Accuracy Of GPS Data In E72 Using JSR-179

[rohiteshd]

Hello,

I am using JSR-179 (code snippet below), to obtain GPS location. However, I am not too happy with the accuracy of the data. I am getting GPS location, which sometimes varies by close to 100m, even when I have set the HorizontalAccuracy criteria to 10. Can anyone give me some pointers on this, please?

What I am doing :
My Midlet in an E72 phone, records the GPS location and stores it.
I then plot the recorded GPS location, on Google Maps.

Code Snippet in Midlet to obtain GPS location :
// set the criteria and get the location provider
Criteria criteria = new Criteria();
criteria.setHorizontalAccuracy(10);
LocationProvider locationProvider = LocationProvider.getInstance(criteria);
Location location = locationProvider.getLocation(20);
Coordinates coordinates = location.getQualifiedCoordinates();

Expected Result :
The GPS location obtained, should be within 10mt radius, of the actual location.

Actual Result :
The GPS location accuracy varies. Sometimes, it is well within 10mt radius. Sometimes, it is varying by close to 100mts.

Cheers,
Rohitesh.

[Tiger79]

Actually your expected results should be somewhat different...
Accuracy on the cheaper integrated GPS chips (such as used in mobile devices) have an accuracy of up and around 15 meters...
Thus if you are talking about a radius then you might expect a radius of 30 meters across, after all it could say you are 15 meters north or 15 meters south of your position...
Also the GPS signal is very very important, with 4 GPS satellites you will barely have an "accurate" position, even though the number of satellites doesn't say everything, for example you can have a better accuracy with 4 satellites which are spread over the whole visible horizon than having 7 satellites all clustered into a particular quadrant of the sky (for example being in an office, having your GPS receiver at a window and having a signal from 6 satellites, it will be a bad signal even though 6 satellites might make you assume its a good one)...

[rohiteshd]

@Tiger79

So, what is the best way to get accurate GPS fixes, using Nokia phones? Without it, making any kind of Locational Applications would be very difficult.

Cheers,
Rohitesh

[Tiger79]

as opposed to what ?
You will have to settle with the GPS locations you are getting from the GPS chipset, even if sometimes it means it's 20 meters off...
One way just to have an idea if a location is precies would be to retrieve the number of visible satellites, together with the HDOP (horizontal dilution of position) from the NMEA string of the gps chipset, specifically within the GPGGA sub-element...
This means you will have to learn more about the GPS technology, how it works, what the above means etc...
Otherwise do as I said earlier, settle with the positions you are getting... Just keep in mind you don't have a very accurate or expensive gps chip in your device...
I know people who have 30 cm precision GPS receivers (measurement teams for the city plans here) but those cost around 15.000 euro's a piece, you cannot compare it to a 300 euro cellular phone :P And I can assure you that difference in money has a reason
Those devices for example will only flag a gps position as valid if it is stable throughout a series of measurements and it won't measure for example if there are less than something like 7 or 8 satellites in view...

[rohiteshd]

@Tiger79

Thanks a ton Tiger. Actually, I do know a bit about GPS technology. A coupla years back, I worked on a hobby project, using a uController and a GPS receiver, to record a bunch of locations. So, I know a bit about the number of satellites needed to get a fix, and the HDOP. But I am new to programming this, on a Nokia SmartPhone. So, I don't know, how to get the NMEA string here. Actually, in other hobby project (mentioned earlier), I used to get the whole NMEA string from the GPS receiver, and I had to parse it to get the Lat/Lng values.

Can you pass me some pointers, on how I could get access to the NMEA string, instead of just the Lat/Lng values, using JSR-179?

And thanks a gazillion tons, for the help. Sure do appreciate it.

Cheers,
Rohitesh.

[Tiger79]

well because of several NDA's I have signed with my boss I simply cannot post code, I can tell you though that through the JSR-179 it's possible to retrieve the NMEA string, and to be honest you're lucky because the JSR-179 isn't that big of an API, so just reading out the whole JavaDoc shouldn't cost you more than a couple of hours and it will also give you a good insight in the capabilities of the API itself...
For example I just tried looking it up with Google and I had 4 hits right away in the matter of seconds, took me a lot more time to write this sentence than looking it up...

[rohiteshd]

@Tiger79

Right after I posted that, I was searching through these forums, and found a number of hits! And pretty much all of them had you in them!

Btw, are you with Nokia? I am. Bangalore, India.

Cheers,
Rohitesh.

[Tiger79]

No i'm not with Nokia, I'm just a Champion...
And yup, you might see me a lot regarding GPS, JSR-179 and LBS... I kinda worked a lot with that and specifically on the Nokia platform, I also wrote a couple of Wiki articles about that...

[rohiteshd]

@Tiger79

Tiger, I worked on this for the weekend. One question that still eludes me is, with this kind of accuracy, how can anyone develop a reliable LBS on Nokia SmartPhones? What is the way around it? If getting direct Lat/Lng values is not good enough, what other services do we have access to, which can help us get a better locational fix?

Cheers,
Rohitesh.

[Tiger79]

Well, with a good signal the accuracy can be pretty accurate
I have seen pretty precise positions (up and around 5m) with optimal situations (regarding position and number of satellites)...
That's accurate enough for me and the applications I need it for, and for me that's still a reliable platform...
There are no "better ways", i.e more precise, than a gps fix...
Using cell-id's is way worse, depending on where you are (within a city or countryside) the precision might be inbetween the 100 meters and a couple of km's...
Just to be clear : the retrieved gps coordinates are the most precise way to get a location, BUT if the signal is weak/bad then keep into account that the precision may be something like a radius of around 15 meters or more, whilst if the signal is good you will have pretty good precision (up to 5 meters)... It might sound a lot to you, but this system is actually able to pinpoint your position on a planet of more than 510 million square km's !!!
If you really really need to have more precision then you might look around for some more stand-alone professional GPS receivers, maybe with a Bluetooth connection... But I don't think many users would want to walk around with those and/or buy one seeing their prices...

[rohiteshd]

@Tiger79

Point taken!

I think I will fool around with the NMEA string for a while. Not proceed with the application, till a fair enough Locational Fix can be obtained (will ascertain this by the number of satellites and the HDOP value). Any thoughts on that?

Cheers,
Rohitesh.

[Tiger79]

well there are several options you could consider...
For example, you might want to give some kind of visual feedback to your user...
Like a GPS icon which is red when the signal isn't very good (thus possibly an unprecise location), orange if it's somewhat better and green if ti's a good signal...
To determine which of these three categories your current signal is you can implement your own algorithm, whcih takes into account the number of available satellites, and maybe the HDOP...
You won't make the signal any better, simply because you can't, BUT you will give some feedback to the user telling him when you have a good signal and when you don't simply because that can happen just as well...

[rohiteshd]

@Tiger79

Perfect! That sounds good. Will implement it. And after am done, will post a comment here, on how it went down with my users!

Note : As this is a hobby project, my users are mostly my friends!

Cheers,
Rohitesh.