TMC is an independent, primarily volunteer organization that relies on ad revenue to cover its operating costs. Please consider whitelisting TMC on your ad blocker or making a Paypal contribution here: paypal.me/SupportTMC

Does using AP increase safety?

Discussion in 'Model S' started by sillydriver, Sep 14, 2016.

  1. sillydriver

    sillydriver Member

    Joined:
    Oct 19, 2014
    Messages:
    542
    Location:
    Middleburg, va
    Does using AP increase safety? We can infer Elon thinks it does, given his comparison of the death rates on roads against Tesla’s one fatality. But I’m still raising this question because I think 1) it’s important and 2) it’s not so simple. Unfortunately it’s also likely to be contentious.


    I’m not asking “Is using AP safe?” because that clearly depends. It depends on the circumstances of its use. Obviously, turning on AP at random during a drive and letting it run without further intervention will probably result in a crash. None of us would do this, but the point is that the safety of AP is a function of the combination of the driver and AP itself, since 1) AP is only designed to be turned on in certain settings (for example, divided highways) and 2) it sometimes needs driver intervention even in those settings. In particular, the driver must be ready and able to take control immediately, and most of us have plenty of personal experiences where such quick intervention was necessary for one reason or other.


    So the safety of AP depends on the conscientiousness, attentiveness and skill of the particular driver using it, which is why “Is using AP safe?” isn’t meaningful. What is meaningful is a comparison of that driver’s projected accident rate using AP, given the choices they make when using it, and their projected accident rate just driving the same routes by themselves. Or in terms of statistics, the meaningful comparison is between the accident or death rate of a population of drivers using AP, and a population of equally conscientious, attentive and skilled drivers who are driving Teslas without using AP over the same routes where the others use AP. This comparison would show whether using AP increases safety.


    The problem is that this is nothing like Elon’s comparison. First, AP is “intended for use only on highways and limited-access roads” (according to the manual) and I assume our AP use is at least skewed in that direction. I’m having trouble finding statistics on fatalities per vehicle mile on interstates, but I assume divided highways are significantly safer (which makes the comparison for AP tougher) than all roads in the aggregate. Second, Teslas are intrinsically safer than the average car, so there would be fewer deaths in the comparison group of non-AP cars if they were all Teslas. Also, overall fatality per vehicle mile statistics include motorcycles, which are less safe. Third, I expect the average Tesla driver is a safer than average driver, even when when they drive ICE cars, simply because their being able to afford a Tesla means they are probably older. The result is that the figure for fatalities per vehicle mile that Elon should compare to is significantly lower than the nationwide average number. Of course Tesla’s one fatality isn’t statistically meaningful anyway, but Elon’s comparison is not a good argument that using AP increases safety.


    But what about other arguments in favor of AP being safer? One could argue that if AP is used exactly as instructed in the manual, then it is extremely safe: safer than driving. But that’s because the driver would always cover for AP’s mistakes: the manual requires that we must “always be prepared to take immediate action.” The difficulty is that I suspect that about the only human capable of always being prepared to take immediate action – who takes no extra time to refocus their full attention – is Sully Sullenberger. So the relevant comparison against AP is Mr. Sullenberger driving an intrinsically safe Tesla down a limited-access highway, which is about as safe as safe can be. I don’t see how AP being used exactly as instructed can be safer than that.


    But then there are the rest of us fallible mortals who are trying to do what the manual wants within the psychological limits of attention and skill. So for us fallible drivers, which alternative is riskier: 1) correcting the wheel in an instant when we are completely surprised by AP balking at a tight curve, an asphalt patch covering lane lines, a reflection driving toward the afternoon sun, or the crest of a hill; or 2) driving ourselves through a tight curve, over an asphalt patch, into the afternoon sun, or over the crest of a hill? I think there is no question that (1) is much less safe than (2). In other words, AP reduces safety in these cases, it doesn’t increase it.


    But these cases need to be balanced against the cases where AP would be safer, which is where the driver is sleepy, or loses attention because of a monotonous road, or because of a long stretch of stop and go traffic. Which way does the overall balance go? Is the safety increase in these situations enough to boost overall safety beyond that of driving one's self? I don’t know, but I doubt it. I definitely think that driving the car one’s self is currently (under 7.1) slightly safer as long as these conditions don’t apply.


    All of this is not meant as a criticism of Tesla. They are trying doing great things in automated driving, and I believe they will eventually succeed. So I guess my conclusions are:


    1) Except for sleepiness, monotony and stop-and-go, supposed safety is not the reason to use AP. Convenience is generally what would justify its use.


    2) The safety focus for using radar in 8.0 is a good thing, because that will shift the balance to some degree. It needs to be shifted because AP is not now intrinsically safer than driving.


    3) I hate to say it, but enhanced nags to force us to hold the wheel are probably a good thing. The argument against nags would be stronger if AP was as intrinsically safe as Elon says, but it isn’t.


    I’m looking forward to 8.0.


    What do you think?
     
    • Love x 2
    • Like x 1
  2. Austral

    Austral Member

    Joined:
    Jul 4, 2016
    Messages:
    173
    Location:
    McLean
    #2 Austral, Sep 14, 2016
    Last edited: Sep 14, 2016
    AP drives better than 75 percent of the drivers around me in the Philadelphia, DC metro, and Hampton Roads region. Fact is that it is too easy to get a license to drive in the US but everyone thinks they are the best. In addition, I should like to note that the behavior of people on their cell phones matches that of drunk drivers. Over 30,000 people a year are killed on the roads every year (10 times the number killed in 9-11). It's not speed that kills; it's bad drivers with no skill. You can get a license in the US without ever touching a skidpad (majority reading this comment probably never heard of the term "skidpad"). That's insane. In the DC metro area, you don't even have to prove you know how to parallel park. I discovered that after moving back from Germany and had to retake my exam. The entire controversy over AP is just bash@# crazy considering the bigger picture on driver safety. AP, prior to v8.0, is absolutely an improvement in safety for all. #justmytwocents
     
  3. sillydriver

    sillydriver Member

    Joined:
    Oct 19, 2014
    Messages:
    542
    Location:
    Middleburg, va
    It's a fair point that an idiot texting while eating a filet-o-fish may be better off using AP. I should have included that in the list of conditions where AP is the safer choice.
     
  4. Austral

    Austral Member

    Joined:
    Jul 4, 2016
    Messages:
    173
    Location:
    McLean
    It's all good. Pardon my old man rant. :)
     
  5. chillaban

    chillaban Member

    Joined:
    May 5, 2016
    Messages:
    581
    Location:
    San Jose, CA
    In terms of safety, I think you are understating the significance of "sleepiness, monotony, stop-and-go" effects. Humans are terrible at paying attention to what's in front of them. Every day on my drive to work in the Bay Area I see one or two rear-ending accidents, and I've already gotten one of my cars totaled by a driver that's simply not paying attention. Sensors always pay attention, and deficiencies in sensors can be addressed and rolled out to the whole fleet. There is no "be better at paying attention and ignoring family member texting you about a health problem" OTA for humans. Well if there is, I haven't gotten one yet.


    But in general, what Elon was saying was two-fold. One, Autopilot/TACC virtually eliminates rear-ending accidents to truly exceptional events (such as T-boning an unrecognized car or a lane-offset approach to an oddly shaped vehicle that's not visually recognized by the camera as being a car).

    Secondly, in non-AP mode, the car's collision mitigation systems err on the side of trusting the driver. As a result, it delays warnings and emergency braking action until the very last second when it's absolutely sure that the driver isn't going to react in time. It does this because humans tend to find false emergency braking / collision alarms to be extremely annoying and after a few false alerts will probably just ignore the system. In AP mode, the car is in full control and knows what it's trying to do. So if it sees an obstacle, it can initiate action (whether that's light or heavy braking or steering in 8.0) much earlier and more gradually. In this sense, it's "safer" in that it can gradually execute avoidance maneuvers well in advance, rather than being in a penalty box and only being allowed to take last-minute emergency-input maneuvers. It's also worth mentioning, even in AP mode with a distracted driver, sudden changes to input by the AP tend to grab our attention. If the car changes from maintaining speed to slowing down, even if we were distracted at the moment, it's probably enough to immediately cause us to glance at the road, and take action to override the AP or agree with the AP and slam on the brakes. In that regard, AP's inputs do buy an additional safety margin vs just the driver.


    (Oddly enough, you can see a lot of folks who are not happy about the second behavior. v7.1 already has a lot of circumstances where it will suddenly start applying max regen braking to slow down for something it thinks is an obstacle, which either shows up as a "ghost car" on the display or as an exaggerated diagram of the car next to you leaving its lane and cutting into yours. We really don't have much of a tolerance for false braking events even when they frankly have little to no safety implications in terms of "causing" a chain reaction rear-ending by slowing down 20mph with 50kW of regen braking….)
     
  6. sillydriver

    sillydriver Member

    Joined:
    Oct 19, 2014
    Messages:
    542
    Location:
    Middleburg, va
    #6 sillydriver, Sep 14, 2016
    Last edited: Sep 14, 2016
    I agree that 8.0's reducing rear-ending in monotonous stop-and-go is a big deal: maybe bigger than I give it credit for. Personally, I'm not sure I trust 7.1 enough to hand over driving to AP in a stop-and-go situation. You mention a couple possible fail situations in parentheses above, another may be where the car being tracked ahead moves to another lane exposing a stopped car in front of it. It's great that 8.0 is bringing more radar data to the task where it should handle this case.

    EDIT: to some degree the Forward Collision Warning helps prevent rear-ending without engaging AP. I triggered it once when I was glancing at the big screen map on a busy street while car's in front hit the brakes. The alarm sounded and I braked well short of the car ahead. Without it it would have been close.

    I think this discussion is good because it is highlighting situations where AP does enhance safety.
     
  7. chillaban

    chillaban Member

    Joined:
    May 5, 2016
    Messages:
    581
    Location:
    San Jose, CA
    I actually use the current AP in stop and go traffic all the time thanks to living in silicon valley. I think the limitations are actually much overstated. I came from a 2014 Audi A6 which literally refused to acknowledge stopped cars exist 100% of the time, and won't recognize a car half-in your lane 100% of the time. That was borderline frightening to use in stop and go traffic…


    I've found that the Tesla system in the current form actually handles most of these situations you mentioned (including cut-out, cut-in, and encountering stopped ordinary cars), more often than not. There's only been one or two occasions where it's been too close for comfort and I intervened before finding out whether or not the car will handle it. As an aside, my rant here is, when in doubt take over control — other drivers on the road are NOT your AP science experiment volunteers!

    8.0 should make these situations even better. But the key is paying attention. It's really easy to detect when AP is not acting as it should as long as your face isn't buried in your smartphone :D. I think part of the problem is that Tesla handles so many of these situations correctly that a lot of drivers who did not read the manual would not be aware of these limitations. Plus I think the manual (dated April 2016, TACC text largely unchanged since TACC was released with v7.0) is describing the worst case limitations from the earlier firmware versions and not referencing a lot of the improvements in 7.1 that have refined these situations.


    (Once I was in the same car as a friend with a similar Audi. We were in stop and go traffic, and the car in front pulled out and revealed a stopped VW Jetta. The Audi immediately started accelerating heavily and he slammed on his brakes, and was like "WTF THE CAR JUST TRIED TO CRASH ITSELF". I asked "did you read the manual?"….. "why would I do that?")
     
    • Like x 1
  8. sillydriver

    sillydriver Member

    Joined:
    Oct 19, 2014
    Messages:
    542
    Location:
    Middleburg, va
    #8 sillydriver, Sep 14, 2016
    Last edited: Sep 14, 2016
    I'm happy to say you are much more an expert than I because I retired to a farm in the county seven years ago and have since then driven in stop-and-go commuting traffic exactly once -- which was solely for the purpose of testing AP.

    It worked.

    Nevertheless I appreciate the importance of avoiding rear-end collisions. Back when I was working in Boston about a decade ago I was rear-ended by some clown in stop-and-go traffic. His car was wrecked, mine was dented. Audis are also good cars. :)


    IMG_0047.JPG IMG_0046.JPG
     
    • Like x 1
  9. msnow

    msnow Active Member

    Joined:
    Jul 14, 2015
    Messages:
    4,275
    Location:
    SoCal
    Very well written and thoughtful thread. Nice job @sillydriver !
     
  10. sillydriver

    sillydriver Member

    Joined:
    Oct 19, 2014
    Messages:
    542
    Location:
    Middleburg, va
    Thanks!:)
     
  11. S4WRXTTCS

    S4WRXTTCS Active Member

    Joined:
    May 3, 2015
    Messages:
    1,180
    Location:
    Snohomish, WA
    The problem with a question like this is what we choose to compare it against.

    In the media it's constantly compared against an unimpaired driver which is a bit silly because where are the unimpaired drivers? In the Seattle area people are either drunk, high, or using their phones. In some cases it's probably all three.

    I see AP as desperate move, but necessary move to save us from ourselves.

    Even if it's less safe in the short term the long term dividends will pay off big time.

    There are four primary negative elements of the current version of AP that hampers it's safety. I'm only referring to the full AP, and not just TACC.

    1.) Abuse in that drivers ignore how you're supposed to use AP. It works deceptively well for what little sensor technology is actually there. This deception leads some experienced drivers into assuming they know it's limitations/etc, but then they don't account for some once in a blue moon event. Something like trust lust, or a stalled car that suddenly appears in front of them.

    2.) Using AP causes a lose of situational awareness. There is just something that happens when I allow the car to steer itself. I try to fight against that by at least having one hand/finger on the wheel for feedback. But, I still find my situational awareness to be lacking compared to what it is when I'm engage. This is precisely why lots of people feel relaxed and energized after a road trip with AP. They do because all that energy that went into situational awareness wasn't used. So of course you're going to feel relaxed.

    3.) Control hand off - In the current AP there have been drivers assuming that AP was still active, but it's wasn't. That they disabled it or overrode it without realizing it. This is why the logs are so important when assessing why an incident happened.

    4.) AP allows a person to drive considerably more miles without the same level of exhaustion, and this leads people to drive way more miles than they would otherwise. The chance of an accident goes up as you're on the road more. It prevents people from taking safer options like getting on a plane. I drive WAY MORE now days now that I have a Tesla than I ever did before, and AP played a significant role in that.

    There are some benefits that enhance the safety.

    1.) Diminished road rage - AP turns you into a softy. Someone is riding their brakes? Oh, well. Or if people are constantly speeding up and slowing down? So what?

    2.) Less speeding - Just set the speed and forget about it. No looking down at the speedometer wondering why you're doing 100mph.

    3.) A co-pilot for distracted moments - We all have distracted moments and you're lying if you say you're never distracted. Sometimes it's a text in a bad moment. Or sometimes it's some beautiful scenery. Or maybe it's a spider that's crawling on you.

    4.) Ceasing lane drifters. Even people who are paying attention sometimes just drift.
     
    • Like x 3
  12. roblab

    roblab Active Member

    Joined:
    Jul 15, 2008
    Messages:
    2,038
    Location:
    Angwin (Napa Valley) CA
    When you use it, you begin to trust it. You learn what it can do well, and what it has trouble with. You understand it. Most people have a hard time just trusting something they don't know about. "I'm a doctor. Trust me." Maybe not, right?

    AP in stop and go traffic is fantastic. Far, far better than the average human. You ought to try it and see.
     
  13. S4WRXTTCS

    S4WRXTTCS Active Member

    Joined:
    May 3, 2015
    Messages:
    1,180
    Location:
    Snohomish, WA
    TACC/AP works pretty much optimally in stop and go. It's absolutely the place I use it most often. It's absolutely worth the price in that alone.

    I feel as if AP drops off a bit when going above 75mph or so.
     
  14. Hota

    Hota Member

    Joined:
    Jun 27, 2016
    Messages:
    53
    Location:
    New England
    That's the crux of the problem. To assess real safety, you can't just look at the system alone without taking into account the driver and his/her interface with the system even if the driver is using it not as intended.

    With AP, I guarantee that most of us probably are distracted a hell of a lot more than without AP because it works so well.

    Those second-long glances at your smartphone that used to be 1 second at a time in your old ICE now slowly creep up to 5 seconds and 10 seconds with AP engaged. You fumble with Slacker 30 seconds at a time trying to find new podcasts. Pretty soon, you'll have gotten to where you were going and realized you were "heads down" maybe 75% of the time only glancing up when you felt a slight deceleration, unexpected drift or a nag from the computer telling you to place your hands on the wheel.

    Now is that safer than the same driver having to manually drive while trying to text a few messages here and there? Maybe...maybe not. Who knows.
     
    • Like x 1
    • Love x 1
  15. RogerHScott

    RogerHScott Active Member

    Joined:
    Mar 2, 2016
    Messages:
    1,356
    Location:
    Lafayette, IN
    I would certainly agree that Tesla's "statistical" arguments about AP safety are deeply flawed. They have the data to present
    the convincing argument: what are the comparative accident rates for Tesla drivers, on AP-approved roads, with and w/o AP?
    I'm not holding my breath waiting for them to share this data.

    As for the likely reality, I think if you look at the circumstances under which at least certain classes of accidents occur you'll see
    a significant correlation with the tired/bored/stressed factor, so preferentially benefiting those drivers is likely to result in a
    disproportionate safety improvement.

    This is a really complicated, multi-dimensional issue. "Safer"/"less safe" are probably simplistic characterizations.
     
  16. EinSV

    EinSV Member

    Joined:
    Feb 6, 2016
    Messages:
    885
    Location:
    NorCal
    #16 EinSV, Sep 14, 2016
    Last edited: Sep 14, 2016
    I tend to be somewhat data-driven, but unfortunately we have limited data so far due to the relatively low number of miles driven on AP.

    What we do know is that in the first 200,000,000 miles on AP, there have been 1 or at most 2 fatalities using AP (depending on the outcome of the investigation in China).

    More significantly, from the information available, a reasonable conclusion is that there has not been a single person killed using pre-8.0 versions of Autopilot when used as instructed in 200,000,000 miles. This is a truly remarkable achievement for AP's first year "in the wild."

    With v 8.0 Autopilot, which includes additional features to help ensure it is used as intended, Elon has predicted that within about a year AP will triple the safety of driving compared to vehicles without Autopilot. I look forward to seeing the data to verify this, but based on the first year of experience with AP this seems like a reasonable guess.

    Edit: Just saw the posts above -- I wouldn't be surprised to see data on v 8.0 w/n the next year or so to back up Elon's safety predictions.
     
  17. RogerHScott

    RogerHScott Active Member

    Joined:
    Mar 2, 2016
    Messages:
    1,356
    Location:
    Lafayette, IN
    Perhaps, but the need to charge guarantees they take breaks and don't do the 12 hour marathon driving that some ICE drivers do.
     
  18. RogerHScott

    RogerHScott Active Member

    Joined:
    Mar 2, 2016
    Messages:
    1,356
    Location:
    Lafayette, IN
    It is only remarkable if we know with equal or greater certainty that 200M miles of equivalent non-AP driving would have had a higher
    death rate. Does anyone have data to that effect?
     
    • Like x 2
  19. green1

    green1 Active Member

    Joined:
    Mar 25, 2014
    Messages:
    4,116
    Location:
    Calgary, Alberta, Canada
    Speak for yourself.
    I am LESS distraced with AP than without, I can now focus my FULL attention on what's happening around me, and be in a much better position to make the necessary changes.

    Those "1 second at a time" glances at a smartphone? They are ZERO second glances. My phone stays in my pocket, not only would it be illegal for me to pull it out, it's also just plain stupid.

    AP is a driver assistance function, it is clearly communicated as such, and if used appropriately adds a LOT of safety.

    If however you're unsafe driving while using AP, you would be even MORE unsafe driving without it.

    Unfortunately, Tesla seems determined to reduce the safety of the system by introducing nags that force people to spend more time looking at the instrument cluster and less time looking at the road. They also already introduced limits to the speed it will operate at which can cause the car to unexpectedly slow large amounts in the middle of a freeway, and they've introduced nags at meaningless times which condition drivers to ignore the nags. All of this makes the system drastically less safe than it was when originally introduced. Very soon it will likely be both easier and safer to just drive manually.
     
    • Like x 2
  20. S4WRXTTCS

    S4WRXTTCS Active Member

    Joined:
    May 3, 2015
    Messages:
    1,180
    Location:
    Snohomish, WA
    It didn't really change my behavior all that much.

    With an ICE I did a marathon drive to SF that took me around 11 hours (I was speeding a bit), and did a marathon trip back.

    With my Tesla I did a leisurely trip to a bit south of Santa Cruz, and then did a marathon drive back. On the drive back when it started to get late I tried to nap for the hour my car was supercharging.

    I don't think it made me any more rested or if there was really any difference. In either case it was probably stupid.

    There are also a number of accidents where Tesla drivers have fallen asleep. Maybe its the combination of smoothness plus quietness. Or maybe they're just old.
     

Share This Page