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

[e-FTW]

Let's all remember that we don't know if he was using AP when the accident happened. And even if he was, it is extremely easy to regain control and deny any odd behaviors by AP. I don't even have to think about it and I'll have disabled AP. And this guy was a software engineer, so he understood how systems like this work.

This is for the media: This is 2018. No one in their right mind would get in to a car that you cannot take control of at will. We do not have self-driving cars. AP is only there to help. Even if AP was driving toward a concrete barrier, every single attentive driver out there can regain control easily and avoid the crash. It would be nuts for Tesla to have designed AP to be hard to disengage.

I know headlines need to be made today, but we will have to wait for the official report to draw conclusions.

 

[SMAlset]

I think it's understandable that Tesla becomes the poster company in the media for these type of accidents right now. Partially because it's new technology, partially to alert Tesla owners to possible issues like with AP, partially because of those out there such as individuals who are entrenched in ICE vehicles in one way or another for personal use or even work related income who dislike Tesla, short sellers who make money from the stock price going down, people who look down on Tesla owners for owning a more expensive car, people who love Elon and Tesla, people who hate Elon and Tesla, and probaby more reasons. However at the end of the day as more BEVs are undoubtedly coming to the roads in the future (be it kicking and screaming still really), there will most certainly be other manufacturers who will encounter issues with their cars and software driver assist systems in accidents such as this one. I'm sure if they are a major manufacturer they will find themselves in the spotlight too especially as their BEV numbers rise and it affects more owners. It's a relatively new segment of the industry at least in popularity so there's that too and people are curious/cautious about adopting it.

As a Tesla owner it is tough seeing Tesla in the news for this, and I'm sure Tesla is taking the loss of one of its customers hard, but regardless of how it turns out, ie whether Tesla's AP was involved in what happened or not, they will take steps to make the car safer. One can't design a car to prevent all accidents from happening but achieving high safety marks is something they strive for with each car.

 

[Cosmacelf]

If the driver was indeed aware that his AP (no one else has corroborated that their AP behaves the same) was drifting left at that point, then it would be even more of a driver error to rely on it.

Look, the bottom line is that you can’t blame AP. The driver should always be looking at what the car is doing. And disengaging AP is easy. There is a left exit near me that AP used to always want to take, and I was always paying attention and forcing it to stay right, usually disengaging AP by doing so.

This rush by the family to blame AP is more about the grieving process than it is about reality.

 

[McRat]

BTW it should be noted that this diagram above is from The Driver's Handbook from South Australia, not California where this accident happened and some of the driving videos crossing solid white lines occur. My Licence - The Driver's Handbook - Driving on the road

Always want to follow the rules of the road for where you are driving. California Rules of the Road make no particular distinction for a solid white line other than to say it divides roadways traveling in the same direction. See my post and link here: Model X Crash on US-101 (Mountain View, CA) and this is from a web traffic school course on road markings pertinent to crossing white lines in California:

"A single solid white line is intended to discourage, but not prohibit you from changing lanes. Although in some states crossing a single solid white line is illegal, it is legal to do so in California, unless the solid white line designates a turn." Section 3.4 Signs, Signals, and Roadway Markings

Without doing an extensive CVC check, there is a law against passing or driving on the shoulder. On freeways, the shoulder has a marker which is a solid white line in most areas. You are not allowed to drive in there except in an emergency, hence the term emergency lane. The wording is a touch ambiguous which says you must drive in the traffic lanes, and the shoulder is not a traffic lane.

Even on an empty road, it is a very dangerous place to drive. This is where the debris field collects. Nails, rocks, glass, sand, and sometimes there will be sharp grade height drop. Stopping in this area is very dangerous when there is traffic, do all you can to avoid stopping there. Many people are injured and killed on the shoulder by passing cars who drift into the shoulder.

Since most drivers will not cross a solid white, even one separating adjacent traffic lanes will form a bit of a debris field.

 

[McRat]

Nobody knows if auto-steering was active on the car at this point correct?

 

[Bubslug]

Good observation, @mongo. The original CHP tweet stating "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" seems to indicate that they initially believed that the vehicle was driving in the gore point before colliding with the barrier. Whether it was AP or driver error that caused the vehicle to be in that non-lane, we don't know for sure. But based on this info, the front-end damage severity, and the trajectory of the wreckage I strongly believe that the car was not swerving or making avoidance maneuvers before impact.

The media tends to embellish things. I would not be surprised at all if the "driver lost control" statement was added by a reporter or news editor. Isn't that what they always say when a car collides with a stationary object?

I mentioned in a previous post that Walter (the driver) was a personal friend of mine. Now that it's out in the media about his prior complaints about AP, I feel comfortable sharing that I heard the same thing from a friends. Apparently he discussed with his wife and at least one other close friend as recently as the week prior to the accident that AP was drifting left at this exact particular junction on previous commutes at around the same time of day. I learned of this allegation a day prior to the I-TEAM news story. Frightening.

It's definitely shocking to me if he did indeed experience this issue with AP, why he would continue to rely on it here. I can only assume he was distracted or otherwise not paying close enough attention. This whole thing is very tragic.

Yes, shocking that he would continue to use AP, if he had experienced failures at that spot, but he may have been taking advantage of AP to respond to an incoming communication on his phone, and momentarily forgot he was approaching the spot he normally disengaged AP for safety. Tesla could in theory know his driving habits on the Bayshore, given their ability to upload logged data from their products. I'm guessing you can "opt out" of this access however, and maybe Walter did.

 

[brkaus]

Why he would continue to use something that put his life at risk repeatedly makes no sense - in my opinion stupid.
Why would he be in the 85 HOV exit lane if he needs the south bound 101 HOV lane? (Ok, not proven but it appears to me that way).

On local highway (divided, 3 lanes in each direction), when main road is backed up, I see people take the exit and get back on the next ramp. There are no lights or stop signs, just ramp to/from cloverleaf. They get about a 1/2 mile "boost".

Not saying the driver was doing this, but could have been forced by someone gaming the system.

 

[bro1999]

If the driver was indeed aware that his AP (no one else has corroborated that their AP behaves the same) was drifting left at that point, then it would be even more of a driver error to rely on it.

Look, the bottom line is that you can’t blame AP. The driver should always be looking at what the car is doing. And disengaging AP is easy. There is a left exit near me that AP used to always want to take, and I was always paying attention and forcing it to stay right, usually disengaging AP by doing so.

This rush by the family to blame AP is more about the grieving process than it is about reality.

You can most certainly blame AP (assuming it was on in this crash). The driver should be alert at all times with eyes on the road, but when you are given the impression your car car drive by itself for stretches (Who would ever get that impression after activating a feature called "Autopilot"? ) you are likely more prone to let your attention wander compared to someone driving manually. About 5 seconds of distraction was probably all that was needed to allow this crash to happen.

Plenty of blame to pass around (including Caltrans, the driver), but to simply dismiss Tesla has any responsibility because no one "should" ever get in an accident with AP is pretty silly.

 

[SMAlset]

Without doing an extensive CVC check, there is a law against passing or driving on the shoulder. On freeways, the shoulder has a marker which is a solid white line in most areas. You are not allowed to drive in there except in an emergency, hence the term emergency lane. The wording is a touch ambiguous which says you must drive in the traffic lanes, and the shoulder is not a traffic lane.

Even on an empty road, it is a very dangerous place to drive. This is where the debris field collects. Nails, rocks, glass, sand, and sometimes there will be sharp grade height drop. Stopping in this area is very dangerous when there is traffic, do all you can to avoid stopping there. Many people are injured and killed on the shoulder by passing cars who drift into the shoulder.

Since most drivers will not cross a solid white, even one separating adjacent traffic lanes will form a bit of a debris field.

We're not talking shoulder white lines here but those of merging roadways where they are used. This is typical in our area with all the morning metered entrance ramps and other major roadways that combine. The video that prompted the white line crossing discussion (originally thought by poster showing AP did it in error) has the merging roadway single white line (see below). Believe we have 3 lanes continuing on the left and a separate merging/exiting 2-lane roadway to the right. You can legally cross this white line to reach your desired highway path, assuming you've checked traffic. Lots and lots of this example here in the bay area.

 

[mongo]

On local highway (divided, 3 lanes in each direction), when main road is backed up, I see people take the exit and get back on the next ramp. There are no lights or stop signs, just ramp to/from cloverleaf. They get about a 1/2 mile "boost".

Not saying the driver was doing this, but could have been forced by someone gaming the system.

The 101-S to 85S carpool lane flyover is a one way trip to 85 S.
Google Maps

 

[BobbyKings]

Can we please,

please,

please,

please stop speculating.



In the name of dignity and integrity.

Please.

 

[Peter Lucas]

Even in old Streetview pic it looks like someone was confused about that not being a lane:

View attachment 288665

I don't think that photo shows the location of the accident. I know that the signs look right. I found another Google street view which showns a barrier that more closely matches photos of the accident. (The highways signs a bit further back) The barrier is in a "lane" that is ever widening as it approaches the barrier. I could imagine that if you executed an autopilot lane change at the right (wrong) moment, the car could put itself into this (not) lane. But it would require some serious inattention.
I understand that Tesla says that there are 200 autopilot trips on this stretch of road per DAY. So it was either a one-off malfunction. Or more likely "operator error". Poor guy. And poor family.

 

[NeverFollow]

As soon as the right signal is on, the dashboard shows that Autopilot happily complies to the command
to cross the white solid line with absolutely no hesitation:

When you are using AP, can you please explicitly describe how the AP behaves when you trigger the right or left sign?

I would like to know if the AP understands that you would like to change lane
BUT the AP verifies that it is safe to do so,
in particular that there is no car in the other lane
(like I would do by looking at the left or right mirror and checking twice the blind-spot area).
And then slowly change lane, and turn the turn sign off automatically after completion.

Or do you need to check by yourself first if it is safe to change lane before triggering the turn sign signal?
So then the AP starts moving the car in the corresponding lane
(But you are still in charge, meaning that if you put back the turn signal,
then the AP will merge in the original lane.

How much delay occur between the moment you trigger the turn signal and the moment the car starts changing lane?

When I learned driving, I was taught to check first if it was safe to change lane,
and then trigger the turn signal and wait then about 3 seconds
to be sure it was safe to do so and that other cars might notice my intention to change lane,
and then finally I would change lane.

 

[Icer]

Nobody knows if auto-steering was active on the car at this point correct?

I have seen zero bits on this.
Tesla blog response,

• Due to the extensive damage caused by the collision, we have not yet been able to retrieve the vehicle’s logs.
• We are currently working closely with the authorities to recover the logs from the computer inside the vehicle. Once that happens and the logs have been reviewed, we hope to have a better understanding of what happened.

Tesla's last data from the car must indicate AP was on.
Otherwise they would have simply stated
"last we know, AP was off".

 

[mongo]

I have seen zero bits on this.
Tesla blog response,

• Due to the extensive damage caused by the collision, we have not yet been able to retrieve the vehicle’s logs.
• We are currently working closely with the authorities to recover the logs from the computer inside the vehicle. Once that happens and the logs have been reviewed, we hope to have a better understanding of what happened.

Tesla's last data from the car must indicate AP was on.
Otherwise they would have simply stated
"last we know, AP was off".

I firmly disagree.
It is quite plausible, and likely, that they have no time relevant data for this car. The last data they had from vehicle could have been at midnight when it sent a cumulative status update packet to Fremont. The cars are not live streaming every event.

 

[SMAlset]

@BobbyKings, don't see anything wrong with speculating here on what could have happened as everyone here knows we don't know what happened yet and pretty much have framed discussion that way. If you have followed this thread from the beginning, you know that new people have come into the thread mid-stream all along the thread and offerred their own thoughts many times expressing theories or information already posted. Happens in long threads. I think people have been pretty respectful in their discussions of the victim, and we all consider him part of our extended Tesla family, and we are pretty much discussing the car and roadway aspects. His death in this has affect many of us deeply even if we didn't know him. The differing points of view actually help educate people. I also think we've exposed many important issues to be aware of that could be improved, especially pertinent for those that drive this section of road, myself and husband included, and whether you are driving a Tesla or other car or on AP or not. I know I've learned a lot from even this thread about the roadway, barrier types, lane markings and aspects of our Tesla I wasn't aware of. I realize you are out of country and probably can't imagine how heavily trafficked this section is and how many accidents have already happened at exactly this point. Better that people who drive the area regularly or come through on business or pleasure are made aware of how things are if they are reading this forum. At least that's my view.

 

[McRat]

We're not talking shoulder white lines here but those of merging roadways where they are used. This is typical in our area with all the morning metered entrance ramps and other major roadways that combine. The video that prompted the white line crossing discussion (originally thought by poster showing AP did it in error) has the merging roadway single white line (see below). Believe we have 3 lanes continuing on the left and a separate merging/exiting 2-lane roadway to the right. You can legally cross this white line to reach your desired highway path, assuming you've checked traffic. Lots and lots of this example here in the bay area.

They are everywhere, and often the 'white line jumper' will come to a full stop in a moving lane waiting to cross the white line to jump some congestion. It is a good reason to have a dash cam because if you do hit somebody who suddenly stops in their moving lane, you'll hit them in the rear. We have 'stoppers' at the I-15 to 91 interchange several times an hour and a few collisions a week. Since there is another ramp shortly after the gore point, you are somewhat forced into the line jumper's lane.

But this impact was in a triangle formed by 2 single white lines. Driving in there is driving in the shoulder, or more accurately, not driving in a traffic lane.

And there is a catch all CVC 23103. This allows a LEO to cite you if they believe you are driving in a dangerous manner. It can be pretty much anything. It's a 2-pt like a DUI.

 

[bhzmark]

it has been confirmed that this had not be reset the day before the accident and the previous accident had been on march 11

Multiple factors likely contributed to this collision. Likely some driver error (unless other cars somehow forced him into the barrier) but certainly, based on the info we have now, Caltrans failure to reset the barrier for so long after the last accident is deadly. I hope someone is fired over this negligence to send a msg to the rest of the agency that they must do their jobs and prioritize the the relatively cheap and easy maintenance that can save lives.

In addition to the barrier, the whole design of the highway was a cluster and should be redesigned. The use of simple chevron and plastiic barriers etc, would also be cheap and easy improvements.

some Caltrans person saw the video of the accident and had it hit home.

I reported that attenuation barrier being collapsed on Sunday using the form on the CALTRANS Traffic or Work Zone Concern:

NTSB determined there was a lack of adequate highway markings. https://www.ntsb.gov/news/press-releases/Pages/PR20170328.aspx

 

[LMNOP]

no one else has corroborated that their AP behaves the same

There was at least one poster here who corroborated exactly this earlier in this thread: Model X Crash on US-101 (Mountain View, CA)

the bottom line is that you can’t blame AP

I don't think anyone is trying to blame AP or find fault in any single thing. We're only trying to determine if AP was a factor in the collision. I think this is very important for the future of autonomous driving and the public's acceptance of it. If the public, or Tesla, or the state agencies that maintain the roads and highways can learn something from this crash to prevent others like it, that's the preferred outcome of this discussion.

Of course you can't blame AP, because AP is not supposed to be fully autonomous or perfect and every Tesla driver knows that they have to stay in control. Walter was a very smart man, a career software engineer, and a technology enthusiast. He understood the limitations of AP as well (if not better) as anyone.

As in most serious crashes, there were multiple failures here that contributed to the cause and the fatality of this crash. I don't think any single factor can be pointed as "blame". In this case, knowing the driver and how he typically used AP, I think AP was a factor. The driver may have been distracted or not paying attention; also a factor. The improperly reset attenuation barrier was also a factor in the severity of the damage and injury to the driver. The lane markings and lack of stripes, rumble strips, or chevron patterns in the gore point are also a factor. The position of the sun may have also been a factor. The position of other cars on the road -- also a factor. I think you get my point.

I am also a software engineer (not at Apple). I do not write software for self-driving cars, but the principles are the same regardless of the type of software. Whenever there is a critical failure in a system, you must identify and rank all the factors that contributed to the failure. Autonomous vehicles are a very challenging problem, because many of the factors are external (i.e. roads, other vehicles, weather, lighting, etc). Most well-designed systems require multiple points of failure to fail in a catastrophic way, and I think that's what happened here. So ... what can we learn from it? We have to identify all the factors and see which of them are practical and feasible to mitigate in future similar scenarios.

Could it be that something as simple as some striping or chevron paint in the gore point could have prevented this? Maybe.
If Caltrans had reset the attenuation barrier prior to the crash, would Walter still be alive? Maybe.
If the vehicle's front radar was programmed to alert an inattentive driver to an upcoming stationary object in the path of travel, could that have prevented this? Maybe.

It's 2018 and autonomous driving is in it's infancy. I remember when Walter first got his new Model X and I rode in it with him for the first time, I was very impressed with the car in general and particularly the technology features. I was jealous. I wanted one, too. Walter was an early adopter of technology like me, and was likely just as excited about the promise of a fully autonomous driving future as I am. Self-driving cars are going to be more and more a part of our lives in the coming years. If one good thing can come out of this tragedy, it would be improvements in the technology or in the physical roads/markings to make it safer and easier for both human and computer drivers in the future.

I think it's understandable that Tesla becomes the poster company in the media for these type of accidents right now.

Last point: I don't think Tesla is being singled out here. It's self-driving technology (and particularly the failures) that have grabbed the public's attention recently. The fatal self-driving Uber crash with the pedestrian seems to be getting a lot more national press coverage than this crash. It was a Volvo, not a Tesla.

 

[bonnie]

I
Why he would continue to use something that put his life at risk repeatedly makes no sense - in my opinion stupid.
Why would he be in the 85 HOV exit lane if he needs the south bound 101 HOV lane? (Ok, not proven but it appears to me that way).

From everything I've read, he doesn't seem to be stupid.

There is no info yet on whether he was using AP or not, but as I said earlier ... IF he had been having problems in that area with AP, I'm relatively sure he would not have had AP engaged.