generates a signal that stands out from ground clutter. If that car stops then the unit can track the slowing object and see that it stops. There may be a problem detecting stationary objects with the radar when those objects never move so there is never an initial differentiated signal to lock on to. An example of this would be the case where you are following a car which changes lanes to reveal a stopped car in front of the vehicle being tracked. The stopped vehicle might be viewed as ground clutter and ignored.

Now consider the opposite problem. It Tesla thinks ground clutter is a stopped object then it applies the brakes at an incorrect time. This is why Tesla turns off the braking if the driver pushes on the accelerator, sharply turns the steering wheel or applies the brakes. Each of thos actions is taken as the driver sayin "I know what to do." Not having this behavior would cause a lot of problems. Imagine a sharp turn in the road and the car seeing a large roadside object and coming to a stop despite the driver's actions.

 

[WABarb]

I think many people are missing the point that @ecarfan made about AEB

Tesla calls it Automatic Emergency Braking, AEB. It is not called "active braking assist". Here is a description of it from the software V6.2 release notes, quote:

Automatic Emergency Braking

Automatic Emergency Braking — a new Collision Avoidance Assist feature — is designed to automatically engage the brakes to reduce the impact of an unavoidable frontal collision.

Automatic Emergency Braking , will stop applying the brakes when you press the accelerator pedal, press the brake pedal, or sharply turn the steering wheel.

Automatic Emergency Braking is enabled by default. You can temporarily disable this feature via the AUTOMATIC EMERGENCY BRAKING setting in Controls > Settings > Driver Assistance > COLLISION AVOIDANCE ASSIST. Automatic Emergency Braking will re-enable when you next drive.

Note: Automatic Emergency Braking operates when you are driving at speeds between 5 mph (8 km/h) and 85 mph (140 km/h).
-----------------------------------------------------------------------------------------------------------------------------------------------

So now you have some information about how Tesla's AEB works.

 

[lklundin]

Your attempt at humor is badly misplaced.

Considering that the only injury was to a husband's pride, I think it is OK. (This is not even considering that he most likely is not speaking the truth and is now publicly getting a lesson in the capability of a 21st century vehicle).

I will go on to speculate that the pedal misapplication could be explained not by the urgency of a manicure, but rather a pedicure...

To be serious again, a staggering number of people get killed due to traffic, so I cannot even imagine the amount of media attention we will see once the inevitable happens, that a car on autopilot gets implicated in a traffic fatality.

 

[Drivin]

I cannot even imagine the amount of media attention we will see once the inevitable happens, that a car on autopilot gets implicated in a traffic fatality.

It is not only inevitable, but it can be a huge and dangerous trend.

Stephen Hawking warns artificial intelligence could end mankind - BBC News

1. Start killing off mankind.
2. Have the AI edit the logs to blame it on Rio.
3. Profit.

Of course we may not hear about it all if the AI starts editing the news and we only get updates on Facebook about things the AI wants us to know.

 

[int32_t]

I think calling it circular is misleading. I expect the logs are consistent.

Are you being funny or just not getting it? I'm having a hard time figuring it out since this is just text, after all. It seems there's more than logs going in circles here. (conversations too)

Assume for a moment that this was a software or hardware fault and nobody pressed the accelerator. Unlikely, but humour me for a moment.

• Somehow (mangled bits, huge noise spike, er, DC level on throttle lines, mishandled exception, you name it) the inverter receives a command to pick.it.up -- all the way.
  
• Software dutifully notes the command sent (100% throttle applied.)
  
• Other software carries out the command.
• Still other software notes that command was carried out. (Yes, we're speeding up.)

So, in this hypothetical scenario, of course the logs would agree! Feel free to point out something that could, but I can't think of any type of logging that could actually prove that this was user error short of a dash cam aimed at the pedals. The log is getting its info from the same place the inverter is getting its signal ... the accelerator pedal and accompanying firmware.

Replace the first bullet with a user stomping the pedal and you get the same net result -- logs in agreement.

 

[vandacca]

Assume for a moment that this was a software or hardware fault and nobody pressed the accelerator. Unlikely, but humour me for a moment.

• Somehow (mangled bits, huge noise spike, er, DC level on throttle lines, mishandled exception, you name it) the inverter receives a command to pick.it.up -- all the way.
  
• Software dutifully notes the command sent (100% throttle applied.)
  
• Other software carries out the command.
• Still other software notes that command was carried out. (Yes, we're speeding up.)

...
Replace the first bullet with a user stomping the pedal and you get the same net result -- logs in agreement.

There are 3 ways that bullet point 1 can happen:
1.) Driver steps on gas
2.) Software instructs acceleration
3.) Some kind of 3rd party hack/spike/short
I think what Tesla is saying is that #2 definitely didn't happen. I'm sure Tesla believes that #3 is highly unlikely. Therefore, #1 is the most likely option. The only way to completely rule out #2 is to do a thorough analysis on the systems to see if there is any kind of way on injecting a acceleration command (e.g. via a short, surge or a hack). I don't know if Tesla has performed this type of analysis or even if it deems it necessary.