In order to achieve robust and rapid track seek of digital video disk recorder(DVDR), alignment between beam axis and objective lens is important as well as exact control of beam spot velocity. The misalignment occurs due to the difference in inertia of two stage actuators, static friction, and model uncertainties at low frequencies. It worsens the measuring accuracies of position and velocity of beam spot, and more over makes the objective lens swing large at the seek end. Conventional way of minimizing the misalignment through feedback control scheme alone is not only insufficient but also hard to tune. In this thesis to eliminate the effects of the repetitive disturbance and model uncertainties, a iterative learning control algorithm has been developed. The algorithm modifies feedforward reference input at each seek trial till it converges. The algorithm utilizing not error itself but a function of it(it cannot be measured directly), gives a convergence condition and remaining error when learning is completed, which are similar to those of one with error. With these conditions, learning algorithm is designed using the nominal plant model. In simulations and experiments used with the conventional feedback controller, the misalignment and the swing at the seek end reduced less than the half. In consequence the seek motion become more stable and the success rate in the track following at the seek end is increased.
