It was only a matter of time. I’m anxiously waiting for a new CR report of 3/Ys that touches on AP again in same way (on a newer vision only car). There’s definitely a chance they will encounter something but who knows. Maybe they’ll get lucky.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Phantom Braking
- Thread starter 2101Guy
- Start date
It was only a matter of time. I’m anxiously waiting for a new CR report of 3/Ys that touches on AP again in same way (on a newer vision only car). There’s definitely a chance they will encounter something but who knows. Maybe they’ll get lucky.
This is very true, and why I welcome the investigation. No deaths and no accidents.
As was discussed earlier, this is all about a balance between true positives and true negatives and where you draw the line (it is actually not a simple line). If I thought that this problem was insurmountable and a deal killer I would cancel my order.
We need data and transparency about the issue.
Lots of ranting, screaming, and some good civil conversation here. However, nothing is getting resolved.
Instead of 23 pages of arguments, why don't we attempt to solve the problem? Let's put together a testing plan, methodology, and track results in a shared workbook. Work on analyzing camera data, correlating accelerometry data from onboard and external sensors, and investigate the neural net algorithms for how they interpret the sensor data into their decision trees. Then we can find the problem and either fix it ourselves or report it to Tesla.
...or we can just keep bitching for 23 more pages...
And if you've read this far and are thinking to yourself, "We can't do most of that, as we don't have access to internal sensor data, firmware code, or the neural nets and decision trees", I'll agree with you 100%. I trust I've made my point that there's not much we can do as end users. We can report the problem to Tesla. We can report the problem to the NHTSA. And we can report the problem to other users on forums like this. That's it. That's all we can do. We can't solve it - only Tesla can solve it.
So settle down people.
Instead of 23 pages of arguments, why don't we attempt to solve the problem? Let's put together a testing plan, methodology, and track results in a shared workbook. Work on analyzing camera data, correlating accelerometry data from onboard and external sensors, and investigate the neural net algorithms for how they interpret the sensor data into their decision trees. Then we can find the problem and either fix it ourselves or report it to Tesla.
...or we can just keep bitching for 23 more pages...
And if you've read this far and are thinking to yourself, "We can't do most of that, as we don't have access to internal sensor data, firmware code, or the neural nets and decision trees", I'll agree with you 100%. I trust I've made my point that there's not much we can do as end users. We can report the problem to Tesla. We can report the problem to the NHTSA. And we can report the problem to other users on forums like this. That's it. That's all we can do. We can't solve it - only Tesla can solve it.
So settle down people.
EVNow
Well-Known Member
Or is it possible the industry that was wrong about panel gaps or EV profitability is also wrong about lidars ?
We can all keep wondering. Good thing is these things are settled on the ground, not on online forums.
if you're tired of being proven wrong, with sources, maybe you ought try being wrong less often?
Protip: Actually reading the entirety of posts to which you are replying would go a long way here. I know you've repeatedly discounted this idea as too much work, but give it a shot!
BTW, fun link for the couple folks who were touting how great Subarus safety systems are:
10 Car Models With The Most Accidents - Subaru Now Has Three Of The Top 5
An updated report from Insurify reveals the 10 cars with the most accidents. The Subaru Crosstrek, Subaru WRX, and Subaru WRX STI, and now Impreza are in the top 5 models.www.torquenews.com
Apparently 3 out of the top 5 models in at-fault accidents are now Subarus.
Toyota, Honda, and Mazda each make an appearance in the top 10 as well.
Tesla of course does not.
I'm a little confused about this.
The article claims 3 of the top 5 models for at-fault accidents are Subaru's. That sounds reasonable as the WRX models are popular among young people and people who drive recklessly.
But, when you look at the cited sources you realize that isn't necessarily true.
The data actually just says the percentage of owners with a prior at-fault accident. So the main article should simply read "The car models with the highest share of accident-prone drivers".
I find this kinda of funny because I recommended a Subaru Cross Trek to a girl who rear ended someone. I mentioned how Eyesight likely would have helped prevent that. But, at the time Eyesight wasn't a standard feature and was only available on higher end models which really annoyed me.
SalisburySam
Active Member
EVNow
Well-Known Member
Here is a comment on Reddit. I wonder if this is why I rarely ever see phantom braking and some see it all the time ? I’ve the following distance set to max and FCW to Early.
A friend of mine has 84000 miles in, mostly Autopilot with 5 months of FSD Beta. He says with the following distance set to max (7) and Forward Collision Warning to Early, he's never had phantom braking.
Knightshade
Well-Known Member
FWIW there was a thread a while back (at least a year or two now I think?) where someone did testing and found one of the settings around either FCW or AEB definitely impacted this-- don't think they tested follow distance though.
The problem with this is it won't take into account phantom braking due to maps/navigation.
What I noticed when moving to Tesla Vision was that TACC adjusted its set speed according to speed limit changes., and it didn't used to do that. This is both helpful from a drivers aid perspective, but introduces additional PB braking possibilities due to incorrect maps or incorrect reading of a sign.
EVNow
Well-Known Member
That could certainly be one of the issues.
But - is there an issue with people setting very low follow distances and Vision not handling it as well as Radar did ?
Knightshade
Well-Known Member
There's nothing "phantom" about that though- the system has a reason- the speed limit dropped. Even if the data it has on the limit is wrong, the behavior of the driving system is correct.
EVNow
Well-Known Member
Thats the analysis part.
But people would still consider it phantom braking because they can't figure out why it happened and that would lead to complaining in forums or to NHTSA.
Here's an interesting one. If you play it back slowly, there is a curve in the road, and a yellow/black warning sign at the leading edge of the guard rail for the bridge - the curve is such that the car is heading straight for that yellow/black sign:
sleepydoc
Well-Known Member
Exactly. If the algorithm is acting ‘correctly’ according to the program but still doing the incorrect thing then it’s still a faulty algorithm.
It’s kind of like saying ‘the algorithm is designed to brake if it detects a car within 500 feet and a car came within 500 feet so it slowed down’. The car followed the algorithm perfectly, but slowing down 500 feet from a car is totally inappropriate so the end result is still a fail.
CyberGus
Not Just a Member
Clearly NHTSA is omitting the many drivers injured after a phantom-braking incident by their angry wives
Knightshade
Well-Known Member
No, it's not.
It's faulty input.
You want the system to drive the correct speed limit.
That's what it's doing.
The driving system is working perfectly in this case.
The speed limit database has an error. Fixing that does not involve touching the driving algorithm at all.
EVNow
Well-Known Member
I'd classify that as faulty map.
As I've been saying they need a way for us to quickly and reliably be able to make changes to the map. That will solve such issues.
EVNow
Well-Known Member
I had something similar - but after many tests I realized FSD was slowing down because of the curve. So, what I thought was a phantom braking event turned out to be not really that ....
CyberGus
Not Just a Member
I got 99 problems but braking ain’t one
Last edited by a moderator:
Similar threads
