I have the same problem. The app reports my home address a full street away from my actual location. Haven't figured out how to correct it.

 

You are referring to where your car is on the map in the app right? Where the car actually is vs where the app says you are? Have a friend in Dallas with Android and it shows his car no where close to his house when he is home. Other reports here about the car showing to a house or two away or across the street.

I've had very good luck with this regardless of the cars at home or work or on the road. It's been dead accurate. I can even tell you which side of the garage each is parked on by just looking them up on the app. It is that precise. Using an iPhone X.

I don't truly know if this is some locations just aren't mapped that well or if it is device dependent.

 

Yes, referring to when the car is on the map (I'm usually watching it from home, as I work from home and my wife drives it most of the time). It reports where my phone is precisely. It also reports where the car is very precisely when it's parked, including which direction it's facing and I can also tell where exactly it is in a large parking lot. It's remarkable. But then she starts driving and it's wildly inaccurate when it reports which street she's driving on. Sometimes getting it exactly correct, but often reporting she's on a side street or a parallel street. Disappointing when I know this isn't impossible to get right and other apps can do it with few issues.

 

I have the same exact scenario as yours, I sometimes I see her going 80mph and the app says she is on a feeder next to the freeway, she's obviously on the freeway.

 

First, GPS isn't as perfect as everyone thinks it is. There's a lot more variation in position than even the posted specs. Not that the posted specs are wrong, sometimes you need to read deeper.

A car in a garage is a unique situation, dependent on all sorts of factors, the car may not even be able to hear the satellites or even cellular sites.

With good visibility, GPS position gets much better over time, as the position is averaged, but if the car can't hear, well...

 

A car in a garage is a unique situation, dependent on all sorts of factors, the car may not even be able to hear the satellites or even cellular sites.

With good visibility, GPS position gets much better over time, as the position is averaged, but if the car can't hear, well...

You seem to have missed my point. It is VERY reliable when parked in the garage. It knows exactly where it's at, without fail. It's when driving, under open skies, on major highways like I25 in Denver, that it has all of the issues.

 

The on-board navigation has been pretty good for me except when I ask it to take me home. It gets me onto my one-block street OK, but it tells me my destination is on the right; whereas, my house is on the left! Maybe I'll try entering my street from the other direction and see what it says.

 

Tesla's Full Self-Driving patent uses cars as Earth-based GPS satellites

 

I'll do some testing, but it's a non-trivial problem when you are in a moving vehicle. Can you comment on how the app does when you are parked at a known address? When you are parked, the major cause of errors would be the reverse address lookup database - which is almost certainly Google's if you are using the Android app (probably is on the iPhone too). Google's is probably one of the best there is, but there will always be some inaccuracies. I'd be curious if you checked it at half a dozen known addresses a fair distance apart (i.e. not several addresses on the same street) while stopped for at least a minute.

The moving vehicle is a whole different problem. Although the Tesla API reports position, speed, and heading (it's almost certainly track not heading, but close enough), the phone app is probably only using the position itself for the reverse address lookup. A dedicated navigation app probably needs to do better than that, and take into account which road it thinks the vehicle is on, using the velocity and recent history to help make that determination. But I don't see that the engineers working on the Tesla app would think that an exact address of a moving vehicle is all that important. In general I would think the two reasons to use the location in the app is to find your parked car (in which case it's not moving) or to report a stolen vehicle. If the stolen vehicle is moving, the reverse address lookup probably isn't as important as just simply watching the vehicle moving on the map.

First, GPS isn't as perfect as everyone thinks it is. There's a lot more variation in position than even the posted specs. Not that the posted specs are wrong, sometimes you need to read deeper. ... With good visibility, GPS position gets much better over time, as the position is averaged, but if the car can't hear, well...

I used to work on GPS receivers (but not the chipset)... The basic system is only accurate to 45 feet. There are lots of things that will affect the position resolution. One of the biggest is the view the receiver antenna has of the sky. Your phone inside a normal car typically has a lousy view of the sky, and this can make for a poor position solution. Generally speaking, the more satellites the receiver can see, the better the solution will be. However, in the case you're talking about we are using the cars position sensor which presumably is mounted on the top of the car to have a clear view of the sky. I believe with the S/X it's under the rear window - anyone know for sure where it is on the Model 3?

When you have a good number of satellites, the condition of the ionosphere can still make a big difference. The system works by measuring how long the signal takes to get from the satellite to your receiver. However, the ionosphere is non-uniform so depending on the situation, the timing from one satellite can be more or less accurate than from another satellite, and this can effect the precision of the computed position.

The geometry of the satellites also has an effect... you could have a case where you see 4 satellites, but in one case you had an optimum geometry and in another you had a poor geometry... and in that case you could have quite different precision between the two solutions. Since the satellites are in low earth orbit, their positions and thus the geometry is constantly changing. You can be sitting in one spot and get different precision hour to hour as the satellites change position in the sky.

There are other things that affect GPS... driving in a city (we call them urban canyons) can block enough of the sky to limit the number of satellites visible to the receiver.

As to Ed's point about the GPS position getting better over time, if he's talking about a stationary receiver, that can be true (but probably isn't for most receivers). The company I worked for had a GPS accurate to 1cm... but it took about 15 minutes to get that solution, basically averaging the position. That doesn't help in a moving vehicle.

Finally, keep in mind that phones often don't use the GPS for position... because GPS uses more power than triangulating on cell phone towers (which it has to do anyway as part of the cell phone functionality... I've worked on the cellular infrastructure as well). So, depending on the situation, it might be that your phone is in a position that cell tower triangulation isn't giving a very accurate solution, but the phone hasn't decided it's worth turning on the GPS receiver... But that's if you're using your phone in the car (which I do - I run Waze at the same time I use the Tesla navigation system). That doesn't apply to what is being discussed here, where the car's receiver is almost certainly the position source.

I don't know what kind of position sensors Tesla uses in the car itself... It would make sense to use the cell phone triangulation as well as GPS to solve the urban canyon problem, but I really have no idea... Another interesting question is whether Tesla uses the WAAS (Wide Area Augmentation System) satellites - we do in aviation because you get much better position information - (nominally 9 feet) as well as some reliability improvements. However it would increase cost slightly and only be useful in the continental United States, so they may not have bothered.

 

I'll do some testing, but it's a non-trivial problem when you are in a moving vehicle. Can you comment on how the app does when you are parked at a known address? When you are parked, the major cause of errors would be the reverse address lookup database - which is almost certainly Google's if you are using the Android app (probably is on the iPhone too). Google's is probably one of the best there is, but there will always be some inaccuracies. I'd be curious if you checked it at half a dozen known addresses a fair distance apart (i.e. not several addresses on the same street) while stopped for at least a minute.

The moving vehicle is a whole different problem. Although the Tesla API reports position, speed, and heading (it's almost certainly track not heading, but close enough), the phone app is probably only using the position itself for the reverse address lookup. A dedicated navigation app probably needs to do better than that, and take into account which road it thinks the vehicle is on, using the velocity and recent history to help make that determination. But I don't see that the engineers working on the Tesla app would think that an exact address of a moving vehicle is all that important. In general I would think the two reasons to use the location in the app is to find your parked car (in which case it's not moving) or to report a stolen vehicle. If the stolen vehicle is moving, the reverse address lookup probably isn't as important as just simply watching the vehicle moving on the map.

I appreciate the details, learned some new things about GPS. I guess my point is that other apps (like Life360) can track family members via their phones in a very precise way while they are moving in their ICE cars. Why can't Tesla's app do the same when it KNOWS where the car is at all times in a very precise way?

 

So, to add to my post...
Mobile/Car GPSs are a little different from hiking or aviation GPSs. A car GPS assumes that it is in a car. As such, it assumes that the car is on the road. To keep people from complaining that their car is driving into buildings or on the wrong side of the road, GPS manufacturers have added a "snap to road" feature. This attempts to make sure that the car shows itself on a road. When their are close parallel streets, or in areas with lousy coverage (cities with tall buildings) the accuracy gets lousy enough that the "guess" of which road it is on just gets it wrong. This can often be seen when you are taking turns and the car has to guess again, or if the road direction just doesn't make since (the parallel road that you got incorrectly pinned to stops) It happens all the time, but few people ever watch their GPSs that closely. On the Tesla, the on-board nav is SO MUCH better than any other car, that people start using it and see the issues.

Don't believe me, get a hiking GPS and watch what it does. I remember years ago when I was using a hiking GPS as my mobile, it showed the car going out overnight and driving all over the neighborhood, sometimes a mile+. This didn't have snap to road, so it indeed drove through all of the houses.

@BostonPilot I remember the first time that I saw a GPS in about 1982, It was a 2 ft square box and if you set it in one place for about 2 days, it could tell you where it was. (With only the few satellites that were up, it took days to get enough passes to triangulate position.) This was a quite remarkable experience in the day.

And for others, did you realize that the 1990 Census paid for by your taxes, made street guidance available? It created the Tiger Map database with essentially all of the roads in an electronic format that was used for the first street maps on GPSs.