Seems like a lot of teams (ours included) are giving mecanum a shot this year. Here are a few things I've picked up...
-Basic mecanum is pretty easy to achieve using the built in "holonomic drive" function, which either taken in 3 axis values (x=strafe,y=fwd/rev,z=rotation) or a polar coordinate (magnitude and direction). The derived equations are based on the solution to the Inverse Kinematics Problem (see:
http://www.simbotics.org/files/first...irectional.pdf).
-There are various ways to taken in inputs. One method is a twisting joystick. Another method is using two joysticks--one for x and y translation, and one for rotation. A final method is a joystick and two buttons--the joystick handles translation, and the two buttons handle left and right rotation at constant speed.
-To make mecanum better, use closed-loop control on the robot's header angle and/or wheel speeds. This can be accomplished with a gyro and encoders.
-To make mecanum better better, you can consider driver-centric control. This means that, no matter what the robot's header angle is, pushing the joystick away from you means that the robot moves away from you. This can be best accomplished by using the Polar coordinates implementation of mecanum, and using the gyro angle to modify the angle.
-To make mecanum better better better, consider using steering curves (see:
http://www.chiefdelphi.com/media/papers/2421).