Model X Crash on US-101 (Mountain View, CA)

[M0DEL³]

Here are the videos from my testing this morning:

Thanks, great effort! It would be interesting to see what would happen if one reattempted this test with a lead-car that encroached into the gore zone before correcting to the right. I would if AP would follow the lead to the right or continue tracking the lane marker to the barrier?

 

[Skidmark]

I did notice on the second run you got a "hold hands" visual warning. And on both videos the car went slightly to the left on autosteer. Both times you were driving with someone detected in front of you.

I'm really trying to understand how this could have happened, and it was a shocker to me to hear that AP was active during the crash.

My assumption is that traffic was even heavier at the time of the crash (9:25am on a weekday in Mountain View), so the vehicle involved also would have been detecting a vehicle in front while on autosteer. The speed I'm also assuming would be lower because of the traffic on a weekday, but without definitive information on those items, I can't really get more accurate with my testing.

 

[robe900]

Thanks, great effort! It would be interesting to see what would happen if one reattempted this test with a lead-car that encroached into the gore zone before correcting to the right. I would if AP would follow the lead to the right right or continue tracking the lane marker to the barrier?

No! Don’t do that please, you don’t even know if that’s what happened. Don’t risk another life.

 

[Skidmark]

Thanks for sharing. Any chance you did a run in the far left lane?

I didn't, because my assumption is that the crash victim was in the same lane I tested in.

 

[robe900]

I'm really trying to understand how this could have happened, and it was a shocker to me to hear that AP was active during the crash.

My assumption is that traffic was even heavier at the time of the crash (9:25am on a weekday in Mountain View), so the vehicle involved also would have been detecting a vehicle in front while on autosteer. The speed I'm also assuming would be lower because of the traffic on a weekday, but without definitive information on those items, I can't really get more accurate with my testing.

 

[VT_EE]

Here are the videos from my testing this morning:

First Attempt:

Second Attempt:

I looks like your vehicle drifted slightly toward barrier at one point in the video. Notice how the lane markers are much better heading toward the left exit lane. The white line for the straight lane is very faded in comparison. I could see how the computer might initially have issues figuring out what is going on in this situation.

 

[Skidmark]

I looks like your vehicle drifted slightly toward barrier at one point in the video. Notice how the lane markers are much better heading toward the left exit lane. The white line for the straight lane is very faded in comparison. I could see how the computer might initially have issues figuring out what is going on in this situation.

Yes, my car drifted left near the barrier on both attempts, but never crossed the white line.

 

[dm33]

Here are the videos from my testing this morning:

First Attempt:

Second Attempt:

Inconclusive. Autopilot uses the existence of cars around it to help guide its placement. Being right behind a car in the proper lane will guide autopilot. It would have a tougher time if there was no car ahead. If you were in the left most HOV lane it would be harder still, especially if there was a car in the next lane over when the split arrives encouraging autopilot to stay next to the other car and pick the non existent lane.

 

[bmah]

Moderator note: Moved some (reported) snippy posts plus responses into Snippiness. There are probably others that should be moved too.

As a reminder, especially to new TMC members: Various points of view are welcomed here on TMC, but trolling and personal attacks are not. Consider that it's been a crazy week and clearly emotions are running high. Before clicking "Post Reply", please take a few seconds, moments, whatever to consider your writing and how it might be received.

Thank you,

Bruce.

PS. Also please remember that the topic of this thread is a specific, unfortunate incident, not autopilot in general.

 

[ggnykk]

Here are the videos from my testing this morning:

First Attempt:

Second Attempt:

definitely need a test without ANY leading vehicle in front of you.

 

[cybergates]

View attachment 290804[/QUOTE]

No. The car ahead is in the correct place in the lane. The black line you indicated as "painted line gone" was never a lane marker - it is a crack/seam between two different freeway surfaces. The freeway used to be more narrow here and they expanded it to add lanes. They left behind different textures that almost look like lane markers in the wrong places.

Here is your picture but with different labels:
View attachment 290811

This is a big problem with roads - caltrans should be held responsible for painted lines that are hard to read even for humans or plain worn off. Additionally the tesla should also bear the majority responsibility in that AP2 still has problems with it unexpectedly thrashing into another lane as soon as markers disappears - the logical course of action should be to continue on the same trajectory when lines disappear - seems they have made some improvement in recent firmware but I can see how the car could crash into a concrete barrier - seems it should be able to recognize that even if it veers off and stop before hitting concrete... we can say the driver should be alert, but even in the most alert circumstances, I've had the car unexpectedly veer as soon as lines disappear in an intersection and almost hit a car in the adjacent lane.

 

[MXWing]

For all the Sherlock’s here testing:

1.) Have we confirmed Mr. Huang was headed to Apple and which campus?

2.) Traffic conditions and weather conditions and time of day?

3.) AP software version on Mr. Huang’s car?

I see software devs are in full force but QA team is MIA.

 

[kkboss]

Yes, my car drifted left near the barrier on both attempts, but never crossed the white line.

Tesla has already said that the vehicle was in distance 1 setting that would mean if there was a car in front that either went to 85 at the very last moment or drifted left and pulled back to 101s, this vehicle since it was following very close could have got confused about the lane markings.

 

[mongo]

View attachment 290804


This is a big problem with roads - caltrans should be held responsible for painted lines that are hard to read even for humans or plain worn off. Additionally the tesla should also bear the majority responsibility in that AP2 still has problems with it unexpectedly thrashing into another lane as soon as markers disappears - the logical course of action should be to continue on the same trajectory when lines disappear - seems they have made some improvement in recent firmware but I can see how the car could crash into a concrete barrier - seems it should be able to recognize that even if it veers off and stop before hitting concrete... we can say the driver should be alert, but even in the most alert circumstances, I've had the car unexpectedly veer as soon as lines disappear in an intersection and almost hit a car in the adjacent lane.

No swerving reported

CHP Redwood City
@CHP_RedwoodCity
·
Mar 23
Update on collision on US-101southbound at SR-85
Blue Tesla driving southbound on US-10, driving at freeway speeds on the gore point dividing the SR-85 carpool flyover and the carpool lane on US-101 southbound collided with the attenuator barrier and caught fire

 

[CarbonFree]

Once, AP almost did this to me. I was in a similar leftmost HOV lane when it was turning into a divided lane. A similar concrete barrier was at the gore point.

In my case, the car that AP was tracking decided suddenly that it wanted to be on the main highway, and veered suddenly to the right, crossing just in front of the gore point. AP followed it for a while, putting me on a trajectory to hit the gore point, then gave up and turned itself off with a few beeps.

Fortunately I was able to recover. Note that this was AP 1.0, two years ago.

 

[Saghost]

Tesla has already said that the vehicle was in distance 1 setting that would mean if there was a car in front that either went to 85 at the very last moment or drifted left and pulled back to 101s, this vehicle since it was following very close could have got confused about the lane markings.

Interesting theory. Since Tesla obviously got to the car's logs they probably also got to the AP camera crash footage, so Tesla probably knows if something like this happened and we may hear more in coming days.

 

[Driver Dave]

No. The car ahead is in the correct place in the lane. The black line you indicated as "painted line gone" was never a lane marker - it is a crack/seam between two different freeway surfaces. The freeway used to be more narrow here and they expanded it to add lanes. They left behind different textures that almost look like lane markers in the wrong places.

Here is your picture but with different labels:
View attachment 290811

Ok, wow, the crucial line is horribly faded.

As the barrier was already hit previously, it is easy to see how this area of road really needs some maintenance.

 

[bonnie]

Tesla has already said that the vehicle was in distance 1 setting that would mean if there was a car in front that either went to 85 at the very last moment or drifted left and pulled back to 101s, this vehicle since it was following very close could have got confused about the lane markings.

Last month I was in the middle lane of I-84 (Portland), using autopilot, follow distance set 5 or 6. Medium density traffic with a moderate slowdown from normal speeds (probably around 50mph, rather than 70). The car in front of me suddenly swerved to another lane. The reason? There was an empty yellow plastic bucket in the road. I did have enough time to move, but unfortunately there were cars on either side of me, no place to move to & no time to brake completely. So I hit it (no damage).

If that plastic bucket had been a concrete barrier, I would have been screwed. And it would have had absolutely nothing to do with autopilot. My follow distance was adequate. I was alert and paying attention. And there was no way to see that bucket. I wish the car in front of me had reacted just a second or two sooner - it would have given me a little more time to decide what to do.

We're all speculating like mad. I'm only offering another scenario. And if this was similar to what happened to cause this tragic accident, it is really immaterial if autopilot was engaged or not. (Though in my case, my logs would have shown that I hit the brake before impact - but not if I hadn't been paying attention.)

 

[bonnie]

Once, AP almost did this to me. I was in a similar leftmost HOV lane when it was turning into a divided lane. A similar concrete barrier was at the gore point.

In my case, the car that AP was tracking decided suddenly that it wanted to be with on main highway, and veered suddenly to the right, crossing just in front of the gore point. AP followed it for a while, putting me on a trajectory to hit the gore point, then gave up and turned itself off with a few beeps.

Fortunately I was able to recover. Note that this was AP 1.0, two years ago.

Glad you were able to recover.

For those new to AP, two years ago AP consistently tried to take exits, etc. It has evolved a lot since then...

 

[mongo]

Interesting theory. Since Tesla obviously got to the car's logs they probably also got to the AP camera crash footage, so Tesla probably knows if something like this happened and we may hear more in coming days.

The lane split before the barrier is 200 meters or so. Barrier is visible from even further when not blocked by a vehicle.

We're all speculating like mad. I'm only offering another scenario. And if this was similar to what happened to cause this tragic accident, it is really immaterial if autopilot was engaged or not.

Given Tesla's blog post, the last moment occluding car scenario does not seem to fit.

The driver had about five seconds and 150 meters of unobstructed view of the concrete divider with the crushed crash attenuator, but the vehicle logs show that no action was taken.