Cake
  • Log In
  • Sign Up
    • dvhart

      "Autopilot system was engaged by the driver in the seconds before a fatal crash last week, raising more questions about the safety of self-driving technology on public roads."

      Frustratingly, this isn't self driving technology. This is adaptive cruise control and Lane keeping. This is SAE L2 technology, commonly found across mass market vehicles like Honda and Toyota, as well as luxury vehicles. Unfortunately, WSJ isn't wrong.

      Because it is named autopilot and humans refuse to read the fine manual, drivers are lulled into thinking the system is more capable than it is. I expect Tesla to add more "hands on the wheel" requirements, and for the forums to be full of people complaining about no longer being able to read a book on the way to work (again).

      Tesla Says Autopilot Was Engaged in Fatal Crash Under Investigation in California https://www.wsj.com/articles/tesla-says-autopilot-was-engaged-in-fatal-crash-under-investigation-in-california-1522462409

    • Chris

      Hmmm, really interesting point about Lane Keeping technology. I hadn't made that distinction. Unfortunately, here's the headline from Forbes:

      Autopilot Steers Tesla Model X To Fiery Crash

      Yikes. Forbes also interpreted Tesla's blog to have some victim-blaming, especially this paragraph:

      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 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.

    • dvhart

      We will of course have to wait for the NTSB to complete their report, which can take months - and fortunately they are pretty good at being objective. There are open questions about the safety of the road and whether or not AP2 failed to avoid the barrier, or if it steered the car into the barrier. Regardless, taking your hands off the wheel for 5 seconds with L2 driver assistance technology is driver negligence. The Forbes article also suffers from some logical fallacies, for example: "But one reason for the smaller number of fatalities could be that Tesla has far fewer vehicles on the road". The numbers reported were accidents per miles driven - these are relative numbers, and therefor independent from the total number of vehicles on the road. I don't mean to imply Tesla is blameless here, I have long held that marketing an L2 system as "Autopilot" would lead to people trusting it more than they should. I maintain that Honda made a better human interface design decision with their lane keeping, which allows you to override it without disabling it. In a Tesla, if you override the steering, AP disengages - providing the illusion that it is either car or the human driving - and not the car assisting the human driver. This is something they should be required to change in my opinion.

    • yaypie

      My heart breaks for the driver's family and friends. I hesitate to even discuss the technical details of this accident because the most important thing here is that a person lost their life, and it shouldn't have happened. But I think discussion is valuable in the context of saving lives in the future.

      I'm eager to see the NTSB report when it's finished. But from what I've read so far, I think responsibility can be divided between three parties, in the following order:

      1. The driver
      2. Caltrans
      3. Tesla

      I assign the most responsibility to the driver only because the driver of a vehicle always has a responsibility to be in control of that vehicle. It's not victim-blaming to say that the person operating a car might have been partially responsible for a car crash.

      That said, I strongly feel that Caltrans bears significant responsibility for this accident. The only reason I assign more responsibility to the driver is that the driver is ultimately the one most able to adapt to less-than-ideal circumstances in the moment. But as far as I'm concerned, this crash was preventable and Caltrans is the entity most at fault for failing to prevent it.

      Collisions with this same barrier at this same interchange are so common that Caltrans can't even manage to keep the crash attenuator — which is meant to absorb an impact rather than allowing a car to run directly into the concrete barrier — in good repair. After an impact, the crash attenuator must be replaced, but Caltrans has frequently failed to replace the attenuator at this location. When the Tesla hit it, it was already collapsed and provided no protection.

      Furthermore, the entire interchange is poorly designed. The white lines on the approach to the barrier appear to form a lane. There are no painted hash marks indicating that the area is not a traffic lane or hinting that there's a barrier ahead. From a distance, the barrier is difficult to see even on a clear day due to its short height and narrow width.

      Human drivers are confused by this interchange often enough that it's a frequent location of accidents. Why hasn't Caltrans improved it? Why haven't they at least painted hash marks across the approach to more clearly indicate that it's not a lane? Why haven't they mounted a larger, more visible warning sign and a flashing light on the barrier to make it more visible? This is inexcusable negligence.

      I assign the least amount of responsibility in this incident to Tesla. Tesla is very consistent about describing Autopilot as a driver assistance system. When the driver engages Autopilot, a warning is displayed informing them that they must still monitor and maintain control of the vehicle, and that they must keep their hands on the wheel at all times.

      Tesla is transparent about Autopilot's limitations in their manuals and marketing material. Autopilot is not good at detecting small stationary obstructions in the road. It's not good at discerning between actual lanes of traffic and markings that just look like lanes. It's not good at handling unusual scenarios. This is why it requires constant human attention, with hands on the wheel and eyes on the road.

      Autopilot can improve, and I expect it to, but I don't blame it for this accident. I think driver assistance technologies like Autopilot are far more beneficial to safety than almost any other advancement in automobile technology, even though they're still far from perfect.

    • dvhart

      Excellent breakdown - and thanks for sharing the streetview of the scene - this adds a lot of necessary context (I should have thought to do that). I'd agree this appears to be a pretty dangerous situation in general.

    • cabbey

      My 2015 Lincoln has adaptive cruise and lane keep assist. It is effectively the same level of functionality as the current Tesla “Auto Pilot”, but anyone with experience with this feature set would laugh at the suggestion of calling it an auto pilot. I use both on a regular basis. But I would never in a million years consider taking my hands off the wheel or my eyes off the road, let alone both at the same time. I don’t yet have any experience with the Tesla “Auto Pilot” functionality... but on paper, what it is capable of just doesn’t deserve that name. There gets to be a point where aspirational names need to be tempered with more than just warnings.

      So yes, the driver bears significant fault in this, and has paid the ultimate price.

      And yes, CalTrans bears some serious fault in this, as evidenced by how often human drivers have issues with the design and layout of this area.

      But someone at Tesla also bears some not insignificant fault for the misleading name and the over promised capability that is “Auto Pilot”.

    • yaypie

      I've heard this criticism from many people. I assume it's common because a lot of people have an incomplete understanding of how airplane autopilot systems (after which Tesla's Autopilot is named) actually work.

      In airplanes, autopilot doesn't make a human pilot unnecessary. For a long time, the autopilot features in airplanes weren't even computerized; they were relatively simple mechanical systems incapable of analyzing input or making decisions. Even with modern computerized autopilot systems, pilots must remain alert and ready to take control at any moment.

      Airplane autopilot systems are a pilot assistance technology that frees pilots from mundane tasks like maintaining a desired altitude, holding a particular rate of climb, maintaining a given air speed, turning to follow navigation waypoints, etc., but they don't obviate the need for a pilot to actually be responsible and ready to take control of the aircraft at all times.

      Tesla's Autopilot feature has precisely the same characteristics. It assists the driver with the more mundane aspects of vehicle control: maintaining speed, staying within a lane, reducing speed when traffic slows down, etc. It's not an autonomous system, and it isn't capable of dealing with complex or unique scenarios.

      I think Autopilot is very aptly named.

    • cabbey

      Exactly the point, the vast majority of people do not understand just how simple minded an aircraft autopilot is. They think it's way more complex, and capable, than it really is... and they assume based on that that their car is way more capable than it actually is.

      Autopilot may be the perfect name for it from an engineering perspective, but from a marketing perspective it's all wrong.

    • I drive this stretch of 101 daily. It's a complex interchange, that's true. The Gore Point might be better marked but I have never seen anyone confuse it for a lane. Do they make stupid decisions by crossing the GP? Yes. Whether it's to skip to the "head" of the southbound 101 carpool lane or to cut into Southbound 85 carpool lane, it happens all too often. Drivers should know that a solid white line is meant to discourage them from crossing-it rarely does.

      From what's been offered story wise, the driver bears significant blame. I'd say Tesla next because it's not easy to put out a Lithium fire and any fire truck responding to the scene would need much more than the 500 gallons of water it carries to cool the batteries enough to put them out-according to reports, the vehicle had to be disassembled on scene to put the fire out. Surely, that was very time consuming. it's not clear first responders could have made more of a difference had they been able to put the fire out sooner (MV Fire said the nearest hydrant was over 2000' away). I don't remember the condition of the attenuation barrels but rarely are they in great condition-so Caltrans will bear some responsibility for their condition.

      On a side note, my neighbor is an engineer for Tesla and regularly drives one home. Some time ago, it was struck from behind in front of my house by a drunk driver (I asked what the estimated speed at impact was-60 was the guesstimate). The impact took the rear wheel assembly off of the Tesla and pushed it into a mini van that ended up on my neighbor's lawn-about 50' away and up the curb. Except for the missing wheel and some not so obvious deformation of the B pillar, the car was in pretty good condition. So it surprised me to see how badly damaged to Tesla was. I'm sure much of the damage was from impacting the GP and from efforts to extract the driver.

      Sad event for sure. Lots of questions but I believe the real message needs to be "drive the car".

    • just saw something that implied the driver had visited the service department for the autopilot malfunction. It will be interesting to see what comes of that.

      As I say, it’s incredibly sad someone died in that accident.

    • yaypie

      From what's been offered story wise, the driver bears significant blame. I'd say Tesla next because it's not easy to put out a Lithium fire and any fire truck responding to the scene would need much more than the 500 gallons of water it carries to cool the batteries enough to put them out-according to reports, the vehicle had to be disassembled on scene to put the fire out. 

      It's true that lithium-ion battery fires are difficult to put out, but the fire happened after the crash. It doesn't have anything to do with what caused the crash, so I'm not sure how it's relevant to use it as the basis for assigning Tesla responsibility for the crash.

      I do agree, though, that the difficulty of putting out lithium-ion battery fires is a big problem for electric vehicles. But my impression from reading about accidents where electric cars have caught fire is that it seems to take a much more catastrophic crash to set an EV on fire than it does to ignite an internal combustion engine car.

      I was once involved in a multi-car accident on this very same highway (just a few miles south of where this crash occurred) that didn't even result in any injuries, but the ICE car that caused the accident caught fire instantly and burned to a crisp in the middle of rush hour traffic while we waited for firefighters to arrive. It was easier to clean up than an EV fire would have been, but I'm pretty sure an EV car wouldn't have even caught fire in this scenario.

    • The fire is relevant because until the scene is safe, firefighters can’t really do much without significant risk to themselves. And it also makes a difference because it’s clear firefighters do not have enough resources on a single apparatus to put an ev fire out.

      To your point tho, while it may not have contributed to the crash, it does delay the response and that may have made a difference in the driver’s survival.

      Your experience in a relatively minor accident where a vehicle caught fire shows it’s not necessary for a major collision to have a fire.

    • Chris

      Once upon a time I was a geophysicist working in the environmental industry and it's really scary to see what gets in our water. I can't see a fire like this and not think of the toxic fluoride gases and lithium salts being washed away by thousands of gallons of water and later rain. 😬

    • That’s a very good point @Chris.

      I drove past the accident site today and for the most part, it’s cleaned up (visually). No attenuators unless you count the K rail and signage.

    • yaypie

      I’m not sure you understood me. I was specifically citing my experience as an example of an ICE car easily catching fire where an EV almost certainly would not have caught fire.

      ICE cars seem to catch fire all the time, sometimes due to minor accidents, presumably because they’re full of flammable things in the places most likely to be damaged in accidents. But it seems like we only see EVs catch fire in more serious accidents, probably because it takes a lot of damage to actually breach a battery pack and there’s not much else that’s highly flammable in an EV car.

      If anyone knows a good source for actual statistics on this, I’d be very interested.

    • yaypie

      It turns out in 2016 a Greyhound bus ran into a barrier at the 101/85 interchange in South San Jose, which is virtually identical to the interchange in Mountain View where the Tesla crash occurred. Two people were killed and over a dozen were injured.

      The NTSB investigated the Greyhound crash and found Caltrans at fault because the interchange had inadequate markings (it looked too much like a normal lane) and the barrier had insufficient signage. Caltrans had also failed to repair the crash attenuator after a previous crash, so the bus impacted the barrier with full force.

      The NTSB recommended that Caltrans add cross-hatched pavement markings, add more visible reflective signage, and improve their procedures to ensure that damaged crash attenuators get replaced quickly.

      Caltrans apparently did none of those things. Two years later, both this interchange and the Mountain View interchange remain just as deadly as before.

    • Us

      CalTrans inability to repair or make our freeways reasonably safe is a major reason I cage it work these days instead of riding the motorcycle. yeah tax dollars at "work"?

    • Us

      The amount of budget to rebuild and f- with the rest of 101 between menlo park and mountain view is nuts. To paint doesn't seem like a big deal but I do not know the budget items and issues. This stretch has been in a constant repair/change for the last 10+ years.

    • It all could have been done by now except Caltrans wouldn't let two contracts be one. So contractor A did his thing and then contractor B tore A's thing down and did his thing. Where's the sense in that? Hence 10 years of construction.

      It's also true so much money that should be going to roads and infrastructure goes elsewhere. It's more this that I am referring to.

    You've been invited!