AP/FSD related crashes

[AlanSubie4Life]

Yes, it drives itself, making all stops and turns. Only have 2 issues on the normal commute to work. First, is the merge from a State Route (4 lane Hwy) onto the Interstate (6+ lane Hwy). It does not recognize that only the rightmost lane NOW exits to the interstate. Use to be 2 lanes going to interstate, but the traffic pattern was recently changed in the area. Second spot it normally has issue is on the interstate where 4 southbound lanes merge into 3 lanes to go under an underpass. I force into the 3rd lane, but if a car is in front (slower) than it tries to change lanes back into the lane that is ending to pass the slower car.

Thanks for confirming you were using FSD(S). It's good for users to realize that their experience does not represent the entire set of conditions that FSD sees. FSD certainly can fail in interesting ways! Seems like a confirmed accident (obviously subject to attestation in a court of law); very interesting.

For the next attempt (presumably in your new car!), definitely use a cabin camera mount (GoPro or similar, capturing the on-screen display). I recommend braking earlier and not swerving, too. People may say that FSD would have stopped if you had allowed it to, but discretion is the better part of valor.

 

[asterothe]

If I suspect something funny going on I take control. You cannot wait and see with FSD. The stakes are too high to test.

 

[cdotyii]

Does the car automatically stop and make turns for you at stop lights and stop signs or not? If it does, then FSD is enabled, and if it doesn’t, then you’re only using Autosteer. Even though all Teslas come with FSD “capability”, and even if you purchased or got a trial of FSD, that doesn’t mean it was enabled. You have to enable it in settings.

Tesla Data Report Confirms FSD was active at time of incident

 

[AlanSubie4Life]

Tesla Data Report Confirms FSD was active at time of incident

Can you post that here? I don’t really have any doubts - it was a tough situation for FSD, but people tend to be skeptical.

 

[CoolBuddy]

Both incidents happened at different railroad crossings, where I was the first car at the tracks. I have a 55-mile commute one-way each morning, and 98% of the time, the FSD does exactly what it’s supposed to do. Since an update about a month ago, the car has been setting the speed on the rural state route (with a 55 MPH speed limit) to around 61 to 63 MPH, which is consistent with normal traffic flow and what I would usually set cruise control at in other cars.

This is not an attempt at "insurance fraud." It is very concerning to me that the FSD system has failed to recognize a train crossing twice.

I find it so interesting how FSD behaves differently even in the same state. I drive between Norwalk and Tiffin and with Auto speed offset enabled, it drives between 53 and 55 then ramps up to 67 mph at times, which will definitely get you a speeding ticket from OSP. Regarding railroad tracks, I observed if I'm going 40 or under, it will slow down for them. If going 50+, it will not. There's a crossing on OH-162 just West of the OH-4 intersection where it never slows down.

I've also observed that on country roads with stop signs that aren't in map data (prior to v12 - haven't tried these since v12), it was not slowing down on approach. I mention this because it is really obvious if you get to the stop-sign ahead/rail road crossing sign, it isn't slowing down and you need to take over. Basically, for future reference, if you see it's going to be a hard stop, it is better for your car in general to just disengage where you would normally start slowing down.

For the lawyer suggestion, a personal injury lawyer probably isn't going to help, but have you tried Tim "He'll Make Them Pay" Misny?

 

[william geil]

Tesla Data Report Confirms FSD was active at time of incident

Did you manually adjust the steering wheel of the vehicle, or did the FSD make it when your car was about to collide?

 

[im.thatoneguy]

The only newsworthy part of this is how suicidal you were in not intervening earlier.

I could produce a dozen videos like this per day if I was willing to wreck a dozen cars every day.

FSD is not ready for you to be reading your Instagram feed while it drives. You will relatively quickly end up smashed into a bus or train.

 

[OxBrew]

The only newsworthy part of this is how suicidal you were in not intervening earlier.

I could produce a dozen videos like this per day if I was willing to wreck a dozen cars every day.

FSD is not ready for you to be reading your Instagram feed while it drives. You will relatively quickly end up smashed into a bus or train.

Square that with Elon/Tesla's continued hammering that Teslas using "AP" crash x times less frequently than all cars.

Where do you draw the line? The implied message is AP does better than humans. I'm human. When do I let FSD over ride my flawed judgement?

Its a lose lose scenario. Teslas are safer on AP only due to hyper aware drivers. Thats the real message they need to get out there. Without supervision, AP would crash several times a day, as you point out.

 

[stopcrazypp]

Square that with Elon/Tesla's continued hammering that Teslas using "AP" crash x times less frequently than all cars.

Where do you draw the line? The implied message is AP does better than humans. I'm human. When do I let FSD over ride my flawed judgement?

Its a lose lose scenario. Teslas are safer on AP only due to hyper aware drivers. Thats the real message they need to get out there. Without supervision, AP would crash several times a day, as you point out.

AP and FSD stats are different. AP might crash a lot less per mile because it's mostly used on straight higher speed roads. FSD is used more on slower roads with turns and intersections.

And the stat is AP+human. It's a win-win because it also reduces driver fatigue, allowing the human to focus more on the road instead of micro-adjustments.

FSD might not be there yet, as there is still a greater level of unpredictability.

I don't agree there is an overall message that ever suggests to let AP override the human, as there is a constant nag that suggests the opposite. Perhaps someone that haven't used the system or never experienced nags yet might think that way, but as soon as the nag is experienced it seems petty blatantly obvious the system can't be left to its own devices.

 

[cdotyii]

Can you post that here? I don’t really have any doubts - it was a tough situation for FSD, but people tend to be skeptical.

 

[cdotyii]

Did you manually adjust the steering wheel of the vehicle, or did the FSD make it when your car was about to collide?

I manually took over and cranked the wheel to the right while slamming on the brakes

 

[RSpanner]

Not interested and not using the temporary FSD. Dealing with a lot of problems..since it was issued..from window issues, ,super slow response and no response to a number of things...like, Reverse, etc. Pains I do not need.

 

[willow_hiller]

Teslas are safer on AP only due to hyper aware drivers. Thats the real message they need to get out there.

The real questions to me would be:

1. Does FSD make drivers more or less attentive than they would be without it?
2. Is an inattentive driver driving manually more or less likely than an inattentive driver on FSD to cause a collision?

I know automation complacency is a well studied phenomenon, but I hope Tesla's driver monitoring would help combat it somewhat. And I have the feeling that if a driver is going to be inattentive no matter what, they're better off with some form of automation.

I have a feeling that if a driver is fine going 60+ MPH on a rural road with train crossings in dense fog, they don't need advanced technology to get in an accident. This crash could have just as easily happened on cruise control.

 

[Persfoto]

I am looking for information on incidents involving Tesla's Full Self-Driving (FSD) mode. I have owned my Tesla for less than a year, and within the last six months, it has twice attempted to drive directly into a passing train while in FSD mode. The most recent incident occurred on May 8, 2024, and I have dash cam footage from that event.

I am trying to obtain the telemetry data from these incidents. Additionally, I am looking for similar cases or incidents. Unfortunately, I haven't been able to find a lawyer willing to take my case due to the lack of significant injuries—only backaches and a deep bruise on my right elbow, which didn’t require medical attention.

I attempted to upload the dash cam footage but was unsuccessful. Any guidance or assistance would be greatly appreciated.

Dashcam Vids

Shared with Dropbox

www.dropbox.com

Including a video.
Tesla problems

 

[AlanSubie4Life]

View attachment 1049582

Craig, Do you have the speeds?

I don’t think this was 60mph as speculated above Probably 45mph?

 

[willow_hiller]

Craig, Do you have the speeds?

I don’t think this was 60mph as speculated above Probably 45mph?

He said 61-63 MPH in a previous response:

Both incidents happened at different railroad crossings, where I was the first car at the tracks. I have a 55-mile commute one-way each morning, and 98% of the time, the FSD does exactly what it’s supposed to do. Since an update about a month ago, the car has been setting the speed on the rural state route (with a 55 MPH speed limit) to around 61 to 63 MPH, which is consistent with normal traffic flow and what I would usually set cruise control at in other cars.

It is hard to imagine someone with the authority to set speed limits deciding that 55 MPH is appropriate for that road.

 

[JB47394]

And I have the feeling that if a driver is going to be inattentive no matter what, they're better off with some form of automation.

Or just full time attention monitoring.

 

[S4WRXTTCS]

I manually took over and cranked the wheel to the right while slamming on the brakes

I'm confused by why you didn't intervene earlier?

Was it intentional inattention?
Was in unintentional inattention? Essentially unconscious trust in the system.
Was it due to the wanting to know whether it would handle the situation?

Obviously you failed so the question is why?

I get that you posted the video because you were trying to show that the system failed. It obviously did, but it WAS SUPPOSED to fail. It doesn't matter how good it might seem to be at times, but its going to fail.

In fact the more it improves the MORE dangerous it will get because drivers will stop being vigilant. In fact the whole idea of an L2 everywhere system it ridiculously stupid.

All L2 accidents are on the driver unless it can be proven that the driver didn't have a chance to intervene.

 

[Speight]

He said 61-63 MPH in a previous response:



It is hard to imagine someone with the authority to set speed limits deciding that 55 MPH is appropriate for that road.

55 mph is the default speed limit for rural roads in several states. It doesn't mean that it's safe to travel at that speed in all locations along the road or in all conditions. It's common for that to be the legal limit on roads that have curves requiring a much lower speed. Such curves and other hazards might or might not have a lower advisory speed posted. Drivers are expected to be paying attention to the curvature of the roadway, weather, etc. and slow down if the posted limit is too fast for conditions. At least in California, even on state highways, it's not uncommon to have a 55 mph legal limit on segments of road where essentially no one ever goes that fast because there are too many sharp curves.

 

[AlanSubie4Life]

He said 61-63 MPH in a previous response:

Agreed. Just sounded like the general speed. Wasn’t clear to me that was the speed during this specific incident. It might be. It’s a little hard to tell. The video has been pulled now, though it is available on Twitter of course.