In this paper, we propose an iterative learning scheme to deal with the periodic off-track errors in the track-following control system for optical disk drives. The periodic errors could be taken into account more effectively by employing an iterative learning algorithm since the errors of the previous period are used to improve the performance of current period. We show a sufficient condition for the convergence of the learning algorithm in the presence of bounded modeling uncertainty. In addition, the effects of the initial state error on the tracking performance are analyzed. Finally, the proposed learning algorithm is demonstrated to be feasible through experiments applying it to the track-following control for an optical disk drive.