Slashdot Mirror
Forget GPS, Hello WPS
No France writes "A company known as skyhook wireless has announced the commercial availability of its Wi-Fi Positioning System, or WPS. The company has compiled a database of every wireless access point it can find in a given city. When a mobile user running th Skyhook client is in a recorded area, their position is calculated by selecting the surrounding signals and comparing them to the reference database. Currently there are 25 US cities mapped, including New York City, Boston, Chicago, Los Angeles and San Francisco. Apparently this device is accurate to within 20-40 meters, though one has to wonder how well it deals with people moving their wireless access points."
PlaceLab has been doing this for a while, and it's free.
Martin May
Basically, the military figured out how to easily jam GPS in an area. But before then, there were GPS field units available that averaged out the error and got better than 2-3 meters so that it didn't really matter that much...
Sub-5-meter accuracy: As another poster pointed out, SA was turned off a while ago. Pretty much any GPS unit will be accurate to within 5 meters if it has a decent signal. With tricks like code smoothing, most errors are probably less than 3-4 meters even for a moving receiver.
Sub-meter accuracy: A little bit of position averaging + basic DGPS makes this easy for a stationary receiver, even when SA was on DGPS could cure the intentionally added errors. Very difficult to use with a moving receiver unless combined with an inertial navigation system. (Rare except in modern airplane navigational systems)
Millimeter accuracy: Also possible before SA was turned off, but required the receiver to be stationary for a long period of time, and required significant postprocessing of the data using a variant of DGPS. It still requires stationary receivers for nonmilitary systems.
About the only thing that can't be done without a method for decrypting the P code is sub-centimeter positioning of a moving object. Even with the P code available it can't be done without combining a high-grade inertial navigation system with the GPS system.
retrorocket.o not found, launch anyway?
But vanilla GPS just isn't accurate in urban canyons. In London's docklands area (high buildings) GPS gives accuracy in the 200 to 400m region due to multipath effects. As well as obscuring satellites, tall buildings cause reflections of GPS signals which can cause large errors in the pseudo-range calculations making accurate position reporting very difficult. Places like NYC or Chicago are useless for GPS.
I would imagine that supplementing GPS with other position determining mechanisms (like WiFi) could be beneficial in these circumstances.
The client software running on the tracked device measures the signal strength of the access points, forwards the data to the server which calculates the position. The big-brother scenario is avoided as long as you still have to install the client yourself.
The major drawback of the system is that it needs extensive calibration, since they are using not only the available access points, but also the signal strength of these. Normally they suggest calibration in a 5x5m (15x15ft) grid. More calibration points yield a more accurate result.
And now the piece of information you have all been waiting for: accuracy. With a good calibration this can yield accuracies of arround 1m. In my tests (indoor) the accuracies fluctuate a bit, but is at least better than 3m 95% of the time.
Just as the system described in the original post, Ekahau requires no extra hardware (we already have 2-300 APs on campus).