Tam
Well-Known Member
Yes. I drove 2 different Tesla cars. They both are with FSD and 1 with FSD beta. They both do poorly in heavy rain and sun glares during sunrises and sunsets. It's a design issue, not a hardware/software issue.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Tesla Radar Speculation
- Thread starter Takaa
- Start date
Yes. I drove 2 different Tesla cars. They both are with FSD and 1 with FSD beta. They both do poorly in heavy rain and sun glares during sunrises and sunsets. It's a design issue, not a hardware/software issue.
Just to be clear, phantom braking and any kind of AEB activation are two completely different things and there are clear differences in alerts and actions of each.
That does not answer the question I asked. Let me re-ask.
You stated “My Tesla cars are very sensitive to heavy rain. They just suddenly give up with the red steering wheel icon and loud siren for me to take over. Manual driving only for my Tesla cars in heavy rains”
Is this behavior from the first day of ownership or after a certain mileage, days, months of usage or a software update, or changes to your car such as new wipers, use of rainx wiper fluid, etc
beatle
Active Member
Though I often bash HW3 for behaving poorly, it does a pretty good job with rain, even heavy rain. It was raining pretty heavily one night (not torrential, but a good steady rain) and the lane markers were not easy to see. I expected the car to give up and tell me to take over, but it did not, and it did a pretty good job of determining where the real lanes were, and it did not force wipers on high. In fact, I was able to drive with no wipers since I had recently applied Rain-X. Note that I only have EAP, not FSD. Perhaps FSD is worse.
Knightshade
Well-Known Member
FWIW I've only had basic AP give up in CRAZY heavy rain- like "probably human shouldn't be driving in it" heavy.
I've had NoA give up in anything beyond light rain lots of times (and thus drop back to basic AP)
FSDb has been somewhat in between those two-- usually needing moderate to err...lightly heavy? rain before it tells you it's limited or unable to work.
I've had NoA give up in anything beyond light rain lots of times (and thus drop back to basic AP)
FSDb has been somewhat in between those two-- usually needing moderate to err...lightly heavy? rain before it tells you it's limited or unable to work.
There is some evidence (anecdotal at present, and Tesla has never commented) that the degree is braking is modulated based on the presence and closeness of any car behind the Tesla. If so, this perhaps explains why there are so many "this is potentially dangerous" posts here but there are far fewer actual "I got read-ended because of phantom braking" incidents (that have been posted here).
Absolutely, FSD is trying to figure out car movements around it, intersections, traffic signals etc. AP is only looking for lane lanes and adjacent car locations. So I would expect FSD to be more fussy.
Phantom Braking happens when there are no references around to make a good determination that AP/FSD is working fine. So it brakes. Better to brake than continue blindly.
That is the reason why even Mercedes has certified L3 ONLY in high density traffic areas.
Tam
Well-Known Member
What I mean is: The hardware and software work as designed, not because there's a defect in the hardware or a bug in the software.
The radar can detect the speed of a moving obstacle accurately and it can also detect the speed of a stationary obstacle accurately.
There's no defect in the radar so how come the car slams into the parking firetruck?
That's the software part. The programmers intentionally write the codes to make sure the radar ignore stationary obstacles. It's a software design. The design is specifically written like that! There's no bug in the software.
Yes, there's a collision but it works as designed.
No matter how much I clean or recalibrate the radar or debug the software, the result is the same: collisions with stationary firetrucks.
Then why did my car slam on the brakes when I'd approach an overpass or large freeway sign? If radar was accurate, the car would know they weren't in my travel lane. If the software was coded to ignore stationary objects, it wouldn't have panicked.
Bad radar? Bad software? Both?
I would like to know how you came to this conclusion. I drive many roads with no other cars in sight without PB events. And, I have had plenty of PB events with other cars in view. However, if we assume that PB happens due to a lack of references, how did you conclude that the car decides that it is blind or chooses to apply the brakes? After all, after using the accelerator for a few seconds, the car generally is happy to continue on, despite nothing fundamentally changing with regard to the terrain or other traffic. so, wouldn't the car just keep braking?
It seems to me that there are always changes to the image so that the car should not conclude that it is not getting fresh images. One would need to be driving through pretty bleak territory to not see changes in the scene.
Tam
Well-Known Member
New 4d imaging radar can differentiate a dangerous stationary obstacle in front within the lane and the benign stationary bridge above the lane.
It's the function of money: Pay cheap radar and get the results of stationary obstacle collisions and phantom brakes. Pay more for 4d imaging radar to get rid of those 2 problems.
It's by design.
So I should have splurged for premium quality radar when I bought my car. I knew I was making a mistake going with the base package.
I think it's a stretch to say the software was "designed" to collide with firetrucks, rather is was not designed to avoid them. That is a software bug by omission, which as as much a bug as any other bug type. So yes, it is a software bug. Somewhere in Tesla there is (or should be) a list of requirements for FSD, and one of them is almost certainly "The car shall not crash nor cause other cars to crash" (this has been stated several times at developer days). Clearly, when the car crashed, it was not meeting the "not crash" requirement.
That is my hypothesis based upon my experiences with the 2 M3LR I drive. The only times I have had PB were as you described- “pretty bleak territory”
If that was not enough, Mercedes delivered L3 with a caveat- only usable in densely populated areas. One would think they would certify it for areas with no population. So why not?
Similar threads
