Fortunately, we have something close to an apples-apples comparison on this. There was about a 1 year gap between when cars with AP1 started being shipped and when the system was actually turned on. NHTSA found that after the system was activated, accident rates dropped by 40% for the same cars. That’s the rate of accidents per mile driven, regardless of whether AP is even turned on(suggesting it’s an underestimate of the true rate drop) and before the nags were instituted.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
HWY101 accident..
- Thread starter _jmk
- Start date
Fortunately, we have something close to an apples-apples comparison on this. There was about a 1 year gap between when cars with AP1 started being shipped and when the system was actually turned on. NHTSA found that after the system was activated, accident rates dropped by 40% for the same cars. That’s the rate of accidents per mile driven, regardless of whether AP is even turned on(suggesting it’s an underestimate of the true rate drop) and before the nags were instituted.
I had statistics too in university, but apparently I remember that correlation isn’t the same as causation. So AP might have nothing to do with fatality rates at all.
Now that is a statistic that makes sense and sounds reasonable.
I don't care about how others use the system and will not let it affect how I use it. Seems like whenever there's one bad apple, everything gets spoiled.
This also applies to legislation for everything that exists in government. This will bring anxiety to the mass which leads to future stupidity and waste. You can't fix stupidity.
This also applies to legislation for everything that exists in government. This will bring anxiety to the mass which leads to future stupidity and waste. You can't fix stupidity.
fasteddie7
Active Member
The numbers aren’t skewed as media doesn’t do the math. Over 3,600 other people unfortunately passed due to auto accidents that day. Those others don’t get massive clicks and views. Tesla is an easy target. Autopilot saving a person from an accident doesn’t get reported on because it doesn’t generate viewership in the same capacity. Lots of other vehicles that advertise their system with active and passive safety features crash daily, and in a year we’ve got 4 or so high profile Tesla accidents against the 40,100 fatal accidents that gained next to no media attention minus some off cases of celebrity or one off “newsworthy” cases. The folks who knee-jerk react to stories like these hone in on a single case that justifies their techno panic and claim that the tech is killing us while not batting an eye to all of the human error, distracted, or drunk driving deaths that occurred thousands of times that same day. Autopilot in its current state is designed to be used as an assistive feature. Accidents can and will happen with any product, but we should learn from it and not chasitise the technology, but use the info to make the situation safer, from Tesla and all the way down to the fact that the section that was hit was not properly repaired to absorb another impact from an earlier accident.
Olle
Active Member
How come the Tesla 'fleet learning' doesn't prevent the cars from keeping doing this same mistake again? After such a horrible crash it should be strongly motivated to learn this kind of divider, no?
But that includes some crap car in southeast Asia transporting 8 people with no airbags, or even seatbelts.
Sure, that doesn't say Tesla AP is unsafe, but I don't think it really safes as many lives as some think here. 40% reduction in accidents is a pretty good statistic, though. Not sure if that also includes AEB, or just AP, but still not bad.
And you have a point that Tesla gets a lot of media attention, but I wouldn't say that's a bad thing. I mean as any random person what they think of Tesla and they will know the company, know the cars are electric, but probably also think the car can drive itself.
Ask them about Cadillac and they won't know about super cruise, which is a pretty equal AP system. So it has it's benefits that Tesla is known, but it also has it's downsides. Maybe Tesla could limit the downsides by educating the public more about AP, but I don't think that's necessary right now. The average Joe and Jane won't buy a Tesla for at least another year anyways and at some point their friends with a Tesla will educate them about the limits of AP.
Headlines like these are only a problem for the people shorting, or longing stock. In a month nobody will talk about this. And this won't lead to different legislation, Uber has done that already.
Olle
Active Member
One would hope so. It is probably learning algorithms to be able to handle every non predictable and not repeatable situation rather than learning location based pre existing activities. Might there be room to learn just a few locations though like these dividers?
There was a discussion quite some time (years) ago about this. Originally there were indications that fleet learning was in use. Then there was something to the effect that Tesla gets way too much data, and sifting through it was a huge challenge that they hadn't really tackled. I haven't dug through the forum to try and find the discussion - that would be something of an easter egg hunt.
From my own observation of the autopilot's behavior, I can see no indication whatsoever of any fleet learning. The local HOV lanes drive the autopilot nuts - at every transfer point the lanes widen and narrow again, and at each transition the autopilot jerks the wheel violently. It's never learned not to do this.
So I don't believe this "fleet learning" thing was ever actually implemented.
What neural network! Isn't a 3 legged dog still a dog! If Beta Pilot cannot recognize an obstruction in the road that is as wide as 2 small children, Eon Musk should stop touting it. I hope the NHTSA recalls this defective product.
jelloslug
Active Member
greghanssen
Member
Great discussion.. Was the driver 100% at fault for not watching the road? YES! Did the system fail and drive into a solid object? YES! Very similar to the Uber AZ accident in that, yes, the system should have totally avoided the accident but also had the driver been paying attention, he/she could have easily avoided the accident as well.
As for the hands on steering wheel nag, I'm working on a similar system and I can say it certainly takes a level of software to get the detection correct... If the wheel torque detect is not sensitive enough then you get warnings even when your hands are on the wheel but if you set the torque detect too sensitive than it can read 'phantom' hands on wheel when the AP system torques the wheel (when your hands are off the wheel) and the mass of the wheel itself creates a false positive hands on wheel signal. Need decent filtering to hands both cases.
As for the hands on steering wheel nag, I'm working on a similar system and I can say it certainly takes a level of software to get the detection correct... If the wheel torque detect is not sensitive enough then you get warnings even when your hands are on the wheel but if you set the torque detect too sensitive than it can read 'phantom' hands on wheel when the AP system torques the wheel (when your hands are off the wheel) and the mass of the wheel itself creates a false positive hands on wheel signal. Need decent filtering to hands both cases.
WannabeOwner
Well-Known Member
"Hold steering wheel"
That's flashing for just over 9 seconds, up until the point at which driver overrides and AP disengages (Bing-Bong just audible)
So the driver ignored the warning for 9 seconds AND was hand-holding the camera, so not properly in control in the first place, and failed to divert the car, which would have been easy with one hand on the wheel, so came to a halt in the median which seems pretty dangerous to me (if this was a test rather than "not paying attention").
Joe F
Disruption is hard.
It doesn't highlight any such thing, IMHO. The lane markings are certainly not the same: In your case (assuming you shot the video?) the "left" lane line for the lane you were in, vanished, and the Tesla did what it was designed to do, follow the lane line to the left as the "lane" widened. Obviously it was not the left lane line for the lane you were in...
It's not clear to me why this is being compared to the 101 accident. It's also not clear to me why, in this case, the car path was not manually corrected immediately and directed to the proper lane. Was this something done intentionally to (incorrectly) compare to the 101 accident? If this was not done intentionally, the question is, why were you not paying attention to the road?
Just so there is no fuzz in understanding by Tesla or anyone else, I have had numerous occasions when autopilot went into search mode in a lane and when changing lanes. This is particularly likely when using a diamond lane where the lane markings are obliterated or ratty on the asphalt shoulder of the left side. It then weaves as far as threatening to hit the divider wall and run over the botts dots on the other side. It requires wrestling back control of steering or canceling autopilot, so I am always on the alert. Traveling at 65+ mph leaves little time to react if caught off guard.
jelloslug
Active Member
It isn't an AP error. Probably parking sensor error or something not serious (since serious errors also appear on the IC).
bebbiXpress
Member
I'd like to add something more general to this discussion that applies also to driving without any driver assistance features.
It seems as nearly nobody gives a damn about posted speed limits anymore. Going 5mph over the speed limit is kinda common behaviour and socially accepted. And in this case you can clearly see a 45mph limit posted well before that fatal divider! This tells me, the authority who did put that sign up thinks "that's a dangerous section"!
But in the now posted video the AP was set to 59mph and neither did the other cars slow down to 45mph. Also from what Tesla stated about the accident (5 seconds and 150 meters of unobstructed view on divider) you can calculate that the crashed car was going around 67mph! That's more than 20mph over the speed limit.
What does that mean? Well, simple physics: 1st, the kinetic energy would have been half if he went 45mph. Just for comparison, hitting an immovable concrete divider with 45 is like hitting ground after falling from the 7th floor. Hitting it with 67mph is like falling from the 15th floor. 2nd, with 45mph there would have been additional 2.5 seconds time for the driver, EAP or AEB to maybe recognise the divider and react.
So, in the end it's your life and the lives of others you risk just to gain a few seconds on your way!
It seems as nearly nobody gives a damn about posted speed limits anymore. Going 5mph over the speed limit is kinda common behaviour and socially accepted. And in this case you can clearly see a 45mph limit posted well before that fatal divider! This tells me, the authority who did put that sign up thinks "that's a dangerous section"!
But in the now posted video the AP was set to 59mph and neither did the other cars slow down to 45mph. Also from what Tesla stated about the accident (5 seconds and 150 meters of unobstructed view on divider) you can calculate that the crashed car was going around 67mph! That's more than 20mph over the speed limit.
What does that mean? Well, simple physics: 1st, the kinetic energy would have been half if he went 45mph. Just for comparison, hitting an immovable concrete divider with 45 is like hitting ground after falling from the 7th floor. Hitting it with 67mph is like falling from the 15th floor. 2nd, with 45mph there would have been additional 2.5 seconds time for the driver, EAP or AEB to maybe recognise the divider and react.
So, in the end it's your life and the lives of others you risk just to gain a few seconds on your way!
Similar threads
- Article
- Article
