Auto Park Accident

[Landsurfer]

Last week, my wife had an accident while engaging auto park to back into a parking spot in a parking structure. We had been using auto park since its inception in 2015. It had been working well for us. We loved the convenience it gave us. But our luck ended last Thursday. It started well. It was in a company park structure. My wife found a spot, and passed it. Tesla detected to spot and showed the "P" sign. My wife engaged reverse gear, and press start to engage auto park. Tesla started backing toward the spot, but suddenly accelerated quickly, and before my wife could react on it, it smashed right into a parked car on the left rear. Tesla had a big dent at the bumper, with rear back light broken. The other car had even a bigger dent. From the damages, the force was pretty big. To me, it looked like there was a glitch in the sensor or software that had caused this. This need to be taken seriously.

I called Tesla road side assistance who directed me to a service center nearby. The service manager told me that Auto park was a "Test Feature", and you as the driver were responsible for the damages. Well, I asked him whether they were interested to take a look and download some data from the Vehicle to analyze the cause of the malfunction. The service manager said some one would call me. Well 5 days later today, still no one called.

I had been a big fan of Tesla, and had tried every feature on my Model S, especially auto pilot and auto park. This little incident poured some cold water on me. Frankly, I am quite disappointed, especially the way Tesla handle the accident. I am an engineer myself, and understand software/hardware glitch happens. When an accident happens, it is important to get first hand data, so that we can improve upon. 5 days passed, no action was taken. Some useful data might already be overwritten. We, as Tesla owner, supported Elon's vision, by contributing money and time, and being the guinea pig. I can't understand why Tesla would give up a perfect chance to investigate and improve on it.

 

[davidc18]

Sorry to hear about the accident. We have never used the autopark feature because Tesla was clearly not going to take responsibility for any errors the system would make. Take a look at Tesla's statements on FSD. They are not planning on covering cars while using FSD (when/if that is ever enabled). They are a for profit company and will act accordingly. Again, sorry for your loss and I hope your car is back on the road asap.

 

[3s-a-charm]

I honestly struggle with these stories that claim sudden random acceleration. Part of me really believes you - very sincere story - but part of me doesn't believe the car would randomly accelerate like that and that user error is to blame.

The other issue is that IF the car somehow has the software glitch that randomly accelerates at times, the Tesla logs would likely show "the accelerator was pressed to the floor" or some similar claim because the logs read what electronically occurred, not HOW the accelerator was pressed (mechanically vs electronically).

I truly wish you the best with this and have sympathy regarding your situation. I also love/trust Autopark and did have one incident where my car forced the wheel into the curb and kept pushing despite hitting the curb (while I had a client in the car no less) so I understand your plight first-hand. My local Tesla shop was nice enough to fix my rim ($150) but you are going to have a challenge somehow proving that the car accelerated into another vehicle on its own.

Keep us posted... subscribed.

 

[hacer]

...
The other issue is that IF the car somehow has the software glitch that randomly accelerates at times, the Tesla logs would likely show "the accelerator was pressed to the floor" or some similar claim because the logs read what electronically occurred, not HOW the accelerator was pressed (mechanically vs electronically).
....

The accelerator pedal, unlike the brake, is not motor driven. The only way the accelerator can be pressed down is by pressing on it. The logs record BOTH of the potentiometers that measure it's position which have opposite outputs (one goes up while the other goes down as the pedal goes down). Thus when the logs show the two sensors moving in tandem in opposite voltage change it is an extremely reliable indicator that the pedal is being pressed, regardless of what the software may be doing.

 

[xborg]

Here is what Tesla is going to say about the incident
"We've checked the logs, Autopilot has worked as intended"

Unfortunately, Tesla is not responsive about autopilot/autopark incidents.

 

[Maximapolak]

Either she pressed the throttle and she's blaming the car or, there really was a "glitch".

Either way, I wouldn't trust autopark without my foot hovering over the brake pedal.

Hope all works out well for you.

 

[llavalle]

Had a similar occurrence...
Auto Park is definitely not reliable.

 

[Landsurfer]

I honestly struggle with these stories that claim sudden random acceleration. Part of me really believes you - very sincere story - but part of me doesn't believe the car would randomly accelerate like that and that user error is to blame.

The other issue is that IF the car somehow has the software glitch that randomly accelerates at times, the Tesla logs would likely show "the accelerator was pressed to the floor" or some similar claim because the logs read what electronically occurred, not HOW the accelerator was pressed (mechanically vs electronically).

I truly wish you the best with this and have sympathy regarding your situation. I also love/trust Autopark and did have one incident where my car forced the wheel into the curb and kept pushing despite hitting the curb (while I had a client in the car no less) so I understand your plight first-hand. My local Tesla shop was nice enough to fix my rim ($150) but you are going to have a challenge somehow proving that the car accelerated into another vehicle on its own.

Keep us posted... subscribed.

Thanks for your comments. To me, I am more disappointed on Tesla unresponsiveness more than the incident itself. I wanted to see the log myself to see exactly what happened. But 5 days passed, no one from Tesla contacted me on what happened or downloading the log. I am afraid that that part of log is already overwritten.

 

[⚡️ELECTROMAN⚡️]

Thanks for your comments. To me, I am more disappointed on Tesla unresponsiveness more than the incident itself. I wanted to see the log myself to see exactly what happened. But 5 days passed, no one from Tesla contacted me on what happened or downloading the log. I am afraid that that part of log is already overwritten.

Sometimes you have to be a squeakier wheel. You shouldn't have to be, but you don't know the competency level of the person you reported the incident to. That particular person may just not care or didn't understand what happened. Who knows, but there is a lot of incompetency out there, and I'm sure Tesla has it's fair share of incompetent employees. If it was important enough, as this would be to me, I'd be calling everyday until I found someone that took this seriously.

 

[Az_Rael]

Thanks for your comments. To me, I am more disappointed on Tesla unresponsiveness more than the incident itself. I wanted to see the log myself to see exactly what happened. But 5 days passed, no one from Tesla contacted me on what happened or downloading the log. I am afraid that that part of log is already overwritten.

Follow up with them. I believe it states in the manual you have a right to your log data in the event of a non warranty repair needing to be done.

 

[3s-a-charm]

The accelerator pedal, unlike the brake, is not motor driven. The only way the accelerator can be pressed down is by pressing on it. The logs record BOTH of the potentiometers that measure it's position which have opposite outputs (one goes up while the other goes down as the pedal goes down). Thus when the logs show the two sensors moving in tandem in opposite voltage change it is an extremely reliable indicator that the pedal is being pressed, regardless of what the software may be doing.

Because both the accelerator and brake are drive-by-wire systems - if using Autopark, couldn't the vehicle "tell" the accelerator to accelerate? I'm not saying that the brakes and accelerator were pushed or even activated at the same time... but rather that there is a possibility that the software told the accelerator to "go" - or the user told the accelerator to "go" - obviously one of the two happened and it is questionable which one it was. I also don't know that the Tesla logs would demonstrate any difference if there was a glitch in the software that activated the accelerator vs a user's foot activating it... I guess that is my ultimate question here. Is there a difference in the logs if the software activates acceleration or if a user does?

 

[Petra]

Because both the accelerator and brake are drive-by-wire systems - if using Autopark, couldn't the vehicle "tell" the accelerator to accelerate? I'm not saying that the brakes and accelerator were pushed or even activated at the same time... but rather that there is a possibility that the software told the accelerator to "go" - or the user told the accelerator to "go" - obviously one of the two happened and it is questionable which one it was. I also don't know that the Tesla logs would demonstrate any difference if there was a glitch in the software that activated the accelerator vs a user's foot activating it... I guess that is my ultimate question here. Is there a difference in the logs if the software activates acceleration or if a user does?

I think what hacer is trying to say is that the accelerator pedal position sensor values are recorded independently of requested acceleration at the motor control level. A mismatch in position sensor data and requested acceleration data would indicate a possible software control issue, matched position data and acceleration would indicate physical pedal input.

 

[chillaban]

Because both the accelerator and brake are drive-by-wire systems - if using Autopark, couldn't the vehicle "tell" the accelerator to accelerate? I'm not saying that the brakes and accelerator were pushed or even activated at the same time... but rather that there is a possibility that the software told the accelerator to "go" - or the user told the accelerator to "go" - obviously one of the two happened and it is questionable which one it was. I also don't know that the Tesla logs would demonstrate any difference if there was a glitch in the software that activated the accelerator vs a user's foot activating it... I guess that is my ultimate question here. Is there a difference in the logs if the software activates acceleration or if a user does?

The brake is not drive by wire, but anyway, let's talk accelerator pedal first:

The accelerator pedal is purely drive-by-wire. The car can command motor torque for Autopilot and Autopark which overrides you, but the car never moves the pedal on your behalf. The car independently logs such commands versus the accelerator pedal position (which has dual redundancy per industry standard practice), so it is unmistakeable when the pedal has been physically moved, though Tesla cannot prove whether it's your feet or another object near the pedal that caused it.


The brake pedal (connected to Bosch iBooster) is more complicated. It's still a master cylinder hydraulic braking system, so technically NOT brake by wire, but the iBooster setup has a small servomotor that can actuate the pedal on your behalf while sensing your input too! Bosch primarily advertises this ability for hybrid vehicle regen braking (e.g. the car has a chance to intercept slight pressing of the brake pedal in order to command negative motor torque without actuating the physical brakes), but Tesla doesn't use brake pedal regen. Tesla primarily wants iBooster for AP's high-precision and high-speed servo control over the brakes (it's supposed to be able to fully slam down the brake pedal in tens of milliseconds). Long story short: The brake pedal won't move for mild brake applications, but for emergency braking, the pedal DOES move on your behalf.

 

[Tam]

...5 days passed...

Did you follow up with another call? Did the Service Center submitted your request for car log or did they forgot?

Self parking and Autopilot do fail sometimes when certain criteria are met so I don't think Tesla would routinely look up the log each time.

For example, hitting a stationary vehicle is a well known bug to be solved so Tesla won't routinely look up the log for each incidence.

However, in your case, there's something out of ordinary: sudden acceleration which should be compared to the car log to see whether it was the human foot or the system's bug.

 

[3s-a-charm]

The brake is not drive by wire, but anyway, let's talk accelerator pedal first:

The accelerator pedal is purely drive-by-wire. The car can command motor torque for Autopilot and Autopark which overrides you, but the car never moves the pedal on your behalf. The car independently logs such commands versus the accelerator pedal position (which has dual redundancy per industry standard practice), so it is unmistakeable when the pedal has been physically moved, though Tesla cannot prove whether it's your feet or another object near the pedal that caused it.


The brake pedal (connected to Bosch iBooster) is more complicated. It's still a master cylinder hydraulic braking system, so technically NOT brake by wire, but the iBooster setup has a small servomotor that can actuate the pedal on your behalf while sensing your input too! Bosch primarily advertises this ability for hybrid vehicle regen braking (e.g. the car has a chance to intercept slight pressing of the brake pedal in order to command negative motor torque without actuating the physical brakes), but Tesla doesn't use brake pedal regen. Tesla primarily wants iBooster for AP's high-precision and high-speed servo control over the brakes (it's supposed to be able to fully slam down the brake pedal in tens of milliseconds). Long story short: The brake pedal won't move for mild brake applications, but for emergency braking, the pedal DOES move on your behalf.

Very helpful, thank you. I thought for sure my brake pedal moves even with relatively slight braking when in stop-and-go traffic using TACC or Autopilot...?

 

[chillaban]

Very helpful, thank you. I thought for sure my brake pedal moves even with relatively slight braking when in stop-and-go traffic using TACC or Autopilot...?

Are you sure? I usually drive long distances with my shoes off and I've felt behind the pedal before, and my pedal doesn't seem to move at all unless the car is braking quite heavily. there is a "dead zone" if you brake while the car is braking, so there does seem to be some spring action that holds the pedal up. Maybe there's some car-to-car variance on that?

 

[Landsurfer]

Did you follow up with another call? Did the Service Center submitted your request for car log or did they forgot?

Self parking and Autopilot do fail sometimes when certain criteria are met so I don't think Tesla would routinely look up the log each time.

For example, hitting a stationary vehicle is a well known bug to be solved so Tesla won't routinely look up the log for each incidence.

However, in your case, there's something out of ordinary: sudden acceleration which should be compared to the car log to see whether it was the human foot or the system's bug.

I called Tesla again today. I was told that there was not a note that some one should contact me about the logs, while the service rep stated specifically that some engineer from Tesla should contacted me last Thursday. Today they assigned the case back to the service rep and said he would follow up with me. "Electroman" earlier commented on competency level. It may be true...

 

[sorka]

This is an almost weekly occurrence in this forum alone.

Here's the reason I stopped using auto park in one of the auto park crash threads:

Self /Auto Park sucks!!!

A few more threads but there more beyond this of course.

Auto park FAIL
Auto park malfunction. Don't feel I can trust it anymore

 

[ram1901]

The majority of Auto Park fails involve auto park NOT seeing an available spot and/or thinking there is a spot when you happen to slow down near an area that may look like a car park to the vehicle... They do NOT involved 'sudden acceleration'.

Sudden acceleration is always - always the result of someone or something hitting the accelerator by mistake.
In the case of the infamous Toyota sudden acceleration incidents, (for example) it was ones foot pushing on the detached floor mats forcing them against the accelerator, which is why floor mats are now attached to the floor with snap fittings.
Every investigation into sudden accelerations results in similar findings..i.e. driver error or driver unaware they were hitting the
accelerator, rather than the brake.

When in AutoPark mode, Tesla does NOT floor the accelerator. In fact, if anything, it moves very very very slowly into each position.
Same as in Summon's mode. In all such circumstances, drivers are required to always be prepared to take control of the vehicle (and because the vehicle is moving so slowly, if one loses confidence it the car's ability to park, they can easily hit the brake (not the accelerator) to stop the process, and park it themselves.

Drivers need to take responsibility for their actions. When these kinds of incidents occur, invariably the driver blames the car.
The car is a dumb machine that does what the driver tells it to do. If one hits the accelerator/GO pedal, it goes.

Summary: logs are going to show that the driver hit the accelerator (when they wanted to hit the brake) and thus the car accelerated as only a Tesla can do .. and hit whatever was in it's path.

 

[Mihai]

All these autopark accidents can be easily solved by changing how autopark works.


In my pervious BMW, in order to autopark you will have to keep brake pressed, switch to reverse then slowly release brake and see what the car is doing.

Your foot had to always be on the brake and just release it bit by bit (dose it off) and see what the car is doing (the steering was fully automated). If you feel the car goes on a collision path, just press the brake harder and it stops for the moment, you can check your surroundings/mirrors/etc and if all is ok you can slowly release the brake again and the car will keep continuing on the same path.

I am not using the autopark in my Tesla just because I don’t trust it not having my foot on the brake, as soon as I touch the brake it cancels the parking maneuver (very counter intuitive).

By always having to keep your foot slightly on the brake during autopark you get the following:

- No more hitting the accelerator by mistake (your foot is always on the brake)
- Having the chance to slow down or stop without canceling the maneuver (if you think something is wrong or someone walks around your car for example)
- Feeling more in control but not having to figure out the right parking path yourself

My wife has been using the autopark in the BMW almost daily for the last 4 years without inccidents, so hats off to BMW for properly implementing the autopark function in their cars