[probizzle]

I've worked with a similar positioning system.
I haven't worked in a situation where a robot arcs, but I have an idea about how you might approach the problem. If you think about it, in 26.2 ms, your robot won't really make that much of an arc. In that time frame, you can assume that your robot has moved in a straight path.

If you take the initial heading (angle) of your robot and the final heading of the robot (before/after 26.2 ms), and take an average, you should get the approximate heading of what the robot is traveling on. I'm not quite sure, but I have a suspicion that this will (along with all the other trig i'm sure you'll figure out) generate pretty accurate results about how much x and y distance you have covered.

And of course, there are better ways to "average" the heading of the robot, by sampling the gyro many times (interrupt) and taking a more significant average.