hile, GPS geeks! to me!

[topaz]

At work we are exploring the possibility of using reverse geocoding to identify a VoIP user's location for 911 calls: we get their GPS coordinates, look them up in a GIS server and figure out which street address they're at.

There are a few sites which allow reverse geocoding lookups over the Web, mostly on a free-trial basis for evaluating someone's pricey GIS server software. I'm using these to determine how accurately we can get a user's fix. For that, I need data. And that, dear reader and owner of a handheld GPS device, is where you come in.

I need to collect a bunch of street addresses and the latitude/longitude that a GPS receiver reports at each of those locations. On this side I can do the reverse lookup and see how closely it matches your actual address. The more far-flung the locations, the better, as long as they're in the 50 U.S. states. (I'm not sure if we're required to provide 911 service in Puerto Rico or Guam.) I will be happy to tell you the results I get, if that makes it more appealing. :-)

If you'd like to help, please feel free either to post the results here or mail them to twp@rnktel.com, whatever you prefer :-)

Comments

[points.livejournal.com]

Would it be helpful to put you in touch with a friend of mine who works for a company that provieds e911 DGPS information, as well as makes 'low signal' GPS hardware for cell phones? I suspect they've already addressed a lot of these issues, and he may be able to give you other pointers...

[qwrrty.livejournal.com]

Would it be helpful to put you in touch with a friend of mine who works for a company that provieds e911 DGPS information...



Why, yes. Yes, that would be very helpful indeed!

[tugrik.livejournal.com]

It's AGPS, not DGPS, silly pointsdragon. :)

*waves to Topaz*

Greetings! I'm said friend. The company is Global Locate (http://www.globallocate.com). Whatcha wanna know? :)

[qwrrty.livejournal.com]

Surfacing long enough to gasp for air a bit...

Well hi! What do I want to know.... well... ANYTHING!

Seriously, here's our current problem in a nutshell. We're a VoIP company. We need to provide 911 operators with a caller's current physical location, but we don't technically know what that is -- they could have picked up their IP phone and flown to Toledo since the last time we heard from them.

Our current plan is to connect each phone to an A-GPS receiver that phones home once in a while to let us know where they are. We would like to be able to take their GPS coordinates, automagically turn those into street addresses, and deliver whatever address that is to the 911 folks. I am coming to believe that's an impossible problem -- reverse geocoding just doesn't seem to be sufficiently accurate to satisfy the needs of the emergency services folks. But maybe there are some tricks I haven't yet figured out.

If you have time to ruminate on some of this or share anything about what you know, please feel free to write me at twp@rnktel.com. Seriously, we don't have anyone with solid GPS experience here, so we don't have a really clear idea of what's possible, what's implausible and what's completely off the table. So I'd love to talk to someone who's done serious work in this space.

[tugrik.livejournal.com]

I've passed your request on to one of the VPs here. They're all in boardmeeting stuff this week, but he should see it before the week is out. The reverse geocoding you speak of is very doable actually and is already part of the e911 infrastructure. Much of it you'd not have to do; it's handled on the emergency services' ends of things. You just pass them lat/long of the device.

[rwx.livejournal.com]

say i'm standing in the middle of a park when i gps (wireless access), what address would i report then? do you care whether we're near potential wireline or not?

--Corprew

[qwrrty.livejournal.com]

Yeah, this is strictly for wireline services. The idea is that if you buy VoIP service from us, and then move across town or take the phone with you on vacation, and plug it in at your new house, we want to see if we can figure out your address automatically for submitting to the ALI. Which means our primary concern is getting street addresses right.

If you've done something cracked like plugging in your phone to a wireless hub at Burning Man, we won't be able to get anything useful to tell the 911 operator anyway (and I'm sure they'd really love to see "BLACK ROCK CITY" on their terminal).

[ronebofh.livejournal.com]

... "hile"?

[qwrrty.livejournal.com]

Sorry. I finally finished the Dark Tower (http://www.tamisclock.com/archives/000621.html) books and it's still kinda stuck in my head.

[ronebofh.livejournal.com]

So, what is it supposed to mean?

[tfarrell.livejournal.com]

My experience with several handheld GPS devices over some years is that at present, you can reasonably expect to get a reading within about 50 feet of the actual location. However, once in a while odd glitches have made it read about 500 feet wrong for as much as 10 or 20 seconds. So, best to take a number of readings over time, discard any wild anomolies, and average the results.

Note however that I said you can reasonably expect accuracy within 50 feet or so... in Somerville, there are probably 6 or 8 homes within 50 feet of my front door. So, it's not very useful for dense neighborhoods.

[qwrrty.livejournal.com]

Right, these are exactly the concerns I have. Which is why I want to collect data: to determine quantitatively just how accurate we can expect the results to be. Which I didn't get. Oh well. :-)

[tfarrell.livejournal.com]

The problem is, the accuracy is variable. It depends on factors such as the individual receiver type, the type of antenna hooked up to it, cloud cover, precipitation, how tall the surrounding buildings are and how close together they are, how many trees are nearby and how close together they are...

You could probably come up with a quantitative average accuracy, but it would be meaningless in terms of individual cases.