FSD + GPS (in)accuracy = fail

[Mark II]

Often times when I first start out on a drive the GPS accuracy is really poor, as it tries to find more satellites. This would not be an issue, except FSD is an absolute nightmare during this period.

On the display my car is clearly nowhere near the route, sometimes half a block away. As a result FSD keeps trying to turn at the next intersection to route my car back to the route. It also randomly breaks the whole time as it tries to lower the speed for "parking lots" etc.

Most, if not all, GPS receivers know how accurate they are based on number of satellites they have a lock on. I would think FSD would rely less on the GPS and more on the map when the receiver is indicating a large error.

Thoughts?

 

[hugh_jassol]

sounds like you should ask to get the GPS issue resolved. I do not see this - it's got my position right on and I park in a garage at home and concrete parking deck at work

 

[Dewg]

I agree with Hugh, never had a GPS issue. I've seen videos of FSD failing that were GPS issues where mapping data shows an upcoming left, but there was no left because the car wasn't on the map where it thought it was. Personally I'd open a service request.

 

[Mark II]

sounds like you should ask to get the GPS issue resolved. I do not see this - it's got my position right on and I park in a garage at home and concrete parking deck at work

This normally only happens when I have not driven for a few days. Which is typical for GPS receivers. The longer since the last time they were active the longer it takes to lock. That plus the fact we have tall trees all around us in the Seattle area.

I should note, it usually locks after about a half mile or so. Then all is well.

 

[dckiwi]

It's happened to me a couple of times, and both times it was pretty alarming. I was driving on an interstate, and I noticed that the max speed dropped to 30mph. The nav system showed my location as about 500 meters to the right, apparently just plowing through houses in a residential neighborhood. Needless to say I turned off AP until the GPS corrected itself several miles later.

 

[Dewg]

Good Lord, that's totally unacceptable. GPS should be dead on at all times. I start my car and GPS has it right inside my house on the map before I even pull out.

 

[Tronguy]

Yeah.. So, I'm an amateur radio operator and a EE. And I repair/diagnose/sometimes fix strange electronics systems for a living. And I've had those GPS hand-helds where one can see the N satellites the receiver is tracking, watch the box lock to them one by one, getting more accurate as it goes.

But my car's garaged, with a roof directly overhead. And there's always (based upon past data) a dozen satellites buzzing around up there. And I don't have GPS issues like you. I get accurate GPS when parked in there.

I take a wild guess. Your GPS receiver has an antenna and receiver front end and one, the other, or both is damaged. Could be a bad coax connector, bad transistor, who knows? But my bet is that something's damaged and your signals are reduced in amplitude.

There's this thing, FIT rate. FIT stands for Failures In Ten to the 9th hours. An individual single resistor is known to have a FIT of 1.0. Which means that an average resistor's mean time to failure is a billion hours. Or another way to look at it is if one has one billion resistors out there, you're going to get one failure per hour, average.

An average IC has a FIT of around 100. A hot switching power module (heat speeds up failures, see Arrhenius equation) might be 500. A complete circuit board covered with ICs, capacitors, resistors, switches, and what all would be 1000-2000. An entire car: Madly guessing, 100,000. And, at this point, one is talking 1e9/1e5 = 1e4 hours for a car: But if one builds 10,000 cars, that means one is going to get one car failure per hour. Whee.

And that's why there's warranties. Whoever it is who's building GPS modules for Tesla is going to have some of those escape factory testing and fail in the field. What you're describing sounds like a field failure. Since nobody else on this thread is reporting behavior quite like yours, and trees don't stop properly functioning GPS receivers, I'm thinking yours is truly malfunctioning. Service ticket time.

 

[nvx1977]

This normally only happens when I have not driven for a few days. Which is typical for GPS receivers.

No it's not typical at all. It is more typical if you suddenly change geographical locations (more common with phones, since you usually have them in airplane mode while traveling). The A-GPS data goes stale, and you need to redownload the A-GPS data for your new locale in order to establish fast connections with the satellites.

Feels like you need to have your car checked out at an SC.

For all of 2020, I drove my car infrequently due to covid lockdowns. Never had any problems with GPS inaccuracy.

In the 4.5 years I've owned my Model 3, I've had the GPS position me incorrectly 2x. And yes, while it's wrong, AP was very wonky. I tend to agree that Tesla could put a check in to disable AP/FSD while GPS confidence is low. But the frequency that you have GPS issues is not normal.

 

[Mark II]

No it's not typical at all. It is more typical if you suddenly change geographical locations (more common with phones, since you usually have them in airplane mode while traveling). The A-GPS data goes stale, and you need to redownload the A-GPS data for your new locale in order to establish fast connections with the satellites.

It is typical: Time to first fix - Wikipedia

My situation is a warm start. So could take up to 45 seconds, assuming good satellite coverage, possibly longer when view obscured by building and trees.

Feels like you need to have your car checked out at an SC.

For all of 2020, I drove my car infrequently due to covid lockdowns. Never had any problems with GPS inaccuracy.

It's possible you may not have noticed back in 2020. This is not an issue unless I use FSD, which has only been out for the last year.

In the 4.5 years I've owned my Model 3, I've had the GPS position me incorrectly 2x. And yes, while it's wrong, AP was very wonky. I tend to agree that Tesla could put a check in to disable AP/FSD while GPS confidence is low. But the frequency that you have GPS issues is not normal.

I have had GPS be wrong on the freeway, but very seldom like you. Possibly because the freeway has clear view of the sky, unlike the narrow roads between tall trees near my house. Also, by the time I get to the freeway it has a lock anyway.

The issue is only within the first minute or so after leaving the house. I.e. the warm lock time.

 

[Mark II]

Yeah.. So, I'm an amateur radio operator and a EE. And I repair/diagnose/sometimes fix strange electronics systems for a living. And I've had those GPS hand-helds where one can see the N satellites the receiver is tracking, watch the box lock to them one by one, getting more accurate as it goes.

Lol, I am also an amateur radio operator and an EE. I also worked extensively with GPS in the 90s. Yes the receivers have improved significantly since then. So it may be that both my cars have HW issues?

Would be interested to hear from others that have tall dense trees around them.

 

[JulienW]

I park in a HUMONGOUS parking deck (over 1¼ miles in circumference) that has about 8 different enter/exit points. So when I pull out GPS has lost track and have to wait on a GPS lock. Also buildings all around. This can sometime take a few minutes and is limitation of the GPS system and can't easily be overcome. I even pull over sometimes and wait on GPS to lock or I manually drive for a few blocks.

EDIT: Just to add we have a 30 stall Supercharger in the parking deck so this will be a much larger Tesla problem.

 

[nvx1977]

It is typical: Time to first fix - Wikipedia

In that same wiki article, it mentions the use of A-GPS to speed up that first fix issue. Since the car is using the same cell towers as phones (I assume it's still AT&T; haven't verified in forever), the car should be able to grab A-GPS data all the time, ensuring a quick lock with the satellites no matter where you are or how much the satellites have moved since last lock.

I've had sports watches that don't have access to A-GPS, and each time I use it, it takes 2-3 minutes to get a lock. I'm pretty sure Tesla's GPS receiver is leveraging A-GPS.

 

[Dewg]

I park in a HUMONGOUS parking deck (over 1¼ miles in circumference) that has about 8 different enter/exit points. So when I pull out GPS has lost track and have to wait on a GPS lock. Also buildings all around. This can sometime take a few minutes and is limitation of the GPS system and can't easily be overcome. I even pull over sometimes and wait on GPS to lock or I manually drive for a few blocks.

EDIT: Just to add we have a 30 stall Supercharger in the parking deck so this will be a much larger Tesla problem.

We may be stumbling onto something here - some people with FSD problems may be having GPS issues contribute. One of my offices has a pretty massive car park (4 entrances, 6 levels), and my car locks onto GPS the moment I start the car no matter where I am in the car park. As I'm driving around the car park, the GPS has me exactly where I should be on the map. I'm curious why some cars don't lock on instantly, or take several seconds to a minute or two to get correct GPS.

 

[sleepydoc]

I’ve had a similar issue (and posted about it) when leaving the parking garage at work. If you’re in a big/dense enough structure then it loses the GPS signal and it takes a minute to lock on. The compass reading may be wrong, too, although that should fix more quickly than the GPS. Either way the car doesn’t know exactly where it is and possibly not which way it’s going. In my case it started to blow right through a stop sign.

FSD should be able to navigate a road (i.e. recognize and stop at a stop sign) without GPS but if it doesn’t know where it is it’s going to be impossible for it do know where to go. It would be like plopping me down in Rome and saying “find the Colosseum.” All I could do is go down streets, randomly tuning at each intersection.

 

[Mark II]

In that same wiki article, it mentions the use of A-GPS to speed up that first fix issue. Since the car is using the same cell towers as phones (I assume it's still AT&T; haven't verified in forever), the car should be able to grab A-GPS data all the time, ensuring a quick lock with the satellites no matter where you are or how much the satellites have moved since last lock.

I've had sports watches that don't have access to A-GPS, and each time I use it, it takes 2-3 minutes to get a lock. I'm pretty sure Tesla's GPS receiver is leveraging A-GPS.

Ha ha, maybe that is my problem! The cell coverage in my garage is dismal. I barely get 1 bar.

 

[boonedocks]

Often times when I first start out on a drive the GPS accuracy is really poor, as it tries to find more satellites. This would not be an issue, except FSD is an absolute nightmare during this period.

On the display my car is clearly nowhere near the route, sometimes half a block away. As a result FSD keeps trying to turn at the next intersection to route my car back to the route. It also randomly breaks the whole time as it tries to lower the speed for "parking lots" etc.

Most, if not all, GPS receivers know how accurate they are based on number of satellites they have a lock on. I would think FSD would rely less on the GPS and more on the map when the receiver is indicating a large error.

Thoughts?

Is this on your MSLR Refresh?

I have the same issue and have reported it ad nauseam to Tesla / Report Bug / snap shot / emails et.... and have gotten nothing back. My home roof boards are the metallic heat reflective boards covered with shingles.

I have this issue almost every day when leaving and sometimes takes quite a while to correct itself. The main road for my commute is 45mph and there is a residential street running parallel that is 25mph. My car will continually turn on the blinker, brake check anyone behind me, readjust the speed to 25mph (based on the map where it thinks it is at) and is so bad I have relegated to just FullbymySelfDriving when I see my car driving off road on the map.

Our 2017 Model X nor my 2015, 2017 or 2019 Model S had this issue.

Here is a pic of the most resent example that was emailed, in car bug reported, snapshot etc....

 

[sleepydoc]

Is this on your MSLR Refresh?

I have the same issue and have reported it ad nauseam to Tesla / Report Bug / snap shot / emails et.... and have gotten nothing back. My home roof boards are the metallic heat reflective boards covered with shingles.

I have this issue almost every day when leaving and sometimes takes quite a while to correct itself. The main road for my commute is 45mph and there is a residential street running parallel that is 25mph. My car will continually turn on the blinker, brake check anyone behind me, readjust the speed to 25mph (based on the map where it thinks it is at) and is so bad I have relegated to just FullbymySelfDriving when I see my car driving off road on the map.

Our 2017 Model X nor my 2015, 2017 or 2019 Model S had this issue.

Here is a pic of the most resent example that was emailed, in car bug reported, snapshot etc....View attachment 861991View attachment 861990

Does it fix itself eventually or is the GPS always off?

 

[boonedocks]

It will eventually (most of the time) fix itself but sometimes will still be a lane or two off on the expressway.

ie. it will turn on the left blinker to get in the HOV lane when it is already there and the only thing to the left is a concrete divider. Doesn’t actually try to change lanes since it is a solid white line but I have to turn off NOA to stop it from repeating.

 

[Mark II]

Is this on your MSLR Refresh?

Both 2018 MS and 2022 MSLR. Although I don't know if one is worse than the other. I mostly drive the 2022 and my wife the 2018, and she does not use FSD. To stressful for her .

When I did drive the 2018 before the 2022 it would have similar issues with GPS no locking for a while, although I did not have FSD at the time, so it did not bother me much.

 

[Mark II]

It will eventually (most of the time) fix itself but sometimes will still be a lane or two off on the expressway.

ie. it will turn on the left blinker to get in the HOV lane when it is already there and the only thing to the left is a concrete divider. Doesn’t actually try to change lanes since it is a solid white line but I have to turn off NOA to stop it from repeating.

Once mine locks, typically within a mile from my house, it is generally very accurate. I don't have issues on the freeway, since by the time I get there is has already locked.