Perhaps better to be proactive than to wait for the NTSB to impose real restrictions.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Phone Distracted Driver on Autopilot Slams into Firetruck
- Thread starter cwerdna
- Start date
Perhaps better to be proactive than to wait for the NTSB to impose real restrictions.
Tesla developers wanted more safeguards on Autopilot system to make driver pay attention
Here's another example of a trade-off. When will companies start realizing that the best thing to do for their customers is also the best thing for shareholders and no other way!
Here's another example of a trade-off. When will companies start realizing that the best thing to do for their customers is also the best thing for shareholders and no other way!
The new Leaf avoided a crash to a standstill dummy from 40 km/h and with a slow rolling dummy from 70 km/h. I think it is time Tesla shows us a 2018 EuroNCAP test for the S and X with AP2 and 2.5 suites of hw.
thegruf
Active Member
^ that's really impressive for the Leaf (Waymo take note!)
I call well done Nissan.
More importantly though is to see major safety institutions starting to independently test systems against known criteria.
It will also very quickly separate hype from reality.
My only concern is that automotive manufacturers will learn to cheat the tests (who'd have thought it?) so the tests must involve randomization to try to preclude this.
Hopefully these tests will continue to evolve soon to include Autopilot type systems against real world criteria.
This can only be beneficial for consumers and road safety.
I call well done Nissan.
More importantly though is to see major safety institutions starting to independently test systems against known criteria.
It will also very quickly separate hype from reality.
My only concern is that automotive manufacturers will learn to cheat the tests (who'd have thought it?) so the tests must involve randomization to try to preclude this.
Hopefully these tests will continue to evolve soon to include Autopilot type systems against real world criteria.
This can only be beneficial for consumers and road safety.
Ciaopec
Active Member
Couldn't agree more! Unfortunately, we cannot fix stupid. Until people read, understand, and practice we will continue with these silly (and in some cases, tragic) accidents. I've ranted before: I believe more than 90% of drivers cannot effectively use cruise control let alone wrap their heads around the wonder that is a Tesla. It is an extra set of eyes but it is still not autonomous. Sadly, Tesla is always in the spotlight for real (and short sellers perceived) reasons. Ignorance does not help our cause.
Kant.Ing
Member
Stupid is as stupid does. This includes parents or speed racers trusting the touted crumple zone.
Those are pretty low speeds... Even the rolling test is slower than Tesla claims for fully stopped vehicles(that'd be roughly 80.5km/h). The crash discussed here occurred with a fully stopped vehicle while the Model S was traveling about 96.5km/h, far and away beyond the Leaf test there.
The latest news say that she was using AP, and the reason she didn’t stop was she was looking at her phone.
To me it’s pretty evident that Tesla owners are highly uneducated on how the system works, and Tesla needs to have some kind of training video a customer has to watch.
This isn’t the first case either. We’ve had a bunch of cases on the TMC forums where people have admitted to crashing into a stopped car while being distracted.
They usually come back with “I didn’t know” followed by someone sending them the link to the wired article that goes in depth on why it can’t always detect a stopped car.
The Latest: Driver: Tesla's Autopilot engaged during crash
To me it’s pretty evident that Tesla owners are highly uneducated on how the system works, and Tesla needs to have some kind of training video a customer has to watch.
This isn’t the first case either. We’ve had a bunch of cases on the TMC forums where people have admitted to crashing into a stopped car while being distracted.
They usually come back with “I didn’t know” followed by someone sending them the link to the wired article that goes in depth on why it can’t always detect a stopped car.
The Latest: Driver: Tesla's Autopilot engaged during crash
Last edited:
If 90% of drivers can't effectively use the technology, then maybe Tesla shouldn't release it in "beta" and without extensive instructions. Car features need to be both easy to use and to work in a reliable and easy to understand manner. Otherwise they just make the car a hazard to others on the road.
I have no idea why they even call AP1 as beta.
It’s not really going to get any better, and it certainly doesn’t have the sensors to detect a stopped car.
It wouldn’t matter if the driver was using Adaptive cruise control or AP. In either case it’s not going to see the fire truck.
AP as with any L2 system allows more distraction, but these types of accidents are also happening with just adaptive cruise control.
I’m in total agreement with you that features should be easy to use, and understand.
But, that might be more challenging than either of us realize. Even when they know better people do stupid things.
Ciaopec
Active Member
What could be easier to use and understand than something simple like cruise control? Oh wait, how about the old analog temperature controls. I can't tell you how many van drivers, when asked to cool it down a bit, dialed the know from full hot to full cold not realizing there was a wide range of temperatures in between. How many people actually take the time to read an owner's manual (while charging, for instance)? (I'd venture most of the pilots on the forum have read through it).
One doesn't need the depth of knowledge or understanding that wk057 has but you need a better understanding of AP, and situational awareness, and AEB, etc. than the average let's-buy-a-Tesla-because-it's-the-cool-new thing person seems to be willing to master. IMHO
Soapbox is now available for others to use. Thanks for your time.
hacer
Active Member
Actually if the AP camera detects the vehicle in the lane, then it will absolutely stop even when the radar did not ever see it. The trouble is that the camera is not so hot at detecting vehicles in the lane, especially at sufficient stopping distances at high speed. It seems likely that it is especially bad at recognizing fire trucks. In my experience it's not too bad at detecting cars with lit brake lights, I'd guestimate that it gets them 80% of the time in my AP1 car.
But it is sufficiently not-good at recognizing stationary cars that nobody should expect to rely on it, ever.
Right... But with TACC the Tesla driver is likely to be paying more attention, since she still would have to steer. AS makes it much easier for a people to be more inattentive. Plus, I suspect that many drivers assume that a stopped truck would be an easy thing for AP to see, not one of the harder things for it to react to.
This is why at Model 3 Gripes 2 Months later I suggested a MANDATORY set of audio recordings that MUST be listened to at above a certain volume level w/instructions, caveats, limitations, etc. of TACC, AEB, AP, etc. before the driver even can use AP.
Perhaps it should also be tied to each key fob or phone. Someone gets in w/a phone (w/Tesla app tied to the car) or fob (also tied to that car) that hasn't heard it yet, they need to listen to it, as well. That way (as a hypothetical example), at least 2 family members are likely to have heard the instructions. One listener can tell the others to listen before using AP.
I'm not done listening to the latest TWiT podcast but the host (Leo Laporte) has had a Model X w/AP for a little while and while they're talking about this accident at ~20:54 of
Can you imagine that a reasonably competent tech journalist who knows more than the average Joe says this, can you imagine the average (not an active participant on TMC nor any other Tesla forum) Tesla driver?If you go further back, to ~19:00, they show one of the stories on this accident and Leo shortly afterwards says "it should stop" and the Navigant says no and talks about the stationary object limitation.
Last edited:
Canuck
Well-Known Member
The text is full of ambiguity. They call it Beta but it is better than beta?
"To be clear, when we say "beta", we do so to encourage a higher level of vigilance. If this were PC desktop or mobile software, we would not refer to it as such."
And all the crap about fleet learning.
They should really only have one line in that disclaimer.
Autopilot and TACC is experimental at this point, and pay close attention (more than when driving yourself) while using as the system may take actions an attentive driver usually will not.
The Tesla phone app needs to have some sort of priority override where it disables the phone and just outputs the on-screen message:
"Shouldn't you be driving right now instead of trying to use your frickin phone?"
Similar threads
