Forums

Forward Collision Warning False Positives

Forward Collision Warning False Positives

The forward collision warning system seems to throw a lot of false positives. By the way, I think it is worth it for the 3 out of 10 valid warnings. But there are certain things that reliably and repeatably set it off. On the way home I drive an uncharacteristically narrow street for California. Would probably be more typical of streets in a place like England. This street is also sloped and bendy. If a truck or van is parked at the apex of a bend with no other cars around the collision warning system goes off. Every time. It does not happen as reliably with smaller vehicles and its less likely to happen if there are cars parked behind and in front. I know I can turn down the sensitivity of the system (I think) but I would rather not do that. Any thoughts?

gmr6415 | 15/11/2018

I'd rather have false positives than false negatives. And yes you can adjust it.

rxlawdude | 15/11/2018

How would the car know you are going to take the bend and not continue straight into that parked van or truck?

pmagid | 15/11/2018

Did not say there was an easy solution...... In fact, I am not sure what the solution is but the first step to figuring something out is to identify a repeatable and reliably reproduced false positive and then maybe testing can begin of various ideas....

Beagle | 15/11/2018

Car was delivered in May, and never had it happen until about 6 weeks ago. Then it became way too common. In the last 2 weeks it does not seem to be as frequent. Got V9 upgrade last week.

rxlawdude | 15/11/2018

@pmagid, how is it a "false positive?" Again, there is no way on God's Green Earth that the car would know you are not headed into the object directly ahead.

pmagid | 15/11/2018

@rxlawdude, by definition its a false positive because this failure mode does not happen as "reliably" with small vehicles parked at the apex of a turn as does with vans or trucks. If the algorithms are smart enough to figured out how to (mostly) not alert on small vehicles parked in this location then this scenario should be possible to engineer a fix to. It is also a false positive because it is something parked off to the side of which I was aware and was not steering into. Its not a car stopping short while my eyes are on something else. Do I need to quote a definition of false positive?

3lawssafe | 15/11/2018

For me the false warnings were also occurring when there was no incidents.
Change the Autopilot Forward Collision Warning setting from the default to late removed the false warnings.

pmagid | 15/11/2018

Did it stop valid warnings as well?

3lawssafe | 16/11/2018

Not sure, there has not been any warnings or incidents yet..
The manual does not mention any difference between medium and late setting.
The brakes are not applied based on the Forward Collision Warning.

rxlawdude | 16/11/2018

@pmagid, no you don't need to define "false positive."

We disagree on the technical feasibility of your edge case.

ftlum | 16/11/2018

How common are false positives with forward collision alerts?

I have had a couple on my model 3 so far (v 42.4) and with the most recent one, I was diving About 30 miles per hour on surface streets, without EAP engaged. My forward collision alert has been at the default setting. There were no cars in front of me as far as I could see and there were no shadows/bridges. It was a partly cloudy day. There wasn't much undulation of the road, but it wasn't completely flat either.

If false positives are fairly common, I'll adjust my settings (and not other the service center).

MichaelB00012 | 16/11/2018

I've had mine go off while driving down the interstate. It seems rather inexplicable some of the times it goes off.

I have also experienced it going off while driving down a little bit of a bendy road and a parked car making it go off.

rxlawdude | 16/11/2018

@michael, yours is the same edge case as @pmagid's.

But that said, there are true false positives that are seemingly random with NOTHING ahead that should have triggered the warning, and sometimes these cause deceleration that is unexpected.

rpc_in_va | 16/11/2018

I get a warning about 10% of the time I pass a particular spot in my neighborhood. It’s an uphill with a slight curve to the right, and a car is parked at the curb in the right. Definitely does not seem like a “valid” warning. Maybe it comes from the fact that I’m starting to turn the steering wheel right as I go up the hill approaching the car? I’ll have to note more carefully exactly what is going on.

pmagid | 16/11/2018

@rxlawdude I dont think we disagree as much as both of us initially thought. I am also unsure of the feasibility of addressing my "edge case". But I do see value in comparing notes with others and helping to collect information.
I love Tesla and I want the product to continue to evolve and improve. Until I see the code or get an engineer to explain to me the AI and algorithms being used, anything I said (or you said for that matter) about feasibility of solutions would be pure speculation. Also, its the edge cases that Tesla will have to spend TONS of time solving if it ever hopes of getting to true self driving. I think we are 80% of the way there but the last 20% is going to take 80% of the time when all is said and done.

ozonelives | 22/04/2019

On the road leading up to my residence there is a slight right bend and almost without fail I get a collision warning at the apex of the turn if a vehicle is parked on the right side of the road. it's consistent and also the only place I've experienced a false positive albeit with only one month of ownership.

wiscy67 | 22/04/2019

I wonder if you could do the voice command to report bug and describe the repeatable condition while you're at that location?

I got my first false positive collision warning in seemingly normal driving with no other moving cars around but don't know what caused it.

ozonelives | 22/04/2019

Yep...I did use the report a bug feature after about the 4th time it happened.

Reflex | 22/04/2019

I would think the software would take in to account the vehicle trajectory before signalling the alarm (e.g., is the steering wheel turned at the time of the obstacle detection?).

howard | 22/04/2019

Both my GM products also trigger at the same spot as the Tesla on the office access road when a car is parked at the exact right spot close on the right in a slight bend. Don't consider it a flaw so much as a predictable sensing condition.

bckator | 23/04/2019

I get a lot of false warnings driving around my neighborhood which consists of curves with parked cars. Moving warning to late reduces but does not eliminate the warnings. Gives me no confidence that this car could ever drive itself.

Sunergy-NJ | 24/04/2019

Rather have a couple false positives now and then instead of missing a real one. And I'm already used to the beep when a parked car is "in my path" on a curved road.

Mnelson5172 | 15/07/2019

My model 3 has avoided a number of accidents that did not exist. No cars around, driving down a two lane road and the audible alarm sounds and the message pops up that an accident was avoided. Very frustrating.

Joseb | 15/07/2019

I get several frontal collision warnings, at low speeds (20-30mph), 2-3 cars distance, plenty of time, no real danger IMO.
how do you report bugs?

jstat10 | 15/07/2019

My 2018 Outback with Eyesight had many false positives. It would often brake for cars that were exiting the highway in the deceleration lane, and for signs or barriers near the apex of sharp curves. The RAB feature would brake suddenly if I backed out of my driveway too quickly. The driveway is steep, and it thought the road was a wall, I guess. It was a great car.

I traded it in for a model 3. Tesla's TACC is much, much better than Eyesight, and Eyesight is pretty good.

calvin940 | 15/07/2019

@jstat10

"My 2018 Outback with Eyesight had many false positives."

Wait what?? Isn't Outback a Subaru?? Fish's holy grail? The measuring stick by which he constantly tries to beat up Tesla with?

Ha ha ha ha

EVRider | 15/07/2019

My false positives have been understandable, for example when I approach a car in front of me that is in the process of turning or moving out of my lane. The Tesla can’t really detect that the other car will be out of my way by the time I get there. I didn’t get emergency braking, just a collision alert. Since a collision alert only benefits the driver, I suppose they’ll get rid of it when there’s no driver.

pmagid | 15/07/2019

Who here thinks @FISHEV owns or works for a Subaru stealership..... Or at least owns a bunch of Subaru stock.....

pmagid | 15/07/2019

By the way.... I started making a bug report every time I got the false positive mentioned above. Did it work? All I can say is the false positives have almost completely gone away in the latest firmwares whereas before they were literally an every day thing at this point on my commute.

rdh37 | 16/07/2019

@jluiswn,

1. Activate mic
2. Say "Bug Report"
3. Speak quickly as it only gives you a few seconds

PECo CT | 16/07/2019

Like @Mnelson5172, my car screeches forward collision warnings at me when there are no other cars on the road around me. Typically, it happens on sunny days driving around curves up steep hills.

pmagid | 16/07/2019

Just curious:
a) is it repeatable
b) what version of firmware are you experiencing this on

PECo CT | 16/07/2019

@pmagid

a) Yes, it is repeatable.
b) It’s on Version: 2019.20.4.2.