@Daniel:
Long answer:
The map is already created (you can see it via pfodApp->plots), however it's not accurate - That's because odometry drifts heavy over time (think of a counter that counts +5% or -%5 all the time).
The actual robot position has to be corrected and that is something we'll work on soon. The trick is:
1. To reduce the error by combining more sensors (odometry, gyro, compass, gps etc.) - that's typically done using a 'Kalman' filter (TODO).
2. To correct the position using known places (e.g. perimeter hit). But that requires an exact map of the perimeter. Typically point 1 and point 2 is a iterative process (again: Kalman helps here).
Short answer: it's something we should work on soon ;-)
Btw: Latest SVN versions already use odometry to let the robot drive straight when in forward driving state.