My PhD thesis

Beginning September 2001, I am studying full-time for a PhD at the University of Wales, Cardiff.

Basically, I'm studying optimisation techniques related to the design of frequency hopping mobile communications networks.

Mobiles communications networks?

You know, like mobiles phones, with a bunch of phones (mobile stations (MS) in the jargon) all talking to a smaller number of central masts (base transceiver stations (BTS)). It also covers stuff like Bluetooth, although most of my work is based on the cellular model.

Frequency hopping?

Imagine a mobile phone talking to a base station, and the user of the phone is moving around in a city -- in other words, a pretty typical cellular communications case. . If you use just one frequency between the phone and base station, then something called narrowband interference can ruin your signal. Basically, the effects of buildings and whatnot between you and the base station you are sending a signal to tend to knock out small chunks of spectrum at random. If you are trying to send in one of those small chunks of spectrum at the time, your connection vanishes into a black hole.

There's a few ways around this, and frequency hopping is one of them. Basically, the transmitting phone in your hand doesn't stay on a single frequency, it jumps around between a bunch of them every few fractions of a second; then, if narrowband interference does knock a chunk of your signal out, you just wait until the next frequency. On average, that frequency will get through, so you'll only lose signal for a very small period.

However, both transmitter and receiver must be able to jump ahead of time to the same channels, so the list of frequency channels (called a hop set, hop sequence, or mobile allocation list -- the great thing about mobile comms jargon is that there's so much to choose from) must be decided ahead of time. And that's what I'm studying: how to decide what frequencies should be assigned to what transmitters, and in what order they should hop between them, so the mobile network works as good as possible.

Clear?

Optimisation techniques?

Solving the problem posed above is a big, big, biiiig job. We can't do it by just looking at it with a piece of paper. We can't even do it by calculating every possible answer on a computer and just picking the best one, because calculating all the possible answers will take all the computing power on the planet until the sun turns into a small dark cinder. Hardly practical.

So, we are applying optimisation techniques, specifically (for the geeks amongst you, and I imagine all the non-geeks stopped reading ages ago) meta-heuristics. Basically, we don't calculate all the solutions, we just kinda hop around and do a few of them. Then we pick the few and bounce around, looking at similar solutions, on the grounds that next to an OK solution we can find a good solution, and next to that is a better solution again, and so on until we find a solution that, although probably not the best, is pretty damned good.

There's much more to be said, including sexy buzzwords like genetic algorithms and simulated annealing but, really, you either don't want the gory details, or you already know them.