two dimensions out of three

[cellio]

It's funny to see (well, hear) my phone's navigator app react to parking garages. "Do X... oh ok you're going north so do Y... oh you're going west so do Z... oh you're going south do A... oh you're going east do X which I'll pretend I haven't said before..." -- iterate until you reach the exit. It doesn't respond to elevation, only latitude/longitude.

I can think of three possible reasons for this, and I wonder which it is (or if it's something else I haven't thought of):

1. The GPS in the phone doesn't detect altitude.

2. The map data (Google's, in this case) doesn't record elevation. It does you no good to know that the GPS is at a certain elevation if the app can't tell that that's 200 feet above the road, after all.

3. The GPS and map data are available, but the app isn't programmed to take it into account. How often does this really come up, after all?

Comments

[browngirl.livejournal.com]

*reads, nodding contemplatively*
*gets to the cartoon*
*falls off bed laughing*

[eub.livejournal.com]

I suspect it's mostly (3) in play here. As a bug it's probably more silly than it is harmful or annoying?

GPS does inherently pick up altitude, but it's less accurate than location horizontally, i.e. the error ellipsoid is tall. Basically because you do not generally have satellites available near directly overhead and rarely straight downwards -- the constellation you see is pretty 'flat'. So when you triangulate position based on your approximate distance to the satellites, the vertical error is rather high, because of the geometry of the solution.

On top of that, seeing satellites from inside a parking structure is hard, so who knows if your GPS has any idea of your elevation when you're in there.

Originally map data didn't come with much elevation data at all, even relative. (You may recall the days of "turn left from Forbes Ave onto S Neville Way" directions from map sites.) It's fairly easy to paint your road lat/long data with elevation datasets derived from orbital radar ranging, but what that gets you is the *top* of your parking garage, or the trees covering the road, etc. Actual elevation of the roadbed has been a longer time coming. You can speculate that crowdsourced GPS data might be useful here.

[cellio]

Thanks for educating me!

(You may recall the days of "turn left from Forbes Ave onto S Neville Way" directions from map sites.)

I do indeed. I didn't know if the change there was due to better satellite info (or better use of it), or crowd-sourced corrections to the map set. I know I've reported bugs in directions at times and wasn't sure how those got corrected.

I don't consider this a bug at all; an app that (somehow) took this into account would get a "hey, cool!" reaction, but it's far from necessary. As I said, how often does this come up?

(I suppose detecting "driving in circles in a sub-block diameter" could provide something useful.)

[eub.livejournal.com]

By the way, to give credit where it goes, fixes to data like the Forbes/Neville grade separation were probably made by the map data vendors (Navteq, TeleAtlas, et al.) through old-school "vans on the ground", driving around and taking notes on the roads. Just by the timing, before any of the fancy stuff was around much.

[cafemusique.livejournal.com]

My impression, just looking at some of the tracks and profiles I've recorded with my phone is that the elevations reported by most phone GPS hardware are subject to a lot of error. Also, in a garage, there's an obstructed view of the satellites, so the accuracy is even lower. I am not sure that the GPS chips in phones are precise enough to determine what level of a multi-level structure one is on.

[dagonell.livejournal.com]

I would be content with "Exit from Parking Garage".

[cellio]

Good point. I wasn't thinking about "what level of the garage" so much as "are we at street level?".

Hmm, I wonder what you get if you're standing on the top of the wall surrounding Jerusalem's old city and ask for (walking) directions within. Maybe I'll remember to test that someday.

[anastasiav.livejournal.com]

The flip side of this (sorta) is that for many personal fitness devices (ie: FitBit et al) they will sometimes calculate that you're doing activity based on changes in elevation even when you're sitting in a car.

Thus, my friend drove up Mount Washington in a car and was "rewarded" by her fit bit for climbing 796 flights of stairs that day.

[sethcohen.livejournal.com]

I wasn't available to respond to this post when it was first published, but realized that I wanted to get back to it. Most of the cogent things I would have said have been said by others, but there's one more thing I wanted to point out.

Your phone will read the GPS signal and tell you your location based on the signal timing. The phones are only programmed to do so appropriately on the surface of this planet. They will tell you that you're in Pittsburgh (and where appropriately) even if your elevation is twice orbital distance (based on the orbit of the GPS satellites it's receiving signal from) above Pittsburgh.

I have the same problem when I'm in my parking garage at work. I've been coming from enough different routes long enough that now when I see work and get to a stop light, I exit my route so my phone doesn't bitch at me for three minutes as I climb floors in the garage.

[eub.livejournal.com]

Now I'm curious just how high an altitude GPS receivers (consumer, specialty, military) are programmed to solve for. Do they actually handle when their location is above the satellites? Just based on the precision of their ranging, it should be possible to distinguish between convexity and concavity of satellte 'shell' (obvs after they've acquired 4+ satellites). But if dropping that capability saved the designer a millisecond or a dime, I have to assume they'd shed one geeky tear and drop it instantly.

If they do support it, I hope they maintain test coverage. :)