Localization via interference

Posted November 10, 2005

I love it when a paper makes me feel like an idiot. Or rather, does so in a particular way. This paper by some folks at my old alma mater is a perfect example of the phenomenon: just reading the abstract is enough to get you to nod your head, "Yes, obviously that should work," and then slap yourself, "... but why didn't I think of that?!?"

Specifically, they describe a method for little wireless sensor nodes to collectively determine their relative locations. This is a highly useful thing to be able to do, and a lot of bright people (plus me) have thought about it a lot, coming up with methods like acoustic time of flight and triangulation from 802.11 access point signal strengths, etc. The trick is that doing it well really requires some electromagnetic black magic, since ideally you'd want to use radio time of flight, which means sub-nanosecond precision to get a location to the nearest few inches.

And then the Vanderbilt guys figured out that the whole way of thinking about the problem is wrong... don't try to figure out the distance between two nodes using those two nodes. Instead, use four nodes as your analysis unit: two of them transmit on slightly different frequencies, and the other two detect the phase difference in the much slower beat frequency that you see from the combined signal. Then you just need millisecond resolution or so, which is trivial to do on a microcontroller.

Granted, at this point, you don't have a distance, exactly, but just some information about it: some linear combination of distances modulo the carrier wavelength. But now it's just a question of software: do the experiment with several different carrier frequencies and different nodes playing the roles of transmitters and receivers, and you can eventually build up the full picture of all the locations in 3D space.

Awesome. So awesome in fact that I'll forgive them for making me feel like an idiot for not thinking of it first. Heh heh.