A repeatable runout (RRO) is accurately compensated for while moving a magnetic head radially across a disk surface. An iterative learning control algorithm is employed to determine appropriate feed-forward coefficients for an RRO compensation signal for each of a plurality of radial locations across the disk surface. The feed-forward coefficients are determined by performing multiple iterations of continuously moving the magnetic head across the disk surface along a target path while measuring a position error signal that indicates the radial error between the magnetic head and the target path. With each iteration, the iterative learning control algorithm computes new feed-forward coefficients for each of the plurality of radial locations, where the new feed-forward coefficients are selected to reduce the measured position error signal when employed to move the magnetic head along the target path.