Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Cruise control fooled by shadow

This site may earn commission on affiliate links.
I was driving with cruise control on at 7 car lengths. I went under an underpass with a deep shadow....and the car slammed on the brakes. I think it thought the shadow was a car. The car behind me almost hit me. This makes me very reticent to use cruise control!
 
"...has slowly improved..."???? Just how will your insurance company handle this when you get rear ended?

Having BETA systems on public roads is high risk. I'm totally against it. Unless ALL cars have these "safety" systems, combined with a road system where ALL roads have imbedded sensors/tech, we're at risk.

I've got my Nomex suit handy for the flaming! :eek: :) :) :)
 
I wasn’t traveling much in 2019 or 2020 with my 2017 MS. Seems like it got worse when ai started again this year. I though it got better after a bit then on last Friday it did it again. Strange when it happens. Have to always pay attention. I still prefer using AP for long drives.
 
"...has slowly improved..."???? Just how will your insurance company handle this when you get rear ended?

Having BETA systems on public roads is high risk. I'm totally against it. Unless ALL cars have these "safety" systems, combined with a road system where ALL roads have imbedded sensors/tech, we're at risk.

I've got my Nomex suit handy for the flaming! :eek: :) :) :)

I'll be over this weekend to deactivate Autopilot for you. :D
 
I was driving with cruise control on at 7 car lengths. I went under an underpass with a deep shadow....and the car slammed on the brakes. I think it thought the shadow was a car. The car behind me almost hit me. This makes me very reticent to use cruise control!
Today, TACC uses the forward facing radar to set/adjust distance. So I doubt the shadow was the problem. More likely the radar took a hit from some other fixed object or perhaps traffic in an adjacent lane, especially if you were also on a curve. Soon Tesla is changing to “Tesla Vision” system that uses the three front cameras for following distance, and disabling the radar completely. We’ll have to wait and see if that is an improvement.
 
Today, TACC uses the forward facing radar to set/adjust distance. So I doubt the shadow was the problem. More likely the radar took a hit from some other fixed object or perhaps traffic in an adjacent lane, especially if you were also on a curve. Soon Tesla is changing to “Tesla Vision” system that uses the three front cameras for following distance, and disabling the radar completely. We’ll have to wait and see if that is an improvement.
There are numerous reports about shadows causing the problem, still anecdotal evidence but believable IMHO. I've had this happen sometimes on the same overpass, only when there is a shadow, and without any other traffic anywhere near me.

BTW: disabling the radar completely seems like the worst thing to do, as it is very accurate in sensing the distance to the car in front of you (so can respond quickly if that car is slowing down). Doing the same with just (stereo) vision processing feels like a very bad idea to me. Though I am sure there are scenarios where it's better to ignore the radar as its resolution is rather low, e.g. in city driving with parked cars on the side of the road.
 
BTW: disabling the radar completely seems like the worst thing to do, as it is very accurate in sensing the distance to the car in front of you (so can respond quickly if that car is slowing down). Doing the same with just (stereo) vision processing feels like a very bad idea to me. Though I am sure there are scenarios where it's better to ignore the radar as its resolution is rather low, e.g. in city driving with parked cars on the side of the road.

I am slightly skeptical as well. But I think we will find out soon enough. I am expecting Tesla Vision to be deployed with the "FSD on City Streets" feature we are all waiting for, I think Tesla Vision is required for that to work. I could easily believe radar would be overwhelmed with ambiguous information with dozens or hundreds of objects moving in all different directions. Need to use cameras to sort it all out. It is much easier to do on Interstate highways where everyone is (generally) going in the same direction and at roughly similar speeds. Speed choice based on a single object directly ahead works pretty well. Definitely not the case on local roads.
 
Last edited:
I haven't seen any self-driving ranking recently, but Tesla has always been ranked at or near the bottom.
I haven't seen the rankings you are referencing, but that doesn't seem to be an apples to apples comparison at all. Sure, if you can toss in a bunch of expensive sensor (360 Lidar, etc) and computing hardware, and then augment with precision 3D maps, you will almost certainly be better positioned than what Tesla is trying to achieve. On the flip side, Tesla is the only one that actually has sensor + computing hardware on all their vehicles to take a decent crack at self-driving today and have the potential to have a far more generalizable solution if they get it right.

I don't say that to defend their state of FSD today. I would never spend even a fraction of the $10K for what it offers right now. But they have an excellent technical team being led by Andrej Karpathy and I don't disagree entirely with their general philosophy of trying to solve self-driving in the general sense, primarily with vision, without relying on crutches like precision 3D maps which severely restrict the general utility of an FSD approach that relies on such maps. I do think that they should augment with other sensors like cheaper Lidar sensors down the line, but they are currently beholden to their decisions on sensor hardware from 10 years ago and won't survive the customer uproar if they have to announce the need for new sensors to get better FSD without a path to freely upgrade existing cars in the fleet.
 
I believe it's going to be an infinitely long time before Tesla manages L5 or even L4 with their current systems. It's not necessarily a matter of processing power. The software is just too damn hard to write. It's like Intel pouring all that money into the EPIC architecture. Some dogs just don't hunt. The many-times-delayed FSD is an absolute joke in beta, if you like jokes that can kill. That's not L5 or L4. That's L3 in a cape. Call me when I can strap a kid in a passenger seat and wave goodbye as the car pulls away on its own.
 
Thread title: "Cruise Control Fooled by Shadow"

Should read, "Tesla Owners Fooled by Unsafe Cruise Control/FSD"

Sorry, I really am, but this ongoing discussion is going nowhere, and is NOT instilling confidence in the general public.
 
  • Like
Reactions: Jigglypuff
I am slightly skeptical as well. But I think we will find out soon enough. I am expecting Tesla Vision to be deployed with the "FSD on City Streets" feature we are all waiting for, I think Tesla Vision is required for that to work.
It sounds like Tesla Vision gets deployed, and radar gets disabled, for the general US Tesla drivers starting with a release next week:


Then it will be a few more weeks before FSD beat v9 comes out using Tesla Vision.