Slashdot Mirror
'Operational Limitations' In Tesla Model S Played a 'Major Role' In Autopilot Crash, Says NTSB (reuters.com)
Mr D from 63 writes from a report via Reuters: The chairman of the U.S. National Transportation Safety Board (NTSB) said on Tuesday "operational limitations" in the Tesla Model S played a "major role" in a May 2016 crash that killed a driver using the vehicle's semi-autonomous "Autopilot" system. Reuters reported on Monday that the NTSB is expected to find that the system was a contributing factor because it allows drivers to avoid steering or watching the road for lengthy periods of time. The NTSB is also expected to find that Tesla Inc could have taken additional steps to prevent the system's misuse and will fault the driver for not paying attention. "Today's automation systems augment, rather than replace human drivers. Drivers must always be prepared to take the wheel or apply the brakes," NTSB Chairman Robert Sumalt said. The system could not reliably detect cross traffic and "did little to constrain the use of autopilot to roadways for which it was designed," the board said. Monitoring driver attention by measuring the driver's touching of the steering wheel "was a poor surrogate for monitored driving engagement." At a public hearing Tuesday on the crash involving Brown, NTSB said the truck driver and the Tesla driver "had at least 10 seconds to observe and respond to each other."
"""Monitoring driver attention by measuring the driver's touching of the steering wheel "was a poor surrogate for monitored driving engagement." """
How would you monitor their engagement? Eye tracking? Manual corrections to the car's path/speed?
What happens when people ignore the "please grab the wheel?" Does the car pull over and park? Is that what it should do?
--Welcome to the Realm of the Hawke--
As autonomous cars get better and better, we'll see more and more accidents attributed to driver inattention -- the better the car is at driving, the less the human is going to pay attention to the car or the road, and by the time the car tells the driver "Oh hey, I don't know how to handle this situation, you take over!", the driver won't have enough situational awareness to get out of the situation.
Though the flip side is that as the cars get better at driving, the overall accident rate will decrease.
The same problem already exists with airplane pilots , and it can be even worse where the autopilot compensates for some building condition (like icing), and by the time it gives up control to the pilot, the plane may already be in a bad state and the pilot has little time to figure out why.
As autonomous cars get better and better, we'll see more and more accidents attributed to driver inattention
Not if we can have a system which is better at driving than a human. In fact, other than the "cool factor" I'm not sure I see the point of a semi-autonomous system which requires me to watch the road all the time since it is no different from driving myself and potentially a lot more annoying. Frankly, it sounds more like paying to debug a final system which will drive itself.
They allowed the system to be active with seven warnings and over the speed limit.
The NTSB is also expected to find that Tesla Inc could have taken additional steps to prevent the system's misuse
Of course they could have taken additional steps to prevent the system's misuse before the crash because that's exactly what they did right after the crash.
Anons need not reply. Questions end with a question mark.
You can't have it both ways. Really, you can't. Either the car is driving or the person is driving. Expecting that a person will let the car drive AND ALSO stay 100% ready to take over is just not reality. If you are not the one in control, then your mind will not focus on it. Driving is boring enough as it is, expecting someone to babysit a semi-autonomous car is way beyond what we can expect people to do.
Just as an example, 9 years ago when I got my fully loaded Infiniti G37S with technology package, it was one of the first vehicles to have laser-controlled intelligent cruise control. It can match speeds of the cars in front and actively adjust, even brake if necessary. And just that ONE feature of driving assistance sounded like it would be very useful. OMG no. I tried many times to use it and found that just fully automated speed control was enough to disengage me from being an active driver. I could not adapt to it and ultimately decided I would never use it again. It was simply unsafe! Regular cruise control- no problem, I have to pay attention and I bump the speed up and down manually with the thumb control and take other action when necessary. But as soon as that was taken away from me, it became nearly impossible to stay attentive, even though I still had to steer!
Now, maybe different brains work differently and some people can handle semi-automation, but I know I can't. So don't even TRY to give me a car that can sorta drive itself and expect ME to be the ultimate failsafe... that just isn't going to happen. And I expect I am far, FAR from alone in this.
The issue is the duration in which the Tesla system gives warnings - the industry standard recommended timespan between inattentive-driver prompts is 15 seconds, while in this case, there were no inattentive-driver prompts from Autopilot for the two minutes leading up to the accident.
This is the issue when you call something "Autopilot" and give it to a consumer base that is used to being spoonfed fictitious understandings of such systems from superficial TV shows - they are led to believe it does something that it most certainly does not.
Yes, Tesla put all sorts of warnings in their manuals about this, but theres absolutely no requirement to read those manuals before jumping into the car, hitting the highway and engaging the system. Thats where the disconnect from reality and theory occurs - in theory, everyone reads the manual and understands the intimate details of the vehicle before setting off, while in practice people jump into new cars all the time and try things out.
Who here has been the person sat in their rental car for 30 minutes reading the manual before driving off for the first time? I bet the number of people who response affirmatively to that question is .... low.
Thats the issue Tesla need to solve.
It's not quite so easy to spot a flock of birds taking off as you reach the critical speed from which you can no longer abort a take-off. The pilot flying the airplane (captain or F/O) does have his or her eyes outside during the take-off roll until lift-off, as this is a completely manual and visual phase. The other pilot is looking at engine parameters, airspeed and a few other instruments and just taking a few occasional glaces outside.
But once you're above the abort speed (v1), there's little you can do about birds. You can't just yank the airplane left or right to avoid them when you're doing 150 knots. We are actually advised not to change our flight path because the birds act in unpredictable ways and violent maneuvers may actually increase the chance of a bird strike with birds dashing off precisely in the same direction you decided to turn towards. Also, it's easier to react to an engine failure if the plane's path is nice and stable rather than while you're violently banking or pitching.
What makes you think that makes it any easier? They were probably doing well over 200 knots by then so it's even harder to make a tight turn. For information, a 2G turn at 200 knots (passengers screaming in the back) has a radius of 630 meters.
Trust me, I fly the A320. You don't just yank such a plane around to avoid some birds. When you see them, it's usually already too late. Too much inertia, and you don't know which direction the birds are going to take anyway. During approach, at low speeds, you might consider going around (pulling the nose up and adding full power) but it's a risky decision because it might be preferable to continue the approach with a limp airplane rather than losing engines during a go-around if you didn't manage to miss the birds.
With most bird encounters, we end up missing them. Sometimes we hit one or more, but it has no effect on the airplane. Very rarely, an engine gets seriously damaged. Loss of all engines in an airliner happens once every few decades in the whole world. Now imagine we would "yank the plane around" every time we saw some birds. You would not want to fly on those planes.