[mogunus]

The way that it works is using the exact position of the wifi access points relative to the center of the field. The tracker client on the robot downloads the list of these positions from a server. Then, using access-point-specific range and propagation information also contained on the server, it triangulates it's position based on timing differences between the arrival of packets from each access point.

Concievebly, this doesn't have to be a bad thing, with 60 teams all setting up their stuff and running into each other. A bunch of teams' programmers could work together and carpet the area in way more wifi bases than I could alone, and set up a server with their locations relative to the center of the field and propagation characteristics. The accuracy resulting from an effort like that would be very impressive, and actually better than GPS.

WARNING: Now begins the speculation

I'm considering implementing DWPS. The rest of this stuff is mostly coded and working, while this whole section is just theory. DGPS works by setting up one gps station rather close to the mobile one but stationary, of exactly the same kind, and using it to solve the gps equations backwards, to arrive at the timing errors for the satellites being used for triangulation. I could set up a stationary client to run the equations backwards and transmit the results to my mobile station for correction, thus eliminating a lot of error. In GPS applications, this improves the accuracy to about 12" worth of error, using cheaper DGPS hardware.