Input-output data based tracking control

Traditional methods of tracking controller design involve inverting the system dynamics using the model of the system. Model errors will result in performance degradation and may lead to instability. This research proposes a tracking control scheme, in which the inverse dynamics of an LTI system are determined using the input-output data. In many practical applications the system is desired to track a set of trajectories that can be characterized by a subspace of L₂. The proposed algorithm utilizes an expansion of the output in terms of the signal obtained in an identification experiment by generating a set of vectors spanning the desired subspace. The performance of the scheme is compared with the performance of the conventional model-based feedforward and repetitive controllers