Tesla can tell when the accelerator was physically pushed. End of story.Let me describe in a bit more detail how my near rear collision occurred while auto parking, and Tesla's analysis of my data log.
I back into my parking spaces every morning in the parking structure at the hospital where I work. While backing up I use my foot on the brake, releasing it and pressing it as needed to allow my 2017 Model S to "creep" slowly back wards until I am close to the wall behind me. I use the rear camera, and direct visual observation to determine my position. If I am backing up between two cars, the automatic Parking notice may come up, which is what it did last week. I normally ignore it but just out of curiosity tried it when my incident occurred. As noted when I reported this incident, the first time I tried this it worked, but the next day when I tried it a second time. In order to initiate auto park , I must keep my foot on the brake and either click the start button on the screen, or "not this time" button on the screen. Last week when I tried this, I clicked start on the screen, lifted my foot off the brake, and as soon as I noticed the alarming speed at which the back up was occurring, I quickly pressed down on the brake again, fortunately only gently hitting the back wall with no damage. Typical reaction time for something like this is between 300-500 msecs, which is probably about the total time for this entire event. I allowed the auto park to begin when I was about 3-4 feet from the back wall of the garage. As quick as my reaction time is, it would have been physically impossible for me to lift my foot off the brake, push the accelerator and then return my foot to the brake without a full force crash into the back wall within the time between hitting start and stopping the car. I reported this immediately to Tesla who downloaded the log.
After a week, my local service center, ( which I like a lot by the way) , wrote to me saying that the log indicated the the driver "pressed the accelerator." You be the judge: I believe the technician was accurately reporting what the log said, but I suggest that the reading from the log did not distinguish between the acceleration generated by its own algorithm, and me as the driver pressing the accelerator. For those of you on this forum who have been skeptical of these reports of unwarranted acceleration I invite you to try this yourself, and even if the back up speed seems to be working normally see whether you would have the time to release start the auto park, release the brake, depress the accelerator and hit the brake again without hitting a barrier 3-4 feet behind you and damaging your bumper. I am guessing that I will have few takers on this challenge as it would be a foolish and dangerous experiment to try for most of us (if you do, and you damage your rear bumper, don't blame me). I wrote back to the technician thanking him for his response, and outlining why I was skeptical of this reading.
I am sure, even with these details laid out those of you who are inveterate skeptics will continue to doubt this, but I suggest that anyone using auto park be extremely vigilant and cautious during the procedure. I for one will not be using auto park again. I park just fine on my own.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Sudden Unexpected Acceleration today
- Thread starter AZM3
- Start date
Tesla can tell when the accelerator was physically pushed. End of story.
This is very interesting. In a nut shell:
- Auto park moved the car backwards (rather quickly ignoring the barrier)
- But logs are saying that the driver pressed the accelerator and not autopark
I am willing to believe that s-cargo is right and the logs are wrong. This is withing the realm of a bug that could happen.
- Auto park moved the car backwards (rather quickly ignoring the barrier)
- But logs are saying that the driver pressed the accelerator and not autopark
I am willing to believe that s-cargo is right and the logs are wrong. This is withing the realm of a bug that could happen.
arcus
Active Member
To add to that, pressing a brake pedal pauses the Autopark till you hit Resume button on the screen. Page 85 of the manual: https://www.tesla.com/sites/default/files/model_s_owners_manual_north_america_en_us.pdf
I have the same problem with the terms "onshore wind and offshore wind" I guess which way the air blows is always kind of complicated.
kingjamez
Member
No it's not. There are multiple sensors on the accelerator that are completely independent of autopark. If those sensors don't agree then the accelerator is not "pressed" according to the computer.
-Jim
Everything seems fast to s-cargo, he normally moves at a snails pace.
Correct. The logs also note if and when Autopark is engaged, disengaged, what Autopark was doing when it was engaged, etc...
Speaking of S-Car. Go... There is a story about a snail that was mugged by a gang of turtles, when the police asked him to describe the perpetrators, the snail said he couldn't, it all happened too fast. I understand and embrace the skepticism expressed by my fellow Tesla drivers. I work with complex IT every day and we all know that systems only work as well as the code is written. With millions of lines of code its easy to make a miss a specific set of circumstances, particularly when it shows itself only intermittently. Since 2018.10.4 i have had problems with homelink and easy entry that occur in only one specific situation related to when I initiate the easy entry system before I back out of my garage. I reported this as well, and it has only been corrected in this recent update. I understand that most reports of sudden acceleration are reflections of driver error, but do read my detailed account of what happened and judge for yourself. Could I have hallucinated the majority of events that I described? Maybe. When I have time to stop into my service center I plan to review the log and the details of the timing of the events recorded and will report back here. The reality of human factors and the timings involved narrow down the possibilities in this situation, so we shall see. Maybe this will never occur again, who knows. As I have said previously, the main take away from my report is to use auto park with caution.
As a result of the previous discussion, I have come up with a possible explanation for sudden unintended acceleration in Tesla vehicles that also explains how the accelerator pedal sensors can have large outputs without the driver pressing on the accelerator pedal. You can read this explanation by going to Dr. Ronald A. Belt’s Sudden Acceleration Papers - Center for Autosafety and reading the paper entitled “Tesla’s Sudden Acceleration Log Data – What it Shows – 5/1/18”. This explanation provides a testable theory of sudden unintended acceleration in all Tesla Vehicles.
I skimmed it and see no reference to the fact that the two position sensors in the pedal have opposite slopes. The dual slope nature makes the correlated output immune to absolute voltage shifts, as does the system's monitoring of the voltage to the sensors.
Did I miss that detail?
That detail was not discussed because it is not necessary. The outputs of the two accelerator pedal sensors are ratiometric, which means that the outputs are proportional to the voltage drop across the sensors. if the +5V supply drops, or if the ground voltage is raised, the sensor outputs merely adjust their values to be proportional to the voltage drop at the time. This ratiometric property applies to both potentiometer type sensors and Hall effect type sensors.
Dual slope outputs will shift absolutely based on total potential, but will not shift position relative to each other. The ADC also reads the FS voltage, so any shift will be compensated.
Resting position:
Sensor A 10% full scale
Sensor B 90% full scale
Floored position
Sensor A: 90% full scale
Sensor B: 10% full scale
Rationality checks:
- Does A = full scale - B?
- Are either < 10%
- Are either > 90%
Calculated position = (A+(FS-B)) / (2 * FS) = (xA + (xFS-xB)/(2*xFS) for any change (x) in voltage.
[Mongo said][/"Changing full scale range does nothing to the calculated output.
Calculated position = (A+(FS-B)) / (2 * FS) = (xA + (xFS-xB)/(2*xFS) for any change (x) in voltage."]
Below is a calculation of the sensor position for three cases. The first two cases have different scales, with the two sensor outputs in each case lying within the range of the ADC. The calculations show that the ratiometric property holds, yielding the correct sensor output in both cases, as you predicted.
The third case shows what happens when a ground drop is present in both sensor outputs, causing a DC offset of the two signals with respect to the ADC. The correct sensor output is again obtained when both sensor outputs lie within the range of the ADC. But when either one of the sensor outputs lies outside the range of the ADC, two possibilities exist. In one case (the case of low accelerator pedal displacements), the combined sensor output becomes negative for some values of the sensor output and the ground drop. In the other case (the case of high pedal displacements), the combined sensor output remains positive, but still has a dependence on the ground drop.
I believe that this dependence upon the ground drop, which causes the combined sensor output to be negative for cases of low pedal displacement, may make the vehicle susceptible to sudden unintended acceleration.
Calculated position = (A+(FS-B)) / (2 * FS) = (xA + (xFS-xB)/(2*xFS) for any change (x) in voltage."]
Below is a calculation of the sensor position for three cases. The first two cases have different scales, with the two sensor outputs in each case lying within the range of the ADC. The calculations show that the ratiometric property holds, yielding the correct sensor output in both cases, as you predicted.
The third case shows what happens when a ground drop is present in both sensor outputs, causing a DC offset of the two signals with respect to the ADC. The correct sensor output is again obtained when both sensor outputs lie within the range of the ADC. But when either one of the sensor outputs lies outside the range of the ADC, two possibilities exist. In one case (the case of low accelerator pedal displacements), the combined sensor output becomes negative for some values of the sensor output and the ground drop. In the other case (the case of high pedal displacements), the combined sensor output remains positive, but still has a dependence on the ground drop.
I believe that this dependence upon the ground drop, which causes the combined sensor output to be negative for cases of low pedal displacement, may make the vehicle susceptible to sudden unintended acceleration.
For significant ground offset, the range check would throw a fault. ADCs can't output a negative value, and properly coded routines will not underflow to a positive number.
Example: A:1V, B:4V offset of +-1V will trip a bounds fault. Say .5V offset between sensor and ADC due to wiring fault
A: 0.5V B 3.5V total of 4V will throw a fault.
Other way
A: 1.5V B: 4.5 total of 6V will throw a fault
The code with validate the inputs before combining them.
If the system managed to get an equal offsets on top and bottom sides of the sensors, then the return would compress toward midscale. (With limit at 2.5V in/out). However, that would require multiple simultaneous faults.
That's all pretty numbers and all - but still doesn't change the fact (or ignores it?) that brakes always stop the car regardless of how hard the accelerator pedal is pressed.
In all Tesla's the brake system is independent and not regenerative so will always work.
Therefore - if you are pressing the brake, the car will stop. If you can't stop the car, you are NOT pressing the brake pedal - simple.
In all Tesla's the brake system is independent and not regenerative so will always work.
Therefore - if you are pressing the brake, the car will stop. If you can't stop the car, you are NOT pressing the brake pedal - simple.
Mongo said" ADCs can't output a negative value" ][/QUOTE]
None of the ADC measurements in my figure above were negative voltage values that would cause a negative value of the ADC (which, I agree, is impossible). The negative values resulted from arithmetic calculations using the always-positive ADC voltage values.
You are assuming that some bounds test is performed that will trip a bounds fault if the sensor output with zero pedal displacement is offset from the ADC ground. While such bounds testing is clearly possible, it is not always done to the same degree by all auto manufacturers. How do you know that Tesla is performing such bounds checks? For that matter, how do you know that Tesla is using one accelerator pedal position sensor with ascending values and the other with descending values? I have never seen either of these features documented by Tesla, so are these merely assumptions on your part?
None of the ADC measurements in my figure above were negative voltage values that would cause a negative value of the ADC (which, I agree, is impossible). The negative values resulted from arithmetic calculations using the always-positive ADC voltage values.
You are assuming that some bounds test is performed that will trip a bounds fault if the sensor output with zero pedal displacement is offset from the ADC ground. While such bounds testing is clearly possible, it is not always done to the same degree by all auto manufacturers. How do you know that Tesla is performing such bounds checks? For that matter, how do you know that Tesla is using one accelerator pedal position sensor with ascending values and the other with descending values? I have never seen either of these features documented by Tesla, so are these merely assumptions on your part?
I'm a former Teir one electronics supplier. Every input is bounds checked and sets a DTC on out of range (short to batt, short to ground) conditions followed by rationality checks (A vs B). Opposite slope pedal sensors are the industry standard (otherwise you can't detect shorts between sensors and supply/ground issues). @wk057 mention it upthread.
Odd that you would post a bunch of papers with your theories on the internet (Dr. Ronald A. Belt’s Sudden Acceleration Papers - Center for Autosafety) but not submit a single paper for peer review. And yes I've read your explanation for that but if you truly wanted your theories to be tested by the scientific community there is no substitute for publishing in a peer reviewed journal where they would be subject to scrutiny by other scientists. That's assuming your analysis passed peer review and met standards for publication, of course.
Last edited: