The red ellipses represent the area where the transmitter probably definitely is. The green dot in the center is where the transmitter is mostly likely to be. The ellipse is what's known as a "covariance ellipse", "confidence ellipse", or "error ellipse". Computing the ellipse really helps save on bandwidth and computing power for the client machine. All the individual interect points together make for a pretty large file. For comparison, here's a screenshot that shows all of the computed intersections (red dots) for a given cluster:
Currently I'm working on being able to add, remove, and modify receivers on the fly in the web interface. With this feature, you'll be able to use it with the normal KSDR software. After that, I'll be adding gpsd support to the KSDR software.
As for my setup, right now I'm working entirely with simulated data. I made a fairly realistic DF simulator which I've primarily used for wireless CTF challenges. I keep it running all the at
dfsim.net. The only thing the software doesn't simulate (and probably can't) is a realistic confidence value.
I have plans to get some real hardware setup soonish. I have one KSDR on my back porch which my soon move to my roof. I also intend to put one more on the tower at my local hackerspace.