what are the states and transistion matrices used in the robot_pose_ekf stack

edit

Does anyone know what state vector, inputs, transistion matrix and measurement matrix is used with the robot_pose_ekf package?

The package fuses data from odometry, IMU and vo. how does this implement the ekf?I've read that you do a prediction step and then a correction step. During which step is each of the data used?

I'm trying to modify the code to incorporate a gyro, but am unable to understand the code.

Thanks!