OP is famous, this just happened 
Tesla driver mows down 11 construction cones on video and blames Autopilot - Electrek
Tesla driver mows down 11 construction cones on video and blames Autopilot - Electrek
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Automatic Emergency Braking Failure, the Movie
- Thread starter Wingsy
- Start date
He actually said in the OP that he wanted it to make an alert (I think he said "a beep") and slow down. But you do make a good point that if he had something else, it would have been different.
And I wonder why my faith in humanity continues to plummet?
"Blames"? He quite clearly states that he blames himself.OP is famous, this just happened
Tesla driver mows down 11 construction cones on video and blames Autopilot - Electrek
Words have meanings. WTF "journalism" is that?
Faith, humanity, etc.
Last edited:
This is definitely an unfounded expectation.
Construction zones were never officially cleared for AP use.
My complaint about them though is that sometime after the Tesla FSD investor event, someone posted a tweet (with video) about their car making it fine through a construction zone and Elon 'liked' that tweet. That sort of spooked me, like, did they change something and made it better? The video was, I believe, a straight pass between 2 rows of cones, but I think the car stayed within its lane, so it probably didn't even account for them at all.
I then went on a 2,500 miles trip myself and was going through some construction zones on AP and observing what would happen.
For several, similar to what was posted on twitter, it was fine. For one of them, where cones appeared in my lane on the side and then moved into the middle guiding me into the next lane, with a wooden barrier across my lane some distance ahead, AP did not work. It was driving straight into the barrier without reducing the speed.
I sort of expected that, but nevertheless was mesmerized by the view of approaching barrier and encroaching cones. Had to take over at the last moment. I know for sure now that AP would not handle it.
Until the official statements that AP/FSD is smart enough to handle construction zones, your best bet is to turn AP off when you see cones.
I always turn off NoA when approaching a construction zone, but for the majority of the ones I've encountered AP itself did just fine. It's all about the zone and how it is managed or marked. The ones that are properly marked it does fine in. The ones where they are using cones or barrels to override still present lane markings and the ones with no markings at all, not so much.
AlanSubie4Life
Efficiency Obsessed Member
I hope you don’t count me amongst them! I think it is pretty clear what the limitations are of the current system; it cannot be relied upon in any circumstance.
Guess you guys with FSD will at some point see what happens with the new HW3 and whatever new code they may have! I definitely would expect level 2 (similar effective reliability to what we have now but with many more features) for quite a while though! I would expect the newer stuff to work better, but whether it will be good enough anytime soon is the question!
Daniel in SD
(supervised)
All the talk about what AP can and can't do reliably is silly. First, it would take millions of miles of driving to determine what it can handle reliably and no individual has driven that much. Second, even if you know what it can handle reliably you still have to be paying attention to see what the current situation is!
"One of the common misimpressions is that when there is, say, a serious accident on Autopilot, people – or some of the articles – for some reason think that it’s because the driver thought the car was fully autonomous and it wasn’t, and we somehow misled them into thinking it was fully autonomous. It is the opposite.
When there is a serious accident, it’s in fact almost always, maybe always, the case that it is an experienced user and the issue is more one of complacency. They get too used to it. That tends to be more of an issue. It is not a lack of understanding of what Autopilot can do. It’s actually thinking they know more about Autopilot than they do, like quite a significant understanding of it."
Elon Musk
"One of the common misimpressions is that when there is, say, a serious accident on Autopilot, people – or some of the articles – for some reason think that it’s because the driver thought the car was fully autonomous and it wasn’t, and we somehow misled them into thinking it was fully autonomous. It is the opposite.
When there is a serious accident, it’s in fact almost always, maybe always, the case that it is an experienced user and the issue is more one of complacency. They get too used to it. That tends to be more of an issue. It is not a lack of understanding of what Autopilot can do. It’s actually thinking they know more about Autopilot than they do, like quite a significant understanding of it."
Elon Musk
I think your confusing AP/TACC functionality, and AEB functionality.
AEB is a very limited system that is only designed to slow the car by 25-30mph in the event of an unavoidable accident. It's extremely tuned to not having any false positives. That makes sense as you would never want the car to suddenly drop it's speed by such a large amount as quickly as possible. It basically slams the brakes until that speed threshold is met. It's that way for a lot of AEB system, and can only be considered an accident mitigation system. I don't believe the system is at all meant for things like cones.
AP/TACC on the other hand can bring a car to a stop. It tends to have a lot of false positives (depending on SW version), but the speed drop is a lot more gradual.
There have been plenty of cases where it's shown to detect cones so I'm not sure why it didn't detect the cones in your case. My expectation is that it would try to get over, but maybe it couldn't due to a truck in the way. I'm also not sure why it didn't slow down.
If AP was in fact engaged then it obviously failed. Now that failure is expected as it clearly says it's not designed for construction areas. I think it's something they're still working on.
As to your failure as the person responsible for oversight. There isn't much to comment about. I was in similar situation where the construction company lied about how many lanes were going to be closed. At the last second I was able to get into the right lane.
Sometimes humans just glitch.
Last edited:
AlanSubie4Life
Efficiency Obsessed Member
Can you point to the reference for this? I wasn't aware of actually any detection capability of such things. I've heard it can alert of construction areas, but it's not clear to me that would be due to anything the car sees (it could just as easily be Waze-like crowd-sourced data for all I know).
I'd like to see a bit more firm evidence that it has any awareness of such objects. It probably exists; I just have not seen it.
Not that it matters - you can't rely on it anyway - but just out of curiosity.
I can't say any of these are impressive.
But, I think they show there is at least some attempt at handling them.
Corrected your post.OP is infamous, this just happened
Tesla driver mows down 11 construction cones on video and blames Autopilot - Electrek
AlanSubie4Life
Efficiency Obsessed Member
I agree it seems to attempt to handle these situations, with a great deal of uncertainty about how to proceed, at very low speeds. Definitely would not count on it!
Baby neural net steps and a GREAT crawl before you walk start. RUNNING down the Interstate at 75MPH is way too much to expect now. Also remember you as a human may do the same thing EXCEPT you would have read all the WARNING signs that the lane was closing and understood what other cars were moving over for before you got to the barrels. It is FAR too early to expect AP to read the signs and understand what was going to happen.
Maybe it shouldn’t be a commercial product right now, if the neural net is still a baby.
Also, if Tesla sells Full Self Driving today and the CEO promises Robotaxis soon, why would the general public assume their shiny new Tesla with the $6000 FSD option isn’t able to see big traffic cones?
Also, if Tesla sells Full Self Driving today and the CEO promises Robotaxis soon, why would the general public assume their shiny new Tesla with the $6000 FSD option isn’t able to see big traffic cones?
Thankfully, I haven’t experienced this same exact scenario. But, I don’t get it. Even my old 2016 Honda was able to stop for relatively large items such as these cones when straight ahead. Even with the CR reports, Tesla’s system continuously seems to have issues with stationary objects. We’ve also seen videos where the Tesla supposedly swerved by itself to avoid a crash when braking would not be sufficient. But, in this particular instance, it does not even decide to slow down and get behind the truck on the right. In fact, it didn’t look like it ever once occurred to slow down.
And before someone exclaims emergency braking could have caused a rear-end, which we don’t know from this video whether a car was behind it, that would have been the rear driver’s fault for not putting enough space in front to avoid a potential crash on the freeway.
Definitely not ready for prime time AP/FSD/ robotaxi for the general public.
And before someone exclaims emergency braking could have caused a rear-end, which we don’t know from this video whether a car was behind it, that would have been the rear driver’s fault for not putting enough space in front to avoid a potential crash on the freeway.
Definitely not ready for prime time AP/FSD/ robotaxi for the general public.
Daniel in SD
(supervised)
How do you know that the car wouldn't have avoided the cones 99 out of 100 times?
That's not until next year.
Tesla collects millions of miles of data monthly. That's why they will be able to verify FSD sooner than the others. What AP/EAP do in the public's cars today is not indicative of what company test cars are now doing and what they will do in the future. Why is this so hard to understand? My car steered slowly through a windy one lane construction zone with barrels on both sides on my delivery drive home with 40 miles on the odometer in July 2018. More recently it changed lanes with cones. It's merging and NoA lane changing is getting better all the time. It's starting earlier and getting smoother on deceleration behind cars stopped at intersections. The drumbeat of those on this forum that say FSD is never going to happen is suspect.
Reacting to stationary object at high speed is a well known hard problem and if 2016 Honda has this problem solved, I would be really impressed. Can you provide actual evidence that it can stop for a cone in the middle of the road at 75mph?
Similar threads
