Quadrupedal locomotion

Okay, this is unexpected.

I mean, I like old technology and all, but why would an electronics geek -- even a Paleotechnologist -- be reading a 127-year-old book on The Horse in Motion (with a study on quadrupedal locomotion?)

For a while now, inspired by some cool YouTube videos, I've been playing with the idea of creating a quadruped robot. The low-level stuff is pretty straightforward; stick a bunch of servos together, connect them with some brackets (3D printers and Google SketchUp are a dangerous combination -- more on this later), and control the positions using pulse-width modulation. No problem.

The hard part turns out to be figuring out what joints to move, where, how, fast, and when -- especially since the cheap S3003 servos I'm using are not the strongest things in the world.

What I really need is a good automated Physics sim (I'm looking into using Havok, but it's nontrivial) that I could hitch to a Genetic Algorithm to figure out what gaits work best. The GA part I could do; tying it all in to Havok or a similar simulation may take a while.