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

[bhzmark]

Interesting but the 2019 Audi A8 with this tech isnt available in Germany until this fall and won't be available in the US until spring or summer of 2019 and the driving technology may not be turned on right away. And when it is, it will be traffic jam pilot, limited as you note to 37 mph.
Autobahn traffic must be really bad now.

The Audi A8 does.
The Audi A8: the World's First Production Car to Achieve Level 3 Autonomy


Now it will only operate up to a fairly slow 37 mph and on divided highways. This is it's defined ODD (Operational Design Domain).

As a level 3 car it MUST also have a "behavioural competency" to avoid static objects, whilst operating within it's ODD. This is the key difference between the Tesla and the Audi in it's classification of autonomy level.

Now I doubt many users have read the full specifications of SAE autonomy, or the federal guidelines, and while I have seen various comments by Tesla fans on articles of the Audi mocking it's slow maximum speed of operation, I really don't think all Tesla drivers understand this key difference. They are falsely assuming the Tesla has the competency (as have a number of commentators on this thread who assumed AEB would kick in), and as such is a level 3, not a level 2.

My prediction is automotive LIDAR costs will fall and resolution increase, alongside better processing allowing faster speeds defined in the ODD of such systems,. Subsequent to that a trickle down of the tech down from flagship models.

The checklist is summarised in this document, which is an interesting read: https://www.hsdl.org/?view&did=795644

 

[mongo]

Article I section 21651 of the California vehicle code makes it illegal.

That is interesting for the concept of a gore point.

(a) Whenever a highway has been divided into two or more roadways by means of intermittent barriers or by means of a dividing section of not less than two feet in width, either unpaved or delineated by curbs, double-parallel lines, or other markings on the roadway, it is unlawful to do either of the following:

Are you putting single white lines in the category of other markings? If so, I would really want to see where white lines are defined as do not cross. Unless the logic is that any marking that is not explicitly called out as crossible is uncrossible. In which case, as long as the split is not yet more than 2 feet wide, it is still (technically) legal.

 

[mongo]

I could see the barrier from more than 150 meters away, but as pointed out in the picture, the point where somebody could realize that the car is veering off and is going straight to the barrier(on AP) is not 70-80 meters away.

As corroborated in your picture, the lane split begins before the 150 meter mark. If the lanes are properly painted, you can know at that point that the car is going the wrong way (even if it is a driver's first trip down this road, which it wasn't)

Following distance 1: They have put this out on public domain probably as fact but irrelevant. Other than making the driver look reckless and irresponsible.

Not irrelevant, sight distance is dependent on distance to a leading car. The camera is centered and more blocked than the driver's view, but both are occluded. (We also do not know if the leading car was in the gore point for a time.)

The driver had received several visual and one audible hands-on warning earlier in the drive and the driver’s hands were not detected on the wheel for six seconds prior to the collision:

Which let's the reader know that the system was able to record such things.

The second part is where It raises doubts. If the MX was just following the left lane marking on the way to the barrier, the right side of the car would see more damage.

Not with the new lane centering algorithm. That would put it right about where it hit. (Based on post-crash imagery). There is a recreation video upthread where it shows lanes on both sides as it heads to the barrier.

There could be a delay in saving the data before the system would have shut down after the crash. Probably that's why it did not record the hands on the wheel.

If so, they could not say there was no action taken in the last 5 seconds. They knew the distance and speed so they know the time till impact. They can compare that time to the last data timestamps. In general, unless they are running a large buffer without fflush, writes will happen semi instantaneously.

 

[JRP3]

After Tesla's statement in the public domain, it looks like they are worried about selling more cars and the stock price rather than making the roads safer.

Rather ignorant statement since they won't sell cars and the stock price would tank if their cars aren't considered to be safe. Plus they all drive the cars themselves, as well as friends and families.

 

[T34ME]

Good or bad, clear or foggy road conditions, I have always set my following distance the furthest, 7 since I first bought it a year ago, so if the system fails to brake or keep in lane, I can have time to react.

Exactly, that's what I do too. The deceased driver was traveling at 20/25 mph over the posted speed limit, 45 mph, and had his following distance set at "minimum." We can only speculate that if he had his following distance set at "maximum" (the car would brake sooner) and been traveling at the posted speed limit (the driver and car would have more reaction time), would he still be alive today.

My current 6 y.o. car has a primitive TACC with no "autosteer." If I am in the left lane and the road curves to the left, the car will automatically brake if there is a vehicle slightly ahead on my right. My car thinks that the person on my right is intruding into my lane and brakes accordingly. It is annoying at times, but I have come to appreciate that this idiosyncrasy of my car keeps me alert and vigilant. I trust that my car is looking out for my safety. I turn off my TACC if I am traveling on a curving road.

 

[alcibiades]

I did a little experiment today as I drive the same route every day. In manual mode, I drove my car and veered off the same way as its posted in one of the "recreated scenario" videos. I still haven't got my MX and now its part of the preparation to own it. I just wanted to see how much time I have to avoid that.

I realized how shrewd Tesla's overall statement is, especially the one about 5 seconds to act and 150 meters of unobstructed view.
View attachment 291685

I could see the barrier from more than 150 meters away, but as pointed out in the picture, the point where somebody could realize that the car is veering off and is going straight to the barrier(on AP) is not 70-80 meters away. That would mean at 70 mph, he had two seconds to react.
You got to be trained to avoid that.

In hindsight, this looks simple. If you drive the car expecting it to make mistakes and then you may be prepared to avoid this but most people drive the Tesla thinking the car has their back if they make a mistake.

Other statements from Tesla
Following distance 1: They have put this out on public domain probably as fact but irrelevant. Other than making the driver look reckless and irresponsible.

The driver had received several visual and one audible hands-on warning earlier in the drive and the driver’s hands were not detected on the wheel for six seconds prior to the collision:

The first part of the statement is to reinforce the perception that the driver was negligent.
The second part is where It raises doubts. If the MX was just following the left lane marking on the way to the barrier, the right side of the car would see more damage. From the pictures of the damage, it looks like the left side was hit hard. Did he try to steer back into the right lane to 101s?

View attachment 291687
There could be a delay in saving the data before the system would have shut down after the crash. Probably that's why it did not record the hands on the wheel.

After Tesla's statement in the public domain, it looks like they are worried about selling more cars and the stock price rather than making the roads safer.

Luckily the superhuman drivers at TMC never have a kid shout something in the back of a car causing them to turn their head nor do they ever look at a passing car while driving nor do they look down at their MCU to change stations or HVAC settings or check the map.

Thanks for your report on that part of the roadway.

 

[Az_Rael]

The deceased driver was traveling at 20/25 mph over the posted speed limit, 45 mp

I thought the speed limit in the X accident area was 65. The 45 limit was where the bus crashed, which is a different intersection. Someone did the math based on the 5 seconds and 150 meters and calculated the X driver was probably going 67 in a 65.

 

[T34ME]

I thought the speed limit in the X accident area was 65. The 45 limit was where the bus crashed, which is a different intersection. Someone did the math based on the 5 seconds and 150 meters and calculated the X driver was probably going 67 in a 65.

Yeah, it is hard to keep track, I'm having trouble keeping up. The two scenarios seem somewhat similar. Why would CalTrans have one posted at 45mph and the other at 65mph?

 

[Az_Rael]

Yeah, it is hard to keep track, I'm having trouble keeping up. The two scenarios seem somewhat similar. Why would CalTrans have one posted at 45mph and the other at 65mph?

I don’t know the areas well enough. Maybe the 45mph area has a exit ramp that can’t support freeway speeds?

 

[mkspeedr]

I did a little experiment today as I drive the same route every day. In manual mode, I drove my car and veered off the same way as its posted in one of the "recreated scenario" videos. I still haven't got my MX and now its part of the preparation to own it. I just wanted to see how much time I have to avoid that.

I realized how shrewd Tesla's overall statement is, especially the one about 5 seconds to act and 150 meters of unobstructed view.
View attachment 291685

I could see the barrier from more than 150 meters away, but as pointed out in the picture, the point where somebody could realize that the car is veering off and is going straight to the barrier(on AP) is not 70-80 meters away. That would mean at 70 mph, he had two seconds to react.
You got to be trained to avoid that.

In hindsight, this looks simple. If you drive the car expecting it to make mistakes and then you may be prepared to avoid this but most people drive the Tesla thinking the car has their back if they make a mistake.

Other statements from Tesla
Following distance 1: They have put this out on public domain probably as fact but irrelevant. Other than making the driver look reckless and irresponsible.

The driver had received several visual and one audible hands-on warning earlier in the drive and the driver’s hands were not detected on the wheel for six seconds prior to the collision:

The first part of the statement is to reinforce the perception that the driver was negligent.
The second part is where It raises doubts. If the MX was just following the left lane marking on the way to the barrier, the right side of the car would see more damage. From the pictures of the damage, it looks like the left side was hit hard. Did he try to steer back into the right lane to 101s?

View attachment 291687
There could be a delay in saving the data before the system would have shut down after the crash. Probably that's why it did not record the hands on the wheel.

After Tesla's statement in the public domain, it looks like they are worried about selling more cars and the stock price rather than making the roads safer.

You information is good but I don't like the last statement. They need to worry about selling cars (stock price not so much). If they don't sell cars they will be out of business and in the super messed up state of California you need to be extra careful.

 

[22522]

I don’t know the areas well enough. Maybe the 45mph area has a exit ramp that can’t support freeway speeds?

One of the 45MPH signs was from a re-creation done in Indiana (completely different location).

 

[daktari]

What is the stopping time from 65 mph? 3-4 seconds from danger identified?

 

[JRP3]

News article referencing the test drives 2 Tesla drivers tried to re-create the fatal Model X accident and showed a potential limitation with Autopilot

 

[EVie'sDad]

Not sure where you got this information. Below is the only thing Tesla said.

1. "earlier in the drive". How much earlier? You get these warnings every minute or two of not applying Torque.
2. We all no that what they mean my "not detected". I.E. No Torque which means nothing.
3. I am a big fan of Tesla but I am not sure who at Tesla wrote that statement. Seems misleading to the general public.

The driver had received several visual and one audible hands-on warning earlier in the drive and the driver’s hands were not detected on the wheel for six seconds prior to the collision.

Fyi
"It serves as a tragic reminder that drivers need to pay attention at all time and be ready to take over when using Autopilot."

Tesla Autopilot confuses markings toward barrier in recreation of fatal Model X crash at exact same location

 

[22522]

What is the stopping time from 65 mph? 3-4 seconds from danger identified?

I think distance works better:

Driver Care - Know Your Stopping Distance ! - Government Fleet

The way I process it, the barrier is so low (with no tall warning sign) that you can't see it (identify it) until it is too late - if stopping is your strategy.

Road markings are the only triggering cue, on the time scale of stopping the vehicle.

Speculation: the length of the gore area is abnormally long - too long. The length may have been set when California had tall visible signs on the concrete barriers. The way it is now, it is almost like a "sort out your lane" warning, that results in the paint getting worn out as people cross - maybe in bumper to bumper traffic jams.

Awareness of the barrier and the marking of the gore area should be close together in time to trigger action. Right now you forget you are in a gore area before you even see the barrier. Then it is too late for braking action. You have to change lanes.

 

[kbM3]

I haven’t read through all of the replies in this thread’s 100 pages. Forgive me if this has already been discussed.

I understand that auto pilot currently is an assist program and constantly needs to be monitored.

I see 2 limitations that occurred in this instance compared to what needs to happen for full self driving .

1) in determining the driving lane it prioritized lane markings over the physical constraints (the physical split of the driving surface by the gore point). Full self driving needs to be able to make a safe path determination in extremely difficult circumstances, and this situation was not particularly difficult. It was just bad lane markings. It didn’t even need to determine if it was safe to drive off of the road surface entirely (like over a curb, or into a park...) to avoid an accident for instance.

2) Even when lane detection fails, it still needs to be able to detect a physical barrier in the vehicle path and either swerved to another lane or braked.

It just seems to me like these are not particularly difficult scenarios compared to interactions with police officers, or avoiding a wildfire or a million other super rare scenarios, and that we are not as close to FSD as hoped.

 

[mongo]

I haven’t read through all of the replies in this thread’s 100 pages. Forgive me if this has already been discussed.

I understand that auto pilot currently is an assist program and constantly needs to be monitored.

I see 2 limitations that occurred in this instance compared to what needs to happen for full self driving .

1) in determining the driving lane it prioritized lane markings over the physical constraints (the physical split of the driving surface by the gore point). Full self driving needs to be able to make a safe path determination in extremely difficult circumstances, and this situation was not particularly difficult. It was just bad lane markings. It didn’t even need to determine if it was safe to drive off of the road surface entirely (like over a curb, or into a park...) to avoid an accident for instance.

2) Even when lane detection fails, it still needs to be able to detect a physical barrier in the vehicle path and either swerved to another lane or braked.

It just seems to me like these are not particularly difficult scenarios compared to interactions with police officers, or avoiding a wildfire or a million other super rare scenarios, and that we are not as close to FSD as hoped.

I agree, these should not be difficult situations for FSD, but this system is not FSD.

1) Without valid lane markings, there is no split to be aware of, unless the system integrates high resolution positioning/ mapping.

2) For FSD, yes it needs full obstacle detection and avoidance, but it is not there, nor is it claimed to be.

EAP's performance does not mean that the FSD development is not going well. Only that is has not been integrated into the released version. Which is itself an interesting question: ultimately how will Tesla differentiate FSD from EAP? I don't see EAP having less collision avoidance than FSD. EAP will definitely require a driver in the car, but what else? Steering wheel nags? Driver attention monitor camera?

 

[TEG]

Other statements from Tesla
Following distance 1: They have put this out on public domain probably as fact but irrelevant. Other than making the driver look reckless and irresponsible.

You could think it is irrelevant and intended to point fingers at the driver.
*or*, you could think they mentioned it because it is relevant to safety and they are suggesting that you don't use following distance 1 in that area.
For instance, if it turns out that autopilot could follow someone into the gore area if the following distance was too close for the cameras to see enough lane marking lines.

Tesla said that they didn't wait to make a statement about the data because they felt it was safety related.

For me, I now wouldn't use following distance 1 in that area until we hear more about what specifically happened that day.

The second part is where It raises doubts. If the MX was just following the left lane marking on the way to the barrier, the right side of the car would see more damage.
From the pictures of the damage, it looks like the left side was hit hard. Did he try to steer back into the right lane to 101s?

Keep in mind that the wall is offset to the left of the gore area. If you drove straight down the center of the gore area, the left side of the car would hit the wall, and the right side could have less damage.

By the way, I think those weird looking lines/shapes to the lower right of the gore may be some kind of sensors in the road. They seem to connect to a storm drain, maybe with some wiring travelling along the storm drain pipe channel.
I am guessing that they use those to measure traffic flow.

 

[ChicagolandMX]

What those barrels contain?

The makings of a bad day!

 

[TEG]

Speculation: the length of the gore area is abnormally long - too long...

I think one of the reasons why the gore area is so long there is that it was originally designed to have a much longer crash barrier in front of the cement wall.
This was the original barrier there: